Chapter 27: Marketing on a Restaurant Budget: Social Media, Reviews, Local SEO, Events, and the \$0 Marketing Plan

"You don't have a marketing budget. You have a room, a product, twenty-two people, and a neighborhood. Three of those four are already paid for." — constructed; the answer every operator eventually gives to the agency that just called

Overview

A consultant will tell you that you need a brand strategy. A platform will tell you that you need to post daily. An agency will quote you three thousand dollars a month, which is a hundred and fifty-five percent of what you have to spend on marketing for the entire year.

Here is what you actually have. Twenty-three thousand two hundred fifty dollars. That is 1.5% of Bellwether's \$1,550,000 first-year revenue plan, and it sits inside the \$217,000 other-operating line, which also has to pay for repairs, smallwares, supplies, utilities, and credit-card fees. Every dollar you add to marketing comes out of something that breaks.

And here is what that money is required to produce. Not awareness. Not followers. Not a brand "presence." Covers. Thirty-six thousand one hundred forty of them, which Chapter 2 already decomposed into the only form that matters: 9,035 distinct guests, each coming four times a year. That decomposition is the whole chapter. It says that marketing has two entirely different jobs with entirely different prices — finding a guest, and bringing a guest back — and Chapter 23 already priced them. Moving your guests from four visits a year to four and a half is worth \$70,538. Buying the same money by acquiring more guests costs \$98,223 and does not pay back for three years. Your entire annual budget is \$23,250. One of those numbers is three times your budget. The other is more than four times it, and it comes with a three-year wait.

That asymmetry is not a nuance. It is the plan. Most restaurant marketing — almost all of it, in my experience, and certainly all of the marketing anyone will try to sell you — is built around reach: more people, more impressions, more new faces. And the arithmetic of a neighborhood restaurant says that the cheapest cover you will ever sell is the fourth one to somebody who already likes you.

So this chapter starts with what a cover is allowed to cost, spends the middle on the free foundation almost nobody maintains, and ends on measurement — comp codes, a question at the host stand, and an honest reckoning of what the \$0 plan actually costs, which is hours.

In this chapter, you will learn to:

• Compute cost per cover acquired, know which costs belong in the numerator, and defend the denominator against the flattering version of itself.
• Fix and maintain the free local-search surfaces — starting with the previous tenant's listing, which is sitting on your address right now.
• Solicit and respond to reviews inside the platform rules, and use the offset identity from Chapter 23 instead of the revenue-elasticity statistic everyone quotes and nobody sourced.
• Run a social cadence in seventy-five minutes a week and say plainly what it does and does not do.
• Build an email and SMS list from day one, legally, and price what it is worth.
• Pitch local press, and understand why a write-up does nothing at all to your cost structure.
• Design a repeat-visit mechanic and compute the incrementality it must hit to break even.
• Show up in your neighborhood as a business that took something, which is what you are.

Learning Paths

🏗️ Opening — all of it, and §27.2 and the pre-opening timeline in the checkpoint are the parts with deadlines attached. The Google Business Profile work happens ten weeks out, not the week you open, and the email list starts before you have anything to send. 📋 Managing — weight §27.3, §27.7, and §27.9. If you manage someone else's restaurant, review response and the repeat-visit mechanic are the two levers you control without a budget conversation. 🍸 Beverage — §27.4 and §27.8 are yours. A bar program is the most photographable and most event-ready thing in the building, and the industry-night partnership in §27.8 is a beverage play before it is a marketing one. 🚚 Small Format — §27.2 matters more to you: a truck's location data is its entire storefront, and a wrong pin is a dead service. §27.5's list is the only asset a truck carries between locations.

27.1 What marketing must produce: covers, at a cost per cover you can afford

Start with the output, because restaurant marketing goes wrong at the very first step, when somebody substitutes an intermediate measure — reach, impressions, followers, "engagement" — for the thing the business needs.

The thing the business needs is 36,140 covers.

Write the decomposition down where you can see it:

FIGURE 27.1 — What marketing is actually buying                        [the Bellwether plan]

  YEAR-ONE COVERS REQUIRED                                              36,140
    ├─ first visits    (9,035 guests × 1)   ████████                     9,035    25%
    └─ return visits   (9,035 guests × 3)   ████████████████████████    27,105    75%

  Marketing budget                                                     $23,250
    cost per first-visit cover     $23,250 ÷ 9,035    =                   $2.57
    cost per return-visit cover    $0      ÷ 27,105   =                   $0.00

  Contribution per cover (Ch. 23)                                        $18.40
  Guest value, first year          $18.40 × 4                           $73.60
  Guest lifetime value             $73.60 × 3 years                    $220.80

Three-quarters of the covers in the plan are return visits, and no marketing dollar buys them. They are bought by the food, the room, the pacing, and whether the server noticed that the woman at table 12 was celebrating something. Marketing buys the first cover. Hospitality buys the other three. That sentence is the entire relationship between this chapter and Chapter 23, and it is why a marketing plan that promises to solve a repeat-visit problem is selling you something.

Defining the number

Cost per cover acquired is the marketing cost of producing one incremental cover — one cover that would not have happened without the spend.

$$\text{Cost per cover acquired} = \frac{\text{media} + \text{production} + \text{offer cost} + \text{purchased labor}}{\text{incremental covers, evidenced}}$$

Both halves of that fraction are where operators lie to themselves, so both get rules.

Numerator rules. Include the media (the ad spend, the postage, the platform fee). Include the production (the photography, the printing, the website). Include the offer cost — if the campaign carried ten dollars off, the ten dollars is a marketing expense that happens to land on the comp line of your P&L instead of the marketing line. Include labor you actually bought; if you paid somebody to run social, that is media cost by another name. Exclude only the labor you did not buy — the hours the front-of-house partner puts in — but do not pretend those hours are free. We will price them at the end of the chapter.

Denominator rules. Count only incremental covers: covers that would not have happened anyway. A redemption is not automatically an acquisition. If you hand a ten-dollar card to somebody standing at your host stand, they were already standing at your host stand. Chapter 28 formalizes this problem for delivery, where it is worth six figures; here it is worth the difference between a channel you keep and a channel you have been subsidizing for a year.

What a cover is allowed to cost

Three ceilings, and choosing among them is a cash decision more than a marketing one.

Marketing people love the third column. Restaurant operators should be extremely careful with it. Spending against a lifetime value you have not earned yet means paying cash today for margin that arrives in 2029, in a business Chapter 1 told you dies of cash timing more often than of unprofitability. A restaurant that spends \$180 to acquire a guest worth \$220.80 has done sound marketing arithmetic and may still miss payroll in February. The first-year ceiling of \$73.60 is the one an independent should actually manage to, and the \$18.40 is the one that lets you sleep.

🧮 Run the Numbers

Cost per cover acquired, four honest ways.

The budget is \$23,250. Here is the same money divided four ways, from the most flattering to the most defensible.

1. The blunt version. Total marketing over total covers.

$\$23{,}250 \div 36{,}140 = \$0.64$ per cover.

Sixty-four cents against \$18.40 of contribution — marketing costs 3.5% of what a cover contributes. It is a true number and it is nearly useless, because it takes credit for 27,105 covers marketing did not produce.

2. The plan version. Total marketing over first visits only.

$\$23{,}250 \div 9{,}035 = \$2.57$ per cover acquired.

Still flattering — it assumes marketing gets credit for all 9,035 first visits, including the man who lives above the bakery and walked past for six months.

3. The attribution-adjusted version. Assume marketing can defensibly claim half the first visits; the rest are proximity, word of mouth, and luck.

$\$23{,}250 \div 4{,}517 = \$5.15$ per cover acquired. At one-third: $\$23{,}250 \div 3{,}012 = \$7.72$.

4. The version with labor in it. Add the front-of-house partner's five hours a week at a replacement cost of \$25/hour: $260 \times \$25 = \$6{,}500$.

$(\$23{,}250 + \$6{,}500) \div 9{,}035 = \$3.29$. At half attribution: $\$29{,}750 \div 4{,}517 = \$6.59$.

Every one of those numbers is under \$18.40. Read that carefully, because it is the chapter's most important finding and it is not the one you expected. At this budget, on this plan, acquisition is not the risk. Take the harshest version available — one-third attribution and labor at replacement cost, $\$29{,}750 \div 3{,}012 = \$9.88$ — and a first cover still costs less than ten dollars against \$18.40 of contribution on that single visit.

