Chapter 25 — Submittals, RFIs, and Document Control: The Paper Trail That Protects the Project

The Hook: The Log Was in the Wrong Order

It was a Wednesday in the seventh week of the Northgate job — April 16, Year 1, forty-four days after notice to proceed. Dani Okonkwo had been a field engineer for seven weeks and had already learned the two things every field engineer learns first: where the good coffee is, and that nobody will tell you what you are supposed to be doing.

I dropped a laptop on the desk next to them, open to a spreadsheet.

"Submittal register," I said. "Two hundred fourteen rows. It got built off the project manual before NTP. I need you to run it from here on. Find me the next one."

Dani looked at it. "The next one?"

"The next one that's going to hurt us."

I want to be honest about what I had just done, because it is the most common failure of supervision in this business and I did it without thinking. I handed a twenty-four-year-old a document, gave them a goal, and gave them no method. Then I went to a meeting.

Here is what Dani was looking at. Two hundred fourteen rows. Column A: specification section. Column B: submittal description. Column C: responsible subcontractor. Column D: date received. Column E: date sent to architect. Column F: date returned. Column G: status. Sorted, top to bottom, by specification section number — 03 21 00, then 03 30 00, then 04 22 00, then 05 12 00, and on down through Division 33.

It was a beautiful spreadsheet. Somebody had frozen the header row and color-banded the divisions.

Dani spent about twenty minutes trying to answer my question by scrolling. You cannot answer it by scrolling. There is nothing in a specification section number that tells you when anything is needed. Division 03 is concrete and Division 26 is electrical, and on our job the electrical switchgear had a manufacturing lead roughly nine times longer than a concrete mix design. The list was sorted by an accident of the alphabet.

So Dani did the thing that changed the job. It was not clever, and it took most of an afternoon. They opened the CPM schedule Wei Chen had built, and next to each submittal row they wrote the date the physical thing has to be on the site. Then they added one more column — the date the submittal itself had to arrive at Kestrel's desk in order for that to happen — and they sorted the spreadsheet by that column instead.

The order changed completely. Division 26 rows jumped to the top. Division 03 rows scattered through the middle. Whole blocks of Division 09 finishes fell to the bottom, where they belonged.

And three rows came up red.

Concrete design mixes. Footings, foundation walls, slab on grade. The project manual required approved mix designs not less than thirty days before first placement, and first footing placement was May 18. Working backward, the mix designs had to be in Kestrel's hands by March 13. It was April 16. Status: not issued. Thirty-four days late. Nobody was chasing it, because Kestrel self-performs concrete, and the register's "responsible subcontractor" column said Kestrel — and a column that says your own name is a column nobody reads.

Reinforcing steel placing drawings. Bar lists and placing drawings for the same footings. Due to Kestrel March 22. Status: not issued. Twenty-five days late.

Anchor bolts and embeds. Submitted by Ironbridge Steel on March 27, exactly on time. Received at Kestrel March 27. Sent to Caldwell Structural on April 7. Status: out for review.

Dani stared at that row for a while, then came and found me in the trailer.

"The anchor bolts sat here for eleven days," they said. "The log says our review is five. Caldwell's fourteen days aren't up until the twenty-first. But your back-schedule in the procurement file says we needed them approved on the fifteenth."

"Yesterday," I said.

"Yesterday."

That is the whole chapter, in one exchange in a trailer. We had a correct back-schedule. I built it myself in March, and Wei Chen loaded the dates into the schedule file, and it lived in a folder that got opened once a month. And we had a submittal register — the document the team actually touched every single day — sorted by an ordering principle with no relationship whatsoever to when anything was needed.

Five weeks later Ironbridge Steel missed its mill rolling slot, the next opening was five weeks out, and steel erection start slipped from August 4 to August 27, Year 1 — twenty-three calendar days, on the critical path, at a combined exposure of $10,650 per calendar day. We spent $168,000 accelerating.

The document that would have prevented it was a spreadsheet sorted the other way.

Part V Begins Here: The Three Systems That Move Information

You have just spent seven chapters in the field. Layout, subcontractors, labor, equipment, temporary structures, quality, safety — the physical work of putting a building together. Part V steps back into the trailer, and it is about the least glamorous and most consequential thing that happens there.