The risk is the fourth visit. Bellwether's plan does not need a cheaper cover. It needs 9,035 people to come back three more times, and no line of this budget can force that.

The budget is a constraint, not a strategy

One more framing before we spend a dollar. Total contribution on the plan's covers is $36{,}140 \times \$18.40 = \$664{,}976$. The marketing budget is 3.5% of it. That is a small, sane number. It is also, and this matters, 10.7% of the \$217,000 other-operating line ($\$23{,}250 \div \$217{,}000$), which is where the money actually comes from.

⚠️ Where the Money Leaks

A percentage of sales is a budget, not a plan.

"Marketing should be 2–4% of sales" is the most commonly repeated budgeting rule in this industry and it fails in two specific, expensive ways.

It moves the wrong direction. A percentage-of-sales budget shrinks when sales shrink. February is soft, so the budget cuts itself in February — which is precisely the month you needed it. A percentage rule is procyclical, and a procyclical marketing budget amplifies exactly the swing you were trying to smooth. Chapter 33's cash forecast will tell you the same thing about every other discretionary line.

It says nothing about what you are buying. Two operators can both spend 1.5% of sales. One spends it on 240 traceable first visits and a list of 2,400 email addresses. The other spends it on a retainer, a logo refresh, and boosted posts, and at the end of the year cannot tell you a single cover it produced. The percentage is identical. The businesses are not.

The disciplined order of operations: start from the covers you need, decide what a cover is allowed to cost, build the plan that produces those covers, then check the total against the percentage as a sanity test. If your plan needs 4% of sales, you do not have a marketing problem. You have a concept problem, and Chapter 2 is where it lives.

27.2 The free foundation: Google Business Profile, hours, photos, and local search

Before you spend anything, spend nothing, correctly.

Local SEO — search engine optimization for geographically qualified queries — is the practice of making a business findable when somebody searches for a category near a place: restaurants near me, dinner downtown, where to eat Rivermill. For a restaurant this is not a specialist discipline with a monthly retainer attached; it is a maintenance task. Google states publicly that its local results are driven principally by relevance, distance, and prominence. Two of those you cannot move — distance is your lease, prominence accumulates over years. Relevance is the one you control, and it is mostly data entry.

Organic vs. paid is the distinction that organizes every channel in this chapter. Organic placement is earned: you appear because a platform's ranking decided you were the right answer, and you paid nothing for the placement. Paid placement is bought: you appear because you bid. The distinction matters commercially because organic placement has no marginal cost and no off switch — it keeps working when the budget is spent, and it degrades quietly when nobody maintains it.

The Google Business Profile

A Google Business Profile is the free listing a business claims and controls, which populates the panel that appears in Google Search and Google Maps: name, category, address, phone, hours, website and menu links, photos, attributes, questions and answers, posts, and — the part nobody controls — the review average and the reviews themselves.

For a neighborhood restaurant this listing is almost certainly the most-viewed page your business has, and it is not on your website. People do not navigate to restaurant websites. They search a category, look at a panel, look at the average, look at four photos, and either tap directions or tap the next result. The decision takes fifteen seconds and happens on a surface you did not design.

Here is what has to be right, and the specific way each one fails:

Two of those deserve their own paragraph.

Hours are the highest-value field in restaurant marketing, and it is not close. A guest who drives to a closed restaurant does not reschedule. They eat somewhere else, a meaningful fraction never try you again, and some of them leave a one-star review about it — which §27.3 will show you is expensive to undo. Every holiday, every private buyout that closes the room, every week you shift brunch: the special-hours field, in advance. Put twelve dates on a calendar in January and handle them on the first Monday of each month.

The previous tenant. Bellwether is a second-generation space — it was a café. That café has a Google Business Profile sitting on your address right now, with its own reviews, photos, "popular times" graph, and quite possibly its own hours. If nobody deals with it, guests searching your address find a business that closed, and some map surfaces will merge or confuse the two. Claim the old listing if you can, get it marked permanently closed or properly converted, and watch for duplicates for six months. Every operator moving into a second-generation space has this problem and roughly half never notice it.

The rest of the free surfaces

Beyond Google, in rough order of what they are worth to a mid-size-metro American restaurant: Apple Business Connect (free, feeds Apple Maps and Siri, fifteen minutes); Bing Places (free, five minutes, feeds some voice and in-car systems); Yelp — you do not control whether a page exists, only whether you have claimed it and whether the information is right, so claim it; and your reservation platform's own directory, which has photos, hours, and description fields of its own. Then NAP consistency — name, address, phone, identical everywhere, because data aggregators feed each other and an old suite number propagates for years.

And your own website, which for a restaurant needs to do exactly five things on a phone in under ten seconds: what you are, where you are, when you are open, the menu as readable text, and a reservation link. Everything else is decoration. Chapter 3 owns how it should look; this chapter only insists that it load.

🧾 Read the Numbers

```text FIGURE 27.2 — "The profile is the front door" [constructed teaching example] THE ARTIFACT Monthly performance summary from a Google Business Profile, October, month 7 of operation. The free report the platform generates. THE CONTEXT A 68-seat neighborhood restaurant in a gentrifying warehouse district, dinner Tuesday–Saturday plus weekend brunch. 3,190 covers recorded on the POS that month. Profile claimed and complete since ten weeks before opening; 169 reviews at a 4.58 average.

                 HOW PEOPLE FOUND YOU
                   searched the business by name (direct)            1,170
                   searched a category or dish (discovery)           3,010
                   TOTAL SEARCHES                                    4,180

                 WHAT THEY DID
                   profile views                                     3,842
                   website clicks                                      612
                   direction requests                                  488
                   phone calls                                         197
                   menu / reservation link taps                        343
                   TOTAL RECORDED ACTIONS                            1,640

                 action rate on views          1,640 ÷ 3,842  =       42.7%
                 discovery share of searches   3,010 ÷ 4,180  =       72.0%

WHAT IT SHOWS Seventy-two percent of the searches that reached this profile were people looking for a category, not for this restaurant — which is the definition of a discovery channel and the reason the free fields are worth maintaining. Of 3,842 views, 1,640 produced an action. The 488 direction requests are the closest thing here to intent to visit: if half of them convert and a converting party averages 2.5 guests, that is 610 covers, about 19% of the month. WHAT IT DOESN'T There is no cover in this report. Not one. It cannot tell you whether a single one of those 488 people walked in, whether they would have come anyway, or — the expensive question — how many of the 3,010 discovery searchers saw this panel and chose the place two blocks over. "Direction requests" also includes staff, delivery drivers, and the guest who requests directions three times. The 2.5-guest party size is an assumption, not a measurement; the POS knows the real number. THE DECISION Monday: enter special hours for both holiday weeks now, before the searches happen. Add the fall menu as a post and six photographs including one exterior-at-night. Start the host-stand question — "may I ask how you found us?" — for thirty days, so that next month's version of this report can be reconciled against covers instead of admired. THE LESSON The most-viewed page your restaurant has is one you did not build and cannot fully control. The discipline it rewards is accuracy, maintained monthly — not cleverness, and not a retainer. ```

The audit itself is ninety minutes once, then twenty minutes on the first Monday of the month forever: open the profile as a stranger sees it, check every field in the table above, read the Q&A, look at what photos guests have uploaded (they outnumber yours and they are worse), and confirm the menu link still resolves. It is the highest-return twenty minutes in this chapter and the first thing every busy operator stops doing.

27.3 Reviews as marketing: soliciting, responding, and the one-star swing

You will be told, confidently and often, that one star of average rating is worth some specific percentage of revenue. Chapter 23 refused that claim and this chapter refuses it again, for the reason Chapter 1's first case study taught: the figure travels without a source. It appears in trade articles citing consultancies citing studies of a different industry in a different decade with a different definition of revenue. It may well be directionally true — a better average almost certainly sells more dinners — but "directionally true" and "a number you can put in a plan" are not the same thing, and a book that opened by dismantling the 90% failure myth cannot turn around and quote an elasticity it did not verify.

What we can do instead is arithmetic, which needs no source because it is a definition.

The offset identity

To hold an average of $A$ after receiving one one-star review, you need $x$ five-star reviews, where:

$$x = \frac{A - 1}{5 - A}$$

That is not an estimate. It falls straight out of the definition of an average, and it has a property worth pausing on: it does not depend on how many reviews you have. Whether you have 40 or 4,000, undoing one one-star at a 4.6 average takes nine five-stars.

FIGURE 27.3 — Five-star reviews needed to offset one one-star     [arithmetic identity, from Ch. 23]

   average                x = (A − 1) ÷ (5 − A)
   you hold
   4.0    ███                                            3.0
   4.2    ████                                           4.0
   4.4    ██████                                         5.7
   4.5    ███████                                        7.0
   4.6    █████████                                      9.0
   4.7    ████████████                                  12.3
   4.8    ███████████████████                           19.0
   4.9    ███████████████████████████████████████       39.0

   Note the shape: the cost of a bad night is not linear in your average.
   It accelerates. The better you are, the more expensive one bad night becomes.

Two consequences fall out of that curve, and both are counter-intuitive enough that most operators have never worked them through.

A four-star review is a downgrade at any average above 4.0. This is arithmetically obvious and emotionally very hard to accept. A guest who had a perfectly nice evening, wrote three warm sentences, and gave you four stars has lowered your average — and a guest who had a perfectly nice evening is the most common guest you have. It does not mean you should chase five stars or, God forbid, coach guests toward a number. It means you should stop treating four-star reviews as good news in your reporting, and it means the volume of genuinely delighted guests is what defends the average, not the absence of bad ones.

Your exposure is highest when you are youngest. The identity is count-independent, but the drop is not:

(Worked at 12: $(12 \times 4.6 + 1) \div 13 = 56.2 \div 13 = 4.32$; rounded from a starting average carried to two places.)

A restaurant in month two, with a dozen reviews, can lose three-tenths of a star to one guest who waited fifty minutes for a table on the second Friday it ever opened. A restaurant with five hundred reviews barely moves. The first hundred reviews are the fragile period, which is an argument for building volume deliberately and early — and Bellwether's two pre-opening neighborhood preview services in §27.8 exist partly for this reason.

🧮 Run the Numbers

What it takes to climb, and what it takes to fall.

Bellwether ends month six with 160 reviews at a 4.55 average — 728 rating points.

The climb. The chef-owner wants a 4.7 before the winter. How many consecutive five-stars?

$\dfrac{728 + 5n}{160 + n} = 4.70 \;\Rightarrow\; 728 + 5n = 752 + 4.7n \;\Rightarrow\; 0.3n = 24 \;\Rightarrow\; n = 80$

Eighty consecutive five-star reviews, with no exceptions. At the rate a 68-seat restaurant actually accumulates public reviews, that is most of a year of unbroken excellence.

The fall. Now suppose a Saturday goes badly — a no-show cook, a forty-minute ticket time — and produces four one-stars in a week.

$\dfrac{728 + 4}{164} = \dfrac{732}{164} = 4.46$

Nine-hundredths of a star, gone in one night. To get back to 4.55 from there:

$\dfrac{732 + 5n}{164 + n} = 4.55 \;\Rightarrow\; 732 + 5n = 746.2 + 4.55n \;\Rightarrow\; 0.45n = 14.2 \;\Rightarrow\; n = 32$

Thirty-two five-stars to undo one Saturday. That is the number to put in front of a kitchen team, and it is a far better argument for the staffing decision in Chapter 19 than any speech about standards. The asymmetry is the lesson: reviews are cheap to lose and expensive to buy back, and the mechanism that produces them is operational, not promotional.

(Note what this does not say. It does not say four one-stars cost you \$X of revenue. Nobody credible knows that number for your restaurant. It says what it takes to undo them, which is actionable and true.)

Soliciting reviews without breaking the rules

The single biggest determinant of your average is who gets asked. Restaurants that never ask are reviewed disproportionately by the aggrieved, because a bad night generates more motivation than a good one. Restaurants that ask everyone, consistently, get an average that looks like their actual service.

The rules, which are real and which change:

• Ask everyone or ask no one. Asking only the guests you think are happy is called review gating; platform policies generally prohibit it, it is detectable, and the penalty — filtered reviews, or a suspended listing — falls entirely on you.
• Never pay for a review, trade anything of value for one, or write one. Staff reviews are fraud. Purchased reviews are fraud.
• Yelp is different from Google. Yelp's stated position discourages soliciting reviews at all, and its recommendation software filters reviews it judges unnatural — frequently the ones you solicited. So: a general, non-gated ask pointing at Google, and on Yelp, a claimed page with correct information and responses.
• The ask that works is not a card. It is a sentence from the person who served them, at the moment the check is dropped, on a night that went well, said the same way to every table: "If you had a good time, a review helps us more than you'd think." Then a QR code on the check presenter for the people who mean it. No incentive, no promise, no follow-up.

⚖️ Code and Compliance

Reviews, endorsements, and the things that get restaurants in trouble.

• Fake and incentivized reviews. The Federal Trade Commission enforces against deceptive endorsements and has rules addressing fake reviews and testimonials; platforms independently prohibit them and remove listings. There is no version of buying reviews worth the exposure.
• Material connections must be disclosed. If you comp a meal in exchange for coverage, the person covering you has a disclosure obligation and you have an interest in it being met. Put it in writing (see §27.6's comp and influencer policy). Employee and family reviews are not a gray area either; both happen constantly and both are detectable.
• Never disclose a guest's information in a public response. Confirming that a named reviewer dined on a particular night, with a particular party, is a bad instinct and in some contexts a legally consequential one.
• Defamation is a bad tool. Operators occasionally threaten reviewers. Beyond the near-certain reputational cost, several states have laws limiting suits that target public statements, and the Consumer Review Fairness Act restricts contract terms that penalize consumers for honest reviews.

All of this varies by state and platform, and platform policy changes far faster than law. Read the current policy of any platform you solicit on, and take anything consequential to an attorney.

Responding

Review response is the public reply an operator posts to a review. Its purpose is almost universally misunderstood.

You are not writing to the reviewer. The reviewer has already had the experience, formed the opinion, and published it; your reply will rarely change any of that. You are writing to the next person who reads the review — a stranger deciding between you and two other places, who is scrolling the one-stars specifically because that is what careful people do. That reader is evaluating one thing: does this business seem like it would take care of me if something went wrong?