A construction project runs on three information systems, and every project has all three whether anybody designed them or not:

Each one has a failure mode that looks technical and is not. A steel package arrives late (document control). A subcontractor swears nobody told them about the Tuesday pour (meetings). A drywall crew shows up to a wall that has not passed in-wall inspection (planning). None of those are engineering problems. All of them cost real money.

This chapter takes the first system. By the end of it you will be able to build a submittal log that behaves like a schedule instead of an inventory, write an RFI — a Request for Information — that gets a usable answer in one round instead of three, run a drawing set so that no crew ever builds from a superseded sheet, and, the part that separates a project manager from a project administrator, create records today that will still be worth something eight months from now when somebody is arguing about what happened.

🏃 Fast Track: If you have run a submittal log before, go straight to §25.3 (the three back-scheduled Northgate items — read the anchor-bolt column), §25.7 (the bad RFI and the good RFI, side by side), and §25.9 (the CO #14 documentation autopsy, which is the most important table in this chapter). Skim the rest.

🔬 Deep Dive: For where the drawings and specifications that generate all of this come from, revisit Chapter 7. For the procurement side of the back-schedule, Chapter 16. For what happens to a record when it becomes evidence, Chapter 33. Blank forms are in Appendix D.

25.1 What a Submittal Actually Is — and What Review Does Not Do

Start with the definition, because most people carry a wrong one around for years.

A submittal is the contractor's demonstration that what it intends to furnish and install conforms to the contract documents.

Read that again and notice who is doing what. The contractor demonstrates. The contractor asserts conformance. The design professional reviews the demonstration and says whether it appears to conform to the design intent. The direction of the obligation runs from the contractor outward, not from the architect inward.

That is not a semantic point. It is the entire allocation of liability in the submittal process.

⚖️ What the contract says. Standard general conditions — the AIA A201 family, ConsensusDocs, EJCDC, and most owner-drafted forms that borrow from them — carry two linked provisions you should be able to recite from memory:

One: by submitting, you certify. The contractor represents that it has reviewed and approved the submittal, has determined and verified materials, field measurements, and field construction criteria, and has checked and coordinated the information with the requirements of the work and of the contract documents. In plain language: you are telling the architect you already checked it.

Two: review does not relieve you. The design professional's review is for the limited purpose of checking conformance with the information given and the design concept. It does not relieve the contractor of responsibility for deviations from the contract documents unless the contractor specifically informed the design professional of the deviation in writing and the design professional gave written approval of that specific deviation. It also does not relieve the contractor of responsibility for errors or omissions in the submittal itself, for dimensions, for quantities, for fabrication processes, for techniques of construction and installation, for safety precautions, or for coordination with the work of other trades.

The exact wording varies by contract and by design firm, and specified review periods, retention obligations, and stamp language differ by jurisdiction and by agreement. Read yours. But the structure is nearly universal, and it produces a rule you can carry to any job: approval of a submittal transfers almost nothing.

What the stamps mean

Design firms use different stamps and different words. Here is the common vocabulary and what each entry does to your risk and your schedule.

Two practical consequences fall straight out of that table.

The gap between "approved as noted" and "revise and resubmit" is worth weeks. On Northgate that gap was worth twenty-one calendar days on the curtain wall and, on a different chain, the difference between hitting a mill slot and missing it. Which is why one of the cheapest conversations available to you happens at the preconstruction meeting: sit down with the architect and the engineers, name the four to eight submittals that gate a long-lead release, and ask that for those specific items the reviewers use "approved as noted" wherever their comments do not go to design intent. It costs nothing. Nobody is being asked to lower a standard. You are asking for a stamp choice on items where the difference between two stamps is a manufacturing slot.

"Reviewed for general conformance" is not a shield you can hide behind. I have watched a contractor argue, seriously, that because the architect stamped the shop drawings, the architect owned a dimensional error that shut down a fabrication run. That argument almost never works, and the reason is in the certification made at submission: you told them you had verified the dimensions.

🔍 Why this works. The allocation is not arbitrary; it follows information. The contractor and its fabricator hold information the design professional does not — actual field dimensions, the real capabilities of a particular shop, how a specific product installs, what the other trades are doing in the same three inches of ceiling space. The architect holds the design intent. Each party is made responsible for what only it can know. When a submittal review goes wrong, it is almost always because one party assumed the other was checking something only the first party could see.