Which produces a short set of rules. Respond to all of them, or to a defined subset, consistently — responding only to the bad ones makes the page read like a grievance log. Keep it to three sentences; a long response reads as defensive no matter how reasonable it is. Acknowledge the specific thing — "fifty minutes is too long for a table you booked" beats "we're sorry you didn't have a great experience," which is the sentence everyone writes and nobody believes. Never litigate the facts in public, even when you are right and have the ticket times; the next reader cannot adjudicate it and will only notice that you argued. Move it offline exactly once, with a real email address, and then stop. And answer inside 48 hours — which for a Tuesday-to-Saturday operation means Monday, in the same thirty minutes as the profile audit.

🤝 Hospitality

The recovery that never reaches the internet.

Chapter 23 made the case that service recovery is where hospitality earns its keep. Here is the marketing consequence, in arithmetic you now have.

A table waits fifty minutes past their reservation. There are two futures. In the first, nobody notices until they leave, and one of them writes a one-star review at 11:40 p.m. from a rideshare. At a 4.6 average, undoing that costs nine five-stars — call it a month of good nights.

In the second, the manager sees the table at minute twenty, goes over before they have to complain, names the problem out loud, buys the first round, and re-fires their appetizers to come out fast. Cost: maybe \$22 of product. No review is written, because there is nothing left to write about.

Twenty-two dollars, or nine five-star reviews. That is the actual exchange rate, and it is why the manager on the floor has more control over your review average than anyone with a marketing title. The best review-management program in a restaurant is a manager who touches tables at minute twenty.

27.4 Social media that isn't a second job: cadence, content, and what actually converts

Two claims, made plainly, before anything else.

First: social media is real and worth doing. A neighborhood restaurant's guests are on it, they look at pictures of food before choosing where to eat, and a well-maintained account resolves the half-dozen questions a person has before booking. It is also, quietly, the best recruiting channel a small restaurant has — cooks look at your account to decide whether they want to work for you, and Chapter 17's cost-of-turnover arithmetic makes that worth more than most of what you post.

Second: almost every specific number you will be told about it is unsourced. Optimal posting times, benchmark engagement rates, the share of diners who choose a restaurant based on Instagram, how many impressions a video gets versus a photo — these circulate as facts, get quoted in decks, and trace back to a vendor's blog post about a sample nobody described. This book will not give you engagement benchmarks, because it does not have credible ones. What it will give you is a cadence you can sustain and a way to find out what works in your room.

What social can and cannot do

The honest framing: social media is a consideration channel, not an acquisition channel. Followers are not covers. A restaurant with 8,000 followers and a wrong phone number on its Google profile is losing money in a way it cannot see.

A cadence that survives a Friday

Content cadence is the sustainable publishing rhythm a business commits to — how often, in what formats, produced by whom, on what day. The word that matters is sustainable. Every restaurant social account I have ever watched fail, failed the same way: three weeks of daily posting by an enthusiastic owner, then a bad Friday, then nothing for two months, then an apology post. The apology post is the tombstone.

FIGURE 27.4 — A content week that survives a Friday                    [the Bellwether plan]

  DAY   STATUS      TIME    WHAT HAPPENS
  MON   closed      30 min  batch shoot: 6 photos, 2 short clips, whoever is in the building
                            schedule the week; answer every review from the weekend
  TUE   service     10 min  post 1 — the dish that is on tonight, with the price on it
  WED   service      5 min  story only: prep, the hearth lighting, tonight's special
  THU   service     10 min  post 2 — the room at 6:15, before the rush, lights right
  FRI   service      5 min  story only: what's left tonight / bar seats are walk-in
  SAT   service      5 min  story only
  SUN   brunch      10 min  post 3 — brunch, and one reason to come back next week
  ───────────────────────────────────────────────────────────────────────────────────
  TOTAL                     75 min/week  ≈ 65 hours/year  ≈ $1,625 at $25/hr replacement
                            — and none of it appears anywhere in the $23,250

Three posts a week and daily stories, in seventy-five minutes, by a person who is in the building anyway. Note the structure: the only real production happens Monday, when the restaurant is closed. Everything during service is five to ten minutes on a phone. A cadence that requires creative energy during a Friday dinner rush is a cadence that ends in February.

The photography days (\$1,500 pre-opening plus \$1,200 in refreshes) are what keep this from looking like a phone. You get a bank of usable images per session — food, room, hearth, hands, staff, exterior at dusk — and spend them over three months. Buy the usage rights outright. A photographer's standard license may not cover the paid ads in §27.9 or a press outlet's reuse, and discovering that in month eight is an annoying way to lose \$1,500.

What actually converts, as best anyone can tell

Without benchmarks, here is what we can defend from mechanism rather than from statistics. A person looking at your account is, almost always, either deciding whether to come or deciding when. So: a photograph of the specific thing they would order, lit honestly, with the price on it — the Hearth Chicken on a plate at \$29, not a moody shot of a hand pouring oil. The room, at the hour they would arrive — how loud, how dressy, how dark, whether the bar has stools; the single most useful and least posted content in restaurant social media. A reason and a date ("the last week of the corn"; "bar seats open at 5:30 on Thursdays"), which is urgency you did not manufacture. Faces, because guests choose restaurants partly on whether the people seem like people, and because staff posts do double duty for recruiting. And operational truth — closed Saturday for a private event, patio open, power out — which is what people actually need and costs nothing.

And the failure mode worth naming: posting into a channel while the room is empty on Tuesday. If Tuesday runs 62 covers against a 68-seat room, the marketing question is not "what should we post" — it is "what would make somebody choose a Tuesday," which is a menu, pricing, and programming question that Chapter 24 owns and §27.8 pays for. Social can carry the answer. It cannot be the answer.

27.5 Owned channels: email, SMS, and the list you should have started on day one

Everything in §27.2 through §27.4 runs on land you rent. A platform can change its ranking, its distribution, its policies, or its pricing, and every one of those changes has cost restaurants real traffic with no notice and no recourse.

Email and SMS lists are the exception: direct, permission-based channels a business owns outright, where you decide who receives what and when and no intermediary sits between you and the guest. They are the only marketing asset here that appreciates, and the only one that transfers if you ever sell.

Start the list before you have anything to send. Bellwether's plan opens email capture in week −10, because a list built from opening day is worth several times a list started in month nine when somebody notices sales are soft.

Building it

The list gets built at the table, not on the internet. Sources, in descending order of what they are actually worth:

Target for year one: 2,400 email addresses, about 27% of the 9,035 guests. SMS is a much higher consent bar and a much smaller list — target 600, and treat every one of them as expensive, because they are.

The four sentences a server needs, which are worth training as carefully as a wine list:

  1. "Are you on our list? It's once a month — menu changes and the things that sell out."
  2. "It's not a lot of email. I'd tell you if it was."
  3. [scan or hand the presenter]  "It's the code on there."
  4. "No problem at all."          ← the most important one; the ask must be free to decline

What to send, and how often

Once a month, for email. That is not timidity — it is the frequency you can maintain in December without hating yourself. Four short paragraphs: what changed on the menu, what is about to disappear, one thing that is happening, and one reservation link. No newsletter design, no header image, no "a note from the chef" unless the chef actually wants to write one every single month, which they do not.

SMS is different in kind. It is immediate, nearly always read, and intrusive — which makes it right for exactly two jobs: short-notice availability ("six bar seats tonight, walk in") and something genuinely time-bound. Two to four messages a quarter. A restaurant that texts weekly will watch its list unsubscribe itself, and each unsubscribe is a permission you cannot get back.

🧮 Run the Numbers

What the list is worth, and how you would know.

The email and SMS platform costs **\$1,080** a year — \$90 a month for a list of this size, which is the whole of the owned-channel budget.

At month twelve the list is 2,400 addresses. Suppose the monthly note moves 15% of them to make one additional visit over the course of a year:

$2{,}400 \times 0.15 = 360$ additional covers

$360 \times \$18.40 = \$6{,}624$ of contribution

Against \$1,080 of cost, that is a 6.1× return and the cheapest contribution in this chapter.

Now the honest part. The 15% is an assumption. It is not a finding, it is not a benchmark, and I did not get it from anywhere — I chose it because it is a modest number that makes the channel worth funding. The number that matters is the one you measure, and you measure it the way §27.9 describes: put a code in the email, count the redemptions, and check the visit count of the people who redeemed.

Run the break-even backward and you can see how little the channel has to do. To cover \$1,080 you need $\$1{,}080 \div \$18.40 = 59$ incremental covers a year. Fifty-nine covers. Out of 36,140. This is not a close call, and it is why the list is the first thing you build and the last thing you cut.

⚖️ Code and Compliance

Email and text messages are regulated, and text messages are regulated hard.

Email — the CAN-SPAM Act. Commercial email must use accurate header and sender information and a non-deceptive subject line, identify itself as an advertisement where applicable, include a valid physical postal address, and offer a clear opt-out honored promptly (the statute's window is commonly described as ten business days). You may not sell or transfer the addresses of people who opted out. Penalties are assessed per message.

SMS — the TCPA and its state analogues. The Telephone Consumer Protection Act and the FCC rules under it generally require prior express written consent for marketing text messages, with clear disclosure of what the person is agreeing to, and require you to honor STOP immediately. Statutory damages are per message, and several states have their own statutes with private rights of action. A restaurant that texts 600 people without documented consent has created an exposure that dwarfs its entire marketing budget. Carrier requirements (campaign registration, content rules) are contractual rather than statutory but will get your messages blocked just as effectively.

What a working operator actually does: a separate, unchecked consent box for SMS — never bundled with the email box, never pre-checked — plain language about what they will receive and how often, and a stored record of when and how each consent was given. Keep the record even for people who leave; the record is the defense.

This varies by state and the rules move. Verify locally, and have counsel look at your consent language before you send the first message.

🔍 Check Your Understanding

1. Bellwether's blunt cost per cover is \$0.64 and its cost per cover *acquired* is \$2.57. Why is the second number the useful one, and why is it still too flattering?
2. Your review average is 4.8 and you receive a four-star review from a guest who wrote three warm sentences. What happened to your average, and what — if anything — should you do about it?
3. Your email list is 900 addresses and the platform costs \$1,080 a year. How many incremental covers must it produce to break even, and what would you have to do to find out whether it did?

(1: The blunt number takes credit for 27,105 return covers that marketing did not buy; the second counts only first visits — but it still assumes every first visit was marketing's doing, when many were proximity and word of mouth. 2: It went down; a four-star review lowers any average above 4.0. You do nothing about that review — you build volume of genuinely delighted guests, because the defense of a high average is more reviews, not fewer four-stars. 3: $\$1{,}080 \div \$18.40 = 59$ covers, regardless of list size. To find out: put a redemption code in the email and check the prior visit count of everyone who used it.)

27.6 Earned media: pitching, press, and what a write-up is worth

Earned media is coverage you did not pay for — a newspaper review, a "best new restaurants" list, a television segment, a food writer's newsletter, a mention on a local blog. It is called earned because you cannot buy it, which is both its value and its limitation: you cannot schedule it, size it, or repeat it.

Chapter 1 said the thing that matters here, and it bears restating because this is the section where operators most want to believe otherwise. Press is a demand-side event. It does nothing at all to your cost structure. A write-up fills your room; it does not change your food cost, your labor, your rent, or your prime cost by a single basis point. A restaurant with a broken prime cost and excellent press dies with a full dining room, which is disorienting to watch and happens constantly. The restaurant in Figure 1.3 was written up twice and closed in March.

So the correct posture toward press is: pursue it, be glad of it, and never let it change a decision you would not otherwise have made.

Pitching, for a restaurant with no publicist

At \$23,250 a year you do not have a PR agency and should not want one. What you have is a neighborhood, a specific angle, and the ability to write six different emails instead of one.

• Local first, and local means local. The metro daily's food writer, the alt-weekly, the city magazine, the regional food newsletter, the neighborhood association's mailing list. A mention in the neighborhood newsletter produces more covers for a 68-seat restaurant than a national listicle, and it is a hundred times easier to get.
• Lead times are long, and they are what amateurs get wrong. A monthly magazine's openings section may close six to ten weeks ahead. Week −4 is the last week a pre-opening pitch can work, which is why it is on the timeline in the Business Plan checkpoint and not on a to-do list.
• One angle per outlet, and the angle is not "we are opening." Everyone is opening. The hearth and what it changes about how food tastes. A warehouse district's tenth restaurant and what that says about the neighborhood. A chef leaving a fourteen-year kitchen career to bet their savings.
• Make it easy. Three paragraphs, the date, the address, and a link to six high-resolution photos with usage rights already cleared — a writer on deadline runs the piece that has a usable photo. Then follow up exactly once, a week later, and let it go.

What a write-up is actually worth

Honestly: it depends, it decays, and the only durable value is the guests it brought who came back. Whatever the size of the spike, attention normalizes within weeks, so the question that determines whether the coverage was worth anything is: of the people who came because of it, what fraction returned? That is measurable — the host-stand question plus the visit count in §27.9 — and it is the one nobody tracks, because press feels like an achievement rather than a channel.

There is also a real risk, and it is why this section carries a warning rather than a celebration.

⚠️ Where the Money Leaks

The write-up that cost \$4,839.

A modeled Saturday, and the arithmetic is unpleasant.

Bellwether plans 123 covers on a Saturday. A well-placed local review lands on Thursday and Saturday walks in at 160 covers on a room, a kitchen, and a schedule built for 123.

The upside. 37 extra covers × \$18.40 = **\$681** of additional contribution. Real money.

The downside. Ticket times run 45 minutes. Two stations are in the weeds from 7:00 to 9:30. Most of that room is dining with you for the first time — that is what a press spike is — so the first impression a hundred-odd new guests form is of a restaurant that could not keep up.

Suppose the night converts 25 fewer guests into repeat guests than a normally paced Saturday would have. At a guest lifetime value of \$220.80:

$25 \times \$220.80 = \$5{,}520$ of value that will not now exist.

Net: \$681 − \$5,520 = −\$4,839. Plus, quite likely, a handful of one- and two-star reviews written by people who waited, each of which §27.3 says takes nine five-stars to undo at a 4.6 average.

The countermeasure is not marketing. It is operations. When you pitch, you are scheduling demand you cannot size, so schedule labor for the top of the range for two weeks, hold back a block of tables, tighten the menu if you have to, and tell the kitchen what is coming. And the harder discipline: do not expand capacity, add a daypart, or sign anything on the strength of a press spike. Chapter 35's growth decision is made on a trailing twelve months, not on a good October.

The comp and influencer policy

A comp and influencer policy is a written standing rule for how the restaurant handles requests for free food in exchange for coverage — from local accounts, visiting creators, bloggers, and the occasional person who simply asks. You need one, in writing, before you open, for one blunt reason: without a policy, the decision gets made at 7:30 on a Friday by a host who does not want to be rude, and it gets made differently every time.

A workable policy for a restaurant at this scale:

The reason code matters more than it sounds. Comps that are not coded are invisible, and invisible comps are Chapter 1's cost drift wearing a friendly face. If the marketing line is \$23,250 and the uncoded comp line is \$9,000, you have a marketing budget of \$32,250 and no idea what it bought.

27.7 Loyalty, gift cards, and repeat-visit mechanics

Back to the asymmetry. Chapter 23 valued half a visit per guest per year at \$70,538 and priced the same money bought through reach at \$98,223** with a three-year payback. Against a \$23,250 budget, the frequency prize is three times everything you have and the reach bill is more than four times it. If you internalize one allocation principle from this chapter, make it this one: spend on the fourth visit before you spend on the first.**

A loyalty program is any structured mechanism that rewards repeat purchase — points, punches, tiers, credits, or recognition. Restaurants adopt them badly, most often by installing a discount and calling it loyalty. The distinction is worth an identity.

The incrementality bar

Take a discount-based mechanic: **\$10 off**, some number of redemptions. Let $D$ be the discount, $C$ the contribution per cover, and $r$ the share of redemptions that are genuinely incremental — visits that would not otherwise have happened.

• Value created: (incremental covers) × $C$
• Cost incurred: (all redemptions) × $D$ — because everyone redeems, incremental or not

Break-even requires:

$$r \ge \frac{D}{C}$$

That is the whole thing. A ten-dollar discount against \$18.40 of contribution requires 54.3% of redemptions to be incremental just to break even. More than half the people using your offer must be people who were not coming.

🧮 Run the Numbers

The loyalty line, tested honestly.

Bellwether budgets \$4,200** for loyalty and gift cards: \$700 of card stock and processing, and \$3,500** of modeled discount cost on the repeat-visit offer. At \$10 a redemption, that funds 350 redemptions.

If every redemption is incremental ($r = 1.0$): 350 covers × \$18.40 = \$6,440 of contribution, less \$3,500 of discount = **+\$2,940**.

If half are incremental ($r = 0.5$): 175 covers × \$18.40 = \$3,220, less \$3,500 = **−\$280**. Underwater.

Break-even: $r = \$10 \div \$18.40 = 54.3\%$.

Now change one variable. Make the reward an access reward instead of a discount — first reservation on the night the new menu drops, a seat at the chef's counter, the December Saturday that is otherwise impossible to book. $D = \$0$, so the break-even incrementality is zero. Any incremental visit at all is pure contribution, and the cost is a table you were going to sell anyway plus the discipline to hold it.

This is the most useful thing in the section. Discount mechanics have a hard, computable, frequently-missed bar. Access and recognition mechanics have no bar at all, because they cost hours and inventory management rather than margin. A neighborhood restaurant with 68 seats has scarcity to give away and should give that away instead of money.

Four mechanics, ranked by what they cost you

And one design rule that follows straight from the plan: reward the visit, not the spend. A spend-based program pays your highest-check guests to do what they were already doing. Bellwether's problem is not check average — Chapter 24 owns that — it is the fourth visit. A mechanic that counts visits and rewards the fourth one is aimed at the actual constraint.

The distribution you cannot manage without

A loyalty program's real product is not the reward. It is the visit-count distribution: how many guests came once, twice, three times, four or more. If you cannot produce that table, you do not have a loyalty program — you have a discount and a mailing list.

(Modeled distribution consistent with the Bellwether plan's 9,035 guests × 4.0 average visits; constructed for teaching, not measured.)

Read it once and the strategy writes itself. Guests who come four or more times are 3,135 people — 34.7% of your guests — producing 25,640 covers, 70.9% of the room. And 2,700 people, nearly a third of everyone who walked in, came exactly once and produced 7.5% of your covers.

Take 300 of those one-timers and turn them into three-visit guests. That is 600 additional covers, $600 \times \$18.40 = \$11{,}040$ of contribution — almost half the entire marketing budget, from three hundred people you have already met and already paid to acquire. Nothing on the acquisition side of this chapter is that cheap.

Gift cards

Gift cards are the most underrated instrument in an independent restaurant, for three structural reasons: cash arrives before the cost does, the recipient is very often somebody who has never been to your restaurant, and the redeeming guest reliably spends more than the face value.

Modeled for Bellwether's first holiday season: 250 cards at an average \$60 face = \$15,000 of cash collected in December and January. Say 210 redeem within eighteen months at an average check of \$78 against a \$60 card — $210 \times \$18 = \$3{,}780$ of spend above face. Forty go unredeemed: \$2,400 of balance sitting on the books.

If half of the 210 redeemers are first-time guests, that is 105 new guests acquired for the cost of card stock — at \$220.80 of lifetime value, a modeled \$23,184, almost exactly the entire marketing budget. Treat that last figure as an upper bound rather than a forecast: gift-card recipients are a gifted-into-it population and there is no reason to assume they retain like a guest who chose you.