🔄 Check your understanding. Your curtain-wall fabricator submits shop drawings showing a 1/2-inch perimeter joint. The specification requires 3/4 inch. The architect stamps the submittal Approved as noted, with comments about the mullion cap finish and nothing about the joint. You fabricate 38,500 SF of unitized panels to a 1/2-inch joint. Whose problem is it?

25.2 The Submittal Types, and Who Prepares Them

"Submittal" is a category, not a document. What lands on your desk varies enormously in what it is for, who makes it, how long it takes, and how much damage it can do.

Two things to take from that table.

The first submittals due on a job are not the glamorous ones. They are concrete mix designs, reinforcing placing drawings, and anchor bolts — the boring end of Division 03 and Division 05. If your submittal effort starts with finish samples because those are what people want to talk about in the owner meeting, you are already behind.

Closeout submittals are the ones people start latest and regret most. O&M data, warranties, and record drawings are the only category on that list whose preparation cost rises the longer you wait, because reconstructing eighteen months of as-built information at month twenty is a different and vastly more expensive activity than recording it weekly. More on that in §25.8.

🔄 Check your understanding. Rank these four Northgate submittals by how early they were due, earliest first: curtain-wall shop drawings, concrete design mixes, restroom tile samples, low-voltage switchgear.

25.3 Building the Log: A Back-Schedule, Not a Transcription

This is the central skill of the chapter, and it is one calculation repeated a few hundred times.

Most submittal logs in this industry are built the way ours was: somebody opens the project manual, reads the "Submittals" article in Part 1 of each specification section, and types what they find into a spreadsheet in the order they encounter it. That is a transcription. It produces an accurate inventory of what is required and tells you nothing about what to do on Monday.

A submittal log is not an inventory. A submittal log is a schedule. Its rows are dated obligations, its sort order is the order in which those obligations come due, and reading it should feel like reading a look-ahead, not a bibliography.

The five subtractions

Here is the calculation. Start at the only fixed point in the chain and work backward.

📊 Diagram (described): the submittal chain, read right to left.

  SUB          GC          A/E        [REVISE +    RELEASE     FABRICATE    SHIP    ON
  PREPARES ->  REVIEW  ->  REVIEW  -> [RE-REVIEW]-> TO SHOP -> / MANUF.  -> /    -> SITE
                                                                            DELIVER
   <-------------------- you compute in this direction ------------------------------|
                                                                     required-on-site
                                                                     is the ONLY fixed
                                                                     point in the chain

1. Start at the required-on-site date. Not "when we need it." The date the physical item must be standing on the site, taken from the CPM schedule you built in Chapter 14 — from the predecessor of the activity that installs it, allowing for any preparatory work.
2. Subtract installation lead. Anything that must happen before the item can arrive: a rigging window, a crane availability, a pad that must be poured, an opening that has not yet been closed by a wall.
3. Subtract delivery time. Transit from the plant, plus permitting for oversized loads, plus offload and staging.
4. Subtract fabrication or manufacturing time, measured from release, not from order. Release means approved and cleared to build.
5. Subtract the approval cycle — the design professional's contractual review period, plus a realistic allowance for one resubmittal, plus your own review and transmittal time.
6. Subtract the subcontractor's preparation time.

The result is the submittal due date — the day the subcontractor must hand you the package. And then one more subtraction gives you the award date: the day the subcontract or purchase order must be executed, because nobody prepares shop drawings for a job they have not been awarded.

If any of those computed dates is in the past, you have found a problem, and you have found it on a spreadsheet instead of on a job site. That is the entire value proposition.

🧩 Productive struggle. Before you read the worked examples, try this one. Give it four or five minutes and write your dates down.

Northgate's imaging suite needs a lead-lined door assembly. The door has to be installed before the room's finish work starts, which the schedule puts at March 2, Year 2. Transit from the fabricator is 6 calendar days. Fabrication after release is 55 calendar days. The architect's review period is 14 calendar days, and this fabricator has never worked with Halvorsen + Pike before. Your internal review takes 5 days. The door supplier says it needs 20 days to prepare the submittal.

What is the submittal due date, and what would you tell the subcontractor today?

Three Northgate items, worked

Here are the three that mattered most on our job, run side by side so you can see how differently three chains behave. Specification numbers follow MasterFormat's structure; your project manual's exact numbering will differ.

Sit with the bottom row of the anchor-bolt column for a minute, because it is the most useful thing in this chapter.