⚖️ Code and Compliance

A gift card is a liability, not revenue — and it is regulated.

• Accounting. Cash received for a gift card is a liability until the card is redeemed. It is not sales, it does not belong in your revenue line, and it will make your December bank balance look considerably better than your December P&L. Chapter 33 will tell you what happens to operators who spend it as though it were income.
• Expiration and fees. The federal Credit CARD Act of 2009 restricts expiration dates and dormancy or inactivity fees on store gift cards and gift certificates. Many states are stricter, and several prohibit expiration entirely.
• Unredeemed balances. State escheat (unclaimed property) law governs balances that are never redeemed. Rules differ enormously — some states require remittance after a dwell period, some exempt gift cards. You do not get to book breakage as income because the card is old.
• Practical. Track cards in the POS, not in a shoebox, and reconcile the outstanding balance monthly.

All of this varies by state and it changes. Verify locally and ask your accountant before your first December, not after it.

27.8 Events, partnerships, and neighborhood presence

Bellwether budgets **\$5,200** here — \$3,600 for six events and partnerships, \$1,600 for four community sponsorships. It is 22.4% of the marketing budget and it is the hardest section of this chapter to write honestly, because it is where the restaurant's arrival in the Rivermill District stops being a positioning statement and becomes a set of consequences for people who were already there.

What events are actually for

Not exposure. Events exist to put covers on the nights you cannot otherwise sell, and to build the kind of local standing that no channel in this chapter can buy.

Look at the shape of the week the plan projects:

FIGURE 27.5 — The shape of the week, and where the empty seats are     [the Bellwether plan]

  night   covers    68 seats                                     seats sold at 1.0 turn
  TUE       62      ████████████████████████████████             91%   ← the problem
  WED       78      ██████████████████████████████████████       115%
  THU       92      █████████████████████████████████████████    135%
  FRI      120      ███████████████████████████████████████████  176%
  SAT      123      ███████████████████████████████████████████  181%
  ────────────────────────────────────────────────────────────────────
  WEEK     475 dinner covers + ~220 brunch = 695/week = 36,140/year

  Ten covers added to Tuesday, every Tuesday:  520 covers × $18.40 = $9,568/year
  Ten covers added to Saturday, every Saturday: not available. The room is full.

That is where the \$3,600 goes. Six events across the year, each aimed at a Tuesday or an early Wednesday, each costing about \$600 of incremental food, beverage, and labor: a winemaker dinner with the distributor sharing cost, a hearth-cooking demonstration for twenty, an industry night for other restaurants' staff on their Monday, a neighborhood association meeting that takes the room at 5:30 and leaves by 7:00, a farm dinner in September, a January prix-fixe series when the whole city stops going out.

The measurement is simple and most operators skip it: covers on the night versus the same night four weeks prior. If a Tuesday event produces 88 covers against a 62-cover baseline, it produced 26 covers; $26 \times \$18.40 = \$478$ against roughly \$600 of cost — a loss on the night, and possibly still worth it if the 88 people include twenty who had never been and six who come back. State which of those two claims you are making before the event, then check.

The neighborhood, and what Bellwether is doing to it

Chapter 2 established two facts about this plan that most business plans would have buried, and they belong together.

First: this is share-taking, not gap-filling. Bellwether adds 68 seats to a direct-occasion competitive base of 172 seats — a ~40% increase in supply for the same occasion in the same trade area. Unless the occasion demand in the Rivermill District grows by something close to forty percent, some of Bellwether's 36,140 covers are covers that currently belong to somebody else.

Second: the positioning depends on the neighborhood continuing to change. A \$46 dinner check in a former warehouse district works because that district is eight years into gentrification and the population that has moved in can pay it. That is not a neutral observation. It means Bellwether's business model is levered to a process that is raising rents on the people and businesses who were there first — including a fifteen-year-old family restaurant in its own competitive set, whose guests are more price-sensitive than Bellwether's and whose lease will reset on the same rising rents that made this space available in the first place.

Both directions of that relationship are real, and it is worth being precise about them, because the temptation is to pick one and be comfortable.

Bellwether is a consequence of the change and an accelerant of it. It did not start the gentrification of the Rivermill District — eight years of it preceded the lease — and the chef-owner is not a developer; they are a cook with fourteen years of experience and a personal guarantee. Restaurants are usually among the earliest and most exposed businesses in a changing neighborhood, and a great many of them fail in exactly the districts they are accused of gentrifying. But a 68-seat, \$46-check restaurant opening on that block is also a signal to the next landlord about what that block can now charge, and pretending otherwise is not honesty, it is marketing.

And the leverage runs both ways, which the plan must say out loud. If the district's change stalls — a development that does not get financed, an employer that leaves — Bellwether has a \$46 check in a neighborhood that cannot support it and a ten-year lease. That is a risk in the plan, not a virtue, and it belongs in the risk section where a reader can see it.

So what does a marketing chapter have to say about this? Three things, none of which resolve it.

One: you cannot market your way out of it, and the attempt is the worst version of it. There is a recognizable genre of restaurant marketing that borrows a neighborhood's pre-gentrification identity — the warehouse aesthetic, the industrial name, the "part of this community" copy — while serving a price point the neighborhood's prior residents cannot afford. It is effective and it is cynical, and the people it describes can read. If your marketing claims a relationship to the neighborhood, the claim has to be checkable.

Two: the checkable version costs money and shows up in other chapters. Hiring from the neighborhood is a recruiting decision (Chapter 17) with a real cost in training time. Buying from neighborhood suppliers is a purchasing decision (Chapter 9) that will sometimes cost you margin. Giving the neighborhood association the room at 5:30 on a Tuesday is a revenue-management decision (Chapter 24) about a table you could have sold. Each is a commitment with a line item, not a slogan. A restaurant that does them can say so plainly; a restaurant that does not, should not.

Three: the competitive relationship does not have to be adversarial, but it is also not your gift to give. A fifteen-year-old family restaurant two blocks away is not your partner because you would like the story. Reciprocal partnerships that work are the ordinary commercial kind: a shared street event where both rooms fill, a coordinated industry night that moves both Mondays, a co-op ad where both names appear and both pay. Those require the other operator to want it, on terms that benefit them. They may not, and the correct response to that answer is to leave them alone.

What an operator should refuse, specifically: using another restaurant's longevity as content. The photo op with the neighborhood institution, the "we love our neighbors" post featuring a business whose covers you are taking, the earnest caption about community written by a restaurant that added forty percent to the block's supply. If you would be embarrassed to have that other owner read the caption, do not post it.

🤝 Hospitality

Neighborhood presence is a hospitality practice, not a campaign.

The most valuable thing Bellwether can do in the Rivermill District costs nothing and appears in no budget line: be a good business on the block.

Concretely, and none of this is marketing copy — it is a list of behaviors with costs. The trash gets broken down and inside, not stacked on the sidewalk at 11:30 p.m. The staff parks where the neighbors do not need to. When someone complains about hood smoke you go talk to them that week instead of citing the permit. The line cooks from the place across the street get taken care of on their Monday, at a price they can pay. The neighborhood association gets the room at 5:30 for free, and you stay for the meeting. Somebody's kid's team gets sponsored and nobody photographs the check.

Every one has a cost — hours, a table, a sponsorship line, a night's revenue — and none is trackable to a cover, which means none will ever justify itself in §27.9's attribution report. Fund them anyway. In a neighborhood restaurant, over a ten-year lease, local standing is the largest marketing asset you will ever accumulate, and it is built entirely out of things indistinguishable from simply behaving well.

The honest caveat: standing does not make you welcome to everyone, and it does not settle what your arrival costs the people who were here first. It only means you are having that conversation as a neighbor rather than as a press release.

27.9 Measuring it: attribution, comp codes, and the honest \$0 plan

Everything in this chapter is worthless if you cannot tell which of it worked. And the honest opening statement is this: restaurant marketing attribution is genuinely hard, most of it cannot be measured, and the practitioners who claim otherwise are selling a dashboard.

A restaurant has no login, no account, no cart, and no cookie following a guest from an advertisement to a table. Somebody sees something, thinks about it for eleven days, and walks in with a friend who suggested it for unrelated reasons. There is no honest way to attribute that.

So you do two things. You make a small part of it hard-measurable, and you ask about the rest.

Comp codes: the hard evidence

A comp code is a POS discount or promotion code tied to a specific campaign, so that a redemption creates a record. It is the only marketing evidence a restaurant generates that is not self-reported. Four rules make it work. A separate code per campaign — not "PROMO," but NEIGHBOR for the pre-opening mailer, LIST-03 for March's email, GC-FIRST for a first-time gift-card redemption. The code must be rung, not waved; if a server can honor it without entering it, half of them will and your data is gone. The offer's dollar cost is marketing cost, and it lands on the comp line of the P&L rather than the marketing line, so reconcile the two monthly or your true spend is invisible (Chapter 31 builds the statement; Chapter 34 owns the comp report). And codes measure redemption, not incrementality — the hard limit, dealt with below.

The host-stand question: the soft evidence

One question, asked by whoever greets, of as many tables as volume allows: "May I ask how you heard about us?" Logged in one of six buckets, on a clipboard or a tablet, for thirty days a quarter.

It is self-reported, which means it is unreliable in known ways: people misremember, they say "friend" when they mean "friend plus an Instagram post plus a Google search," and they under-report advertising because nobody enjoys admitting an ad worked. It is still the most valuable marketing data a restaurant collects, because it is the only source that connects a channel to an actual cover.

🧾 Read the Numbers