Ironbridge Steel's subcontract was executed March 7. A back-schedule carrying one resubmittal cycle says it had to be executed February 20. The chain was already fifteen days short of feasible on the day everyone shook hands. Nobody knew, because nobody ran the calculation with a resubmittal in it.

We got away with the first fifteen days by pure luck: Caldwell Structural stamped the submittal approved as noted on the first round, which handed back all seventeen days of resubmittal allowance at once. That windfall put us back on the plan.

Then we spent it. The submittal sat in our own office for eleven days instead of the planned five, and the six extra days consumed the margin the lucky first-round approval had returned. Approval came back April 21 instead of April 15.

💰 Money check: the gate nobody had written down. Kestrel's back-schedule tracked an approval milestone of April 15. Ironbridge's actual constraint was a mill release date of April 18 — the last day Hank Duffy could commit steel against a reserved rolling slot. Three days of margin existed between those two dates, and nobody at Kestrel knew it existed, because nobody had ever asked to see Ironbridge's production schedule.

Approval landed April 21. Three days past the release date. The slot was gone; the mill's next opening for those shapes was five weeks out.

There is no crew, no crane, and no piece of equipment on a job site that burns $28,000 a day. The most expensive thing on your project is paper that is not moving.

And it did not stop at money. The acceleration produced trade stacking, a rework event on deck-edge detailing, and a spike in near-misses in weeks 34 through 36 — including the scaffold event on the north elevation. A submittal-log failure in April became a safety problem in November. Read that sentence twice; it is Theme 4 arriving through a door nobody was watching.

The curtain wall: a chain that had a resubmittal allowance and still failed

The curtain-wall column is worth its own paragraph, because it teaches the more advanced lesson.

Vantage Facade Systems was awarded April 9, Year 1 — five days ahead of the computed latest date, on a 266-calendar-day chain. Renata Vogel's team submitted 214 sheets on May 26, exactly on schedule. Kestrel transmitted in five days. Halvorsen + Pike and Caldwell Structural reviewed together and returned it on June 18 stamped Revise and Resubmit.

That return was planned for. The chain carried a 21-day resubmittal allowance, and June 18 landed precisely where the log said a first-round rejection would land.

It failed anyway. The revision took four weeks instead of eleven days, the re-review took three weeks instead of ten, and a third partial round on the lobby glass took two more. Nine weeks beyond plan. Approval moved from week 13 to week 22, and dried-in eventually slipped from March 28 to May 9 of Year 2 — forty-two calendar days.

Why did a correctly built allowance fail? Because the allowance was sized for a normal comment cycle, and what came back was not a normal comment cycle. Roughly forty percent of the north-elevation transition details were marked "by others," and "others" turned out to be nobody: Vantage assumed the precast subcontractor would detail the curtain-wall-to-precast joint, the precast shops stopped at the panel face, and neither subcontract said who owned the joint. That is a scope gap — the thing Chapter 16 warns lives between subcontracts — and resolving it required a commercial decision between two subcontractors plus a design decision from the architect. No review allowance covers that. You cannot buy your way out of a scope gap with schedule float.

The general rule: a resubmittal allowance protects you against comments. It does not protect you against a question the documents cannot answer. Those you have to find in the scope review, months earlier. The full anatomy is in the enclosure chapter, Chapter 9.

🔄 Check your understanding. In the anchor-bolt chain, Caldwell Structural took its full 14-day contractual review. Was Caldwell the cause of the twenty-three-day delay?

25.4 Running the Log: Status, Aging, and Ball-in-Court

A log that is built correctly and read monthly is nearly as useless as a log that is built wrong. The value is in the running.

Status codes

Keep them few and keep them unambiguous. Six is plenty:

Resist the temptation to add "pending," "in progress," or "with Bob." Every one of those hides the only question that matters.

The two reports that run the log

The aging report. What is out for review, and for how long? One page, sorted by days outstanding, descending.

Notice what that report does. It ranks by days out, but it makes the decision on the last column. Submittal 022 has been out twelve days and gates nothing; submittal 019 has been out nine days and gates a mill slot in two days. If you are calling the design team on Thursday about one item, you now know which one, and you know it in ten seconds instead of by memory or by luck.

The ball-in-court report. Whose desk is this on right now? Two hundred fourteen rows collapse to five numbers:

This is the single most useful management report in document control, and almost nobody produces it. It answers a question everyone thinks they know the answer to and almost always gets wrong: who is actually holding up this project?

Look at that table honestly. The answer on Northgate in week seven was the subcontractors, with 138 items not yet submitted at all — and behind that, us, because nobody had issued a single written request for any of them. It was not the architect. It is almost never the architect first. But the architect is who everyone complains about, because "out for review" is visible on a log and "never asked for" is not.

🏗️ From the field. After Dani rebuilt the log, we started a Monday-morning ritual that took eleven minutes and that I have run on every job since. Print the ball-in-court report. Read the top three rows of each section out loud. For each one, name a person and a date — not "we'll follow up," but "Dani calls Renata Vogel today and gets a commitment for Friday." Then send the report to the entire team including the architect and the owner's rep, every Monday, without editorial comment.

The effect of publishing that report is not informational. It is social. Nobody wants to be the top row.

The weekly submittal review with the design team

Thirty minutes, standing, weekly or biweekly, with the architect and whichever engineer has items out. The agenda is three columns and nothing else:

1. Returning this week. What is coming back, in what condition, with what stamp expectation.
2. Going out this week. What we are sending, so the reviewer can plan the hours. Fifteen submittals arriving Friday afternoon is not a submission; it is an ambush.
3. The gating list. The four to eight items whose approval releases a long-lead purchase. Where are they, and does anybody see a problem?

Add one standing question that changes the character of the meeting: "Is there anything in the pile that you think should not have come to you?" Architects will tell you, if asked, that a third of what they receive is either not required, or should have been caught by the contractor's own review, or is a question dressed as a submittal. Every one of those you eliminate buys reviewer attention for the ones that matter.

Escalation, and the single owner

Two disciplines, both boring, both worth money.

Escalation. When a review passes its contractual period, an email goes out that day — not a week later, not with an apology, not with a complaint. Factual, addressed to the reviewer, copied to the owner's representative: "Submittal 019 was transmitted April 7. The specified review period is 14 calendar days, which expires April 21. This submittal releases Ironbridge's mill order, which requires release by April 18. Please advise whether an expedited return by April 17 is possible." That email exists whether or not it works, because it is contemporaneous evidence of notice.

One owner. The log has exactly one person responsible for it. Not "the project engineers." One name. On Northgate that name became Dani's on April 16 and stayed theirs for nineteen months. A log with shared ownership is a log with no ownership, and it degrades in about three weeks — I have watched it happen, and the degradation is always the same: statuses stop being updated first, then dates, and then people quietly start keeping their own private lists, which is the point at which the project has no submittal log at all and does not know it.

25.5 Review Time, Resubmittals, and Who Owns the Clock

The review period is a contract term

Most specifications and general conditions state a review period for the design professional — commonly in the range of ten to twenty-one calendar days, with fourteen being the most common single figure I encounter. Northgate's was fourteen. Yours may differ, and you must read it rather than assume it, including any provisions about resubmittals, about batching, and about what happens on items requiring review by more than one consultant.

⚖️ What the contract says. Three things about that clause that people get wrong.

One: exceeding the review period is not automatically a compensable delay. A late review is a breach of an obligation. A delay requires that the late review actually pushed a critical-path activity, and a compensable delay requires entitlement, causation, and damages — three separate proofs, which is the threshold concept of Chapter 33. A reviewer who takes twenty days on a submittal with thirty days of float has cost you nothing you can collect.

Two: your notice obligation runs on its own clock. Nearly every contract requires written notice of a delay-causing event within a stated number of days of when the contractor knew or should have known. Miss that window and you can be right about everything else and still recover nothing. Notice periods vary by contract and by jurisdiction and are enforced strictly in many of them — check yours, and calendar it.

Three: batching destroys your own argument. A contractor who submits eighty items in one week and then complains about slow review has manufactured its own problem, and every arbitrator and judge in this industry has seen that pattern before. If your submission profile is a wall of paper in weeks 8 through 12 and a trickle thereafter, you have no standing to complain about turnaround, and you should not expect any.

The corollary is a planning obligation: level your submission curve. Look at the log sorted by due date and count submittals per week. If any week has more than the design team can plausibly review, move the ones with float earlier. Ten items a week reviewed in fourteen days beats forty items a week reviewed in forty.

Why items go around more than once

The resubmittal is the hidden schedule cost in every submittal chain. Each cycle costs the revision time plus a full re-review — on Northgate, roughly 17 to 22 calendar days depending on the item. Three rounds on a gating item is a month.