```text FIGURE 27.6 — "The attribution report that admits what it doesn't know" [the Bellwether plan] THE ARTIFACT One-page monthly marketing report, March, month 12. Built from the POS comp-code register and thirty days of host-stand logging. THE CONTEXT Bellwether, end of year one. Marketing spend that month: $1,510. 3,240 covers recorded. Email list 2,380; SMS list 615; 214 reviews at a 4.6 average.

                 COVERS RECORDED (POS)                                    3,240

                 A. HARD EVIDENCE — code redeemed at the POS
                    NEIGHBOR  pre-opening mailer      41 checks      98 covers
                    LIST-03   March email offer       63 checks     152 covers
                    GC-FIRST  gift card, first visit   37 checks     101 covers
                    SUBTOTAL                         141 checks     351 covers

                 B. SELF-REPORTED — host-stand question
                    asked at tables representing 61% of the
                    2,889 covers not already coded            1,762 covers
                    "Google or maps"                            412 covers
                    "a friend brought me / recommended"         559 covers
                    "walked past it"                            288 covers
                    "Instagram"                                  97 covers
                    "read about you"                             44 covers
                    SUBTOTAL, usable answers                  1,400 covers
                    asked, no usable answer                     362 covers

                 C. UNATTRIBUTED                               1,489 covers
                    (362 asked without an answer + 1,127 never asked)

                 TOTAL                                        3,240 covers

                 Of the 351 hard-attributed covers:
                    acquisition-coded (NEIGHBOR)                 98 covers   28%
                    frequency-coded (LIST-03, GC-FIRST)         253 covers   72%

WHAT IT SHOWS Two things worth the whole page. First, 72% of everything this restaurant can hard-prove its marketing produced was a return visit, not an acquisition — which is exactly the shape Chapter 23's arithmetic predicted and exactly the opposite of how the money is usually spent. Second, the largest single self-reported source is other guests: 559 covers from "a friend brought me," against 97 from the channel that consumes the most management attention. WHAT IT DOESN'T It cannot tell you whether any of the 351 coded covers were incremental. Every one of those guests might have come without the offer, in which case the discount was a gift and the campaign produced nothing. It cannot see the people who considered you and went elsewhere. It cannot separate "Google" from "my friend told me and then I searched you," which is probably the truest answer for several hundred of these covers. And 46% of the month is simply unattributed — a number that will never reach zero and should not be tortured toward it. THE DECISION Three, all cheap. Keep LIST-03 and run the incrementality test on it in April: hold out 20% of the list at random and compare visit rates. Kill nothing yet on Instagram's 97 covers — it costs cash of zero — but stop treating it as an acquisition channel in the plan. Raise the host-stand ask rate from 61% toward 85% by moving the question from the greeting to the water drop, where there is more time. THE LESSON An honest attribution report has a large unattributed block and says so. Its value is not precision — it is direction: it told this operator that the money is in frequency and the people are coming from each other. ```

The incrementality problem, and the one test worth running

Comp codes prove redemption. They do not prove that the visit happened because of the offer, and the difference is the whole ballgame.

Take the paid line — \$3,000 of local search and social, capped and tested across ten months — and suppose it produces 240 traceable first visits carrying a \$10 offer.

Two accounting methods, same answer — which is the check you should run whenever you are unsure whether you are double-counting an offer. Pick one and be consistent; do not mix them.

Now break it. If only 60% of those 240 were incremental, you acquired 144 guests, and the real cost is $\$5{,}400 \div 144 = \$37.50$. Cumulative contribution reaches \$36.80 after two visits and \$55.20 after three: break-even moves to visit three, and a channel you were about to scale is now a channel that depends on retention you have not proven.

The test that settles it, and it is nearly free: hold out a random slice. Take 20% of the list, or 20% of the mailing area, and deliberately do not send them the offer. Compare visit rates between the two groups over the following six weeks. The difference is your incrementality. It costs you the offer revenue from a fifth of the list and it is the only thing in restaurant marketing that approaches proof.

The \$0 marketing plan, honestly

Here is the whole thing, and the honesty is in the last column.

Total cash: \$0. Total hours: about five a week, every week, forever.** At a \$25/hour replacement cost that is \$6,500 a year — 28% of the entire marketing budget** — and it appears nowhere in the \$23,250, because the front-of-house partner absorbs it into a salary that is already sitting in the labor line.

That is the honest version of the \$0 marketing plan. It is not free. It is paid in the scarcest resource an owner-operator has, and it is the best trade available to a restaurant with \$23,250, which is why it is the foundation of the plan and not a footnote to it. The failure mode is predictable and nearly universal: the hours get spent in months one through four, service gets busy, and by month seven nobody has answered a review since April and the holiday hours are wrong. Put it on the schedule with a name on it, or it will not happen.

🔍 Check Your Understanding

1. A campaign spends \$1,200 of media and carries a \$8 offer redeemed 300 times. What is the cost per cover acquired if all 300 were incremental? If only 40% were?
2. Your discount is \$12 and contribution per cover is \$18.40. What share of redemptions must be incremental for the offer to break even, and what does that suggest about the mechanic?
3. In Figure 27.6, 46% of covers are unattributed. Name two reasons that is acceptable and one thing you would do to shrink it.

(1: Cost = \$1,200 + (300 × \$8) = \$3,600. All incremental: \$3,600 ÷ 300 = \$12.00 per cover acquired. At 40%: 120 incremental covers, \$3,600 ÷ 120 = \$30.00. 2: $r \ge \$12 \div \$18.40 = 65.2\%$ — nearly two-thirds, which is a very high bar; the discount is too deep and an access or occasion mechanic would carry no bar at all. 3: Acceptable because guests genuinely do not know how they found you and because word of mouth is inherently unattributable; a plan that requires full attribution is a plan that will fabricate it. To shrink it: raise the ask rate by moving the question from the greeting to the water drop, and add codes to more campaigns.)

🍽️ The Business Plan

Checkpoint 27 of 40 — the Marketing Plan.

Part VI opens with the cheapest channel in the book: the room you have already paid for. Here is the Marketing Plan section, as it goes into the document.

The objective

Not awareness. 36,140 covers, from 9,035 distinct guests, at an average of 4.0 visits each. Marketing is accountable for the first of each guest's four visits — 9,035 covers, 25% of the year. The remaining 27,105 covers are produced by the food, the room, and the service, and belong to Chapters 22, 23, and 24.

The budget

FIGURE 27.7 — The $23,250, by what it is buying                       [the Bellwether plan]

  foundation      ████████████████████████████     $6,470   27.8%   photos, site, print
  frequency       ███████████████████████          $5,280   22.7%   email/SMS, loyalty, gift
  neighborhood    ███████████████████████          $5,200   22.4%   events, sponsorships
  acquisition     ██████████████████               $4,300   18.5%   previews, mail, paid
  contingency     █████████                        $2,000    8.6%   unspent by design
  ──────────────────────────────────────────────────────────────
  TOTAL                                           $23,250  100.0%

    = 1.5% of $1,550,000 of year-one revenue
    = 10.7% of the $217,000 other-operating line
    = 3.5% of the $664,976 of contribution those covers produce

  NOT IN THIS BUDGET, AND NOT FREE:
    ~5 hours/week of the front-of-house partner  =  260 hrs/yr
    at a $25/hr replacement cost                 =  $6,500

Pre-opening, weeks −10 to 0 — \$5,250:

Year one, months 1–12 — \$18,000:

$\$5{,}250 + \$18{,}000 = \$23{,}250$, and $\$23{,}250 \div \$1{,}550{,}000 = 1.5\%$.

The channels

The cost per cover

The plan's headline number is \$2.57 per cover acquired against \$18.40 of contribution on that single cover — and it survives every honest adjustment we can make to it.

What this checkpoint settles

That the marketing budget is adequate for acquisition. Even at one-third attribution and with labor priced at replacement cost — \$9.88 a cover — a first cover costs less than ten dollars against \$18.40 of immediate contribution. Acquisition is not this plan's risk.

What it does not settle

Considerably more.

1. Whether the trade area contains 9,035 people who want a \$46 dinner. Chapter 2 computed the capture requirement — 7.7% of the ten-minute drive-time population — and this chapter has done nothing to make that easier. Marketing does not create demand; it competes for it.
2. Whether they come four times. The entire plan rests on 4.0 visits per guest per year, and §27.7's modeled distribution shows what that requires: 3,135 guests at four-plus visits producing 71% of the covers. Not one line of the \$23,250 can force that. Chapters 22, 23, and 24 can.
3. Whether the loyalty offer is incremental. The \$3,500 discount line needs 54.3% incrementality to break even and the plan has no evidence either way. The hold-out test in month four is how it gets settled, and the plan should say so.
4. The neighborhood dependency, in both directions. Bellwether's \$46 check requires the Rivermill District to keep changing; its 68 seats add ~40% to the direct-occasion supply and some of its covers will be taken from operators already there, including a fifteen-year-old family restaurant. Neither fact is resolvable by a marketing plan and both belong in the risk section.
5. The \$139,240 that is not in these covers. Chapter 1's four-variable estimate on the plan's own assumptions produces \$1,410,760 of dining-room revenue against a \$1,550,000 plan. That gap does not close by finding more dinner guests — the Friday and Saturday rooms are already full. It closes, if it closes, in the rest of Part VI and in Chapter 24. Marketing's job is not to find the gap. It is to make sure the 36,140 covers underneath it actually happen.

Open questions carried forward:

1. Can off-premise add contribution without cannibalizing the dining room? (Chapter 28)
2. Can private events fill the Tuesdays that §27.8's six events cannot? (Chapter 29)
3. Does the \$5,250 of pre-opening marketing land in a month the cash forecast can carry? (Chapter 33)
4. Where does the offer's discount cost appear on the P&L, and will anyone reconcile it to marketing? (Chapters 31 and 34)
5. What happens to this plan if the review average settles at 4.2 instead of 4.6? (Chapters 23, 39)

Conclusion

A restaurant marketing budget is small enough that it can only do one thing well, and this chapter argued about which thing.

The arithmetic settled it. Bellwether needs 36,140 covers, which is 9,035 guests times four visits. Marketing buys the first visit; hospitality buys the other three. At \$23,250 the first visit costs \$2.57** and is worth **\$18.40 immediately — a ratio so comfortable that acquisition is simply not the risk in this plan, even after you adjust attribution down by two-thirds and price the owner's hours at replacement cost. The risk is the fourth visit, which Chapter 23 valued at three times this entire budget and which no line of it can purchase. So the money went where the arithmetic sent it: 27.8% to a foundation that must be correct before anything else works, 22.7% to frequency, 22.4% to the neighborhood, and 18.5% to acquisition — the smallest slice, deliberately, in the category everyone assumes marketing is about.

Along the way the chapter refused two numbers. It would not tell you what a star is worth in revenue, because nobody has demonstrated it; it gave you the offset identity instead, which is a definition and therefore true — nine five-stars to undo one one-star at a 4.6 average, thirty-two to undo one bad Saturday. And it would not give you engagement benchmarks, because the ones in circulation trace back to nothing. What it gave you instead is a cadence you can hold and a test you can run.

The honest summary of the \$0 plan is that it costs about five hours a week, forever, and those hours are the scarcest thing an owner-operator has. Everything durable here — the profile, the reviews, the list, the neighborhood — is built out of that time rather than out of the budget, which is a fair description of the whole business.

Chapter 28 takes the opposite case. Where this chapter's channels were nearly free and hard to measure, off-premise is expensive and exactly measurable: a commission rate, a packaging cost, and a throughput cost, against a check you can see. The question it asks is the one this chapter kept deferring — whether a channel that adds revenue adds any money at all.

Key Terms

Cost per cover acquired — the marketing cost of producing one incremental cover: media, production, offer cost, and purchased labor, divided by covers that would not have happened without the spend. Bellwether's plan figure is \$2.57 against \$18.40 of contribution. (Ch. 27)

Organic vs. paid — organic placement is earned through a platform's ranking and carries no marginal cost; paid placement is bought through a bid and stops the moment the budget does. (Ch. 27)

Local SEO — the practice of making a business findable in geographically qualified searches and map results, driven principally by relevance, distance, and prominence — of which only relevance is under an operator's control. (Ch. 27)

Google Business Profile — the free, business-claimed listing that populates the Google Search and Maps panel: name, category, address, hours, links, photos, attributes, Q&A, and reviews. For most neighborhood restaurants it is the most-viewed page the business has. (Ch. 27)

Content cadence — the sustainable publishing rhythm a business commits to: how often, in what formats, produced by whom, on what day. The operative word is sustainable. (Ch. 27)

Earned media — coverage you did not pay for. A demand-side event that fills a room and changes nothing about your cost structure. (Ch. 27)

Email and SMS lists — direct, permission-based owned channels with no intermediary between the business and the guest; the only marketing asset that appreciates, and the only one that transfers with a sale. (Ch. 27)

Loyalty program — any structured mechanism that rewards repeat purchase, through discount, access, occasion, or recognition. A discount mechanic offering $D$ against contribution $C$ requires incrementality of at least $D/C$ to break even. (Ch. 27)

Review response — the public reply an operator posts to a review. Written for the next reader rather than for the reviewer: short, specific, never adversarial, and moved offline exactly once. (Ch. 27)

Comp and influencer policy — a written standing rule for how the restaurant handles requests for free food in exchange for coverage: who approves, from what budget, with what disclosure, and what every staff member says when asked. (Ch. 27)

Spaced Review

1. Without looking back: state the formula for the number of five-star reviews needed to offset one one-star at an average of $A$, and compute it for 4.5 and 4.8.
2. A campaign costs \$2,000 of media and carries a \$15 offer redeemed 200 times. Compute the cost per cover acquired at full incrementality and at 50%. Given contribution of \$18.40, what do you conclude about the offer size?
3. From Chapter 1: a restaurant gets a glowing local review and does 40% more covers for six weeks. Its prime cost is 68%. What happens to the business, and why is that not a paradox?
4. From Chapters 2 and 23: Bellwether needs 9,035 guests at four visits each. Explain why a marketing plan built around reach would be the wrong plan, using the two numbers Chapter 23 derived.
5. The recurring question: an operator moves \$3,000 from the paid-acquisition line into a recognition program — POS notes, a trained habit of greeting returning guests by name, and a held table for regulars on Saturdays. Does prime cost move? Does contribution? What would you have to measure to know whether the trade was good, and how long would you have to wait?