Here is where the rounds come from, in the rough order of how often I see them:

Three practices cut resubmittal rates hard.

A submittal the GC has not reviewed should never reach the architect. This is the single highest-leverage rule in this chapter and it is violated constantly, usually under time pressure, usually with the rationalization that the architect will catch it anyway. Your review is not a formality and it is not a routing step. You are checking completeness against the specification's submittal list, checking dimensions against the field and the model, checking coordination against adjacent trades, and checking for deviations that need to be declared. If you are transmitting packages you have not opened, you are not running a submittal process — you are running a mail room, and you have also just falsely certified that you checked it.

A pre-submittal meeting for complex packages. One hour, before the fabricator starts drawing, with the fabricator, the architect, the relevant engineer, and you. Agenda: what will be submitted, in what packages, in what order, what the reviewers care most about, and what is not in this scope. The curtain wall's nine-week disaster would have surfaced in that meeting as a fifteen-minute conversation about who details the precast joint.

Design-assist engagement for facades and MEP. Bringing the specialty contractor into the design process before the documents are finished — as a paid preconstruction service — collapses the review cycle because the reviewer has already seen and shaped the approach. It costs money in preconstruction and it is not right for every project or every delivery method. It is nearly always right for a unitized curtain wall.

🔄 Check your understanding. Your specification gives the architect fourteen calendar days. You transmit a submittal on the 3rd; the architect returns it on the 25th. What are the three separate questions you must answer before you can say you have a claim?

25.6 Deferred and Delegated Design Submittals

Some parts of a building are not designed by the design team at all. The specification states a performance requirement, and the contractor's specialty engineer designs the component and seals the design.

Common examples: unitized curtain walls, precast connections, wood and metal-plate-connected trusses, fire sprinkler layout and hydraulic calculations, seismic bracing for piping and ductwork, cold-formed steel exterior framing, glass railings, formwork and shoring, and stair and rail assemblies.

This is delegated design, and where the building official permits the design to be submitted after the permit is issued, it is also a deferred submittal. The sequence usually runs:

  SPEC states a           SPECIALTY ENGINEER       ENGINEER OF RECORD        AHJ REVIEWS
  performance      -->    designs and seals   -->  reviews for              -->  the deferred
  requirement             the component            consistency with              submittal and
                                                   the design of the             approves it
                                                   building
                                                          |
                                                          v
                                            ONLY THEN may the component be
                                            fabricated and installed

Three things make this schedule-dangerous.

There are three reviewers, not one. The specialty engineer, the engineer of record, and the authority having jurisdiction — the AHJ, the building department or fire marshal's office with legal authority over the work. Each has its own duration, and the AHJ's is the one you cannot escalate with a phone call from a project manager.

The AHJ step is frequently forgotten in the back-schedule. People remember to allow for the engineer of record and forget that a deferred submittal is a permit action with its own queue. On a busy jurisdiction that can be weeks. Whether a given component may be deferred at all, and how long review takes, varies by jurisdiction and by the building official — ask Frank Petrosyan's counterpart on your job, in the preconstruction meeting, and write down the answer.

Nobody owns the interfaces. Delegated design produces components. Buildings are made of connections between components. The curtain-wall engineer designs the panel and its anchor; the structural engineer designs the slab edge; and the embed plate that connects them belongs to whichever subcontract said so, which is frequently neither. This is the same scope-gap mechanism that cost the curtain wall nine weeks, and it is why the interface details on a delegated-design package deserve a specific line in your scope review.

⚠️ Safety alert: the submittal-to-installation chain on safety-critical components.

Some delegated designs are not aesthetic and not commercial. They hold people up.

• Fall-protection anchorage design — tie-off points, horizontal lifelines, and their attachment to the structure.
• Scaffold engineering — any scaffold outside the manufacturer's standard configuration, and any scaffold over the height where your jurisdiction requires a registered professional engineer's design.
• Shoring and reshoring — formwork shoring, excavation shoring, and the reshoring sequence that carries a green slab.
• Temporary bracing — steel erection bracing and precast panel bracing before final connections.
• Mast climbers, hoists, and suspended platforms.

For these, "installed from an unapproved submittal" is not a paperwork problem. It is a structural condition with people standing on it.

The rule is absolute and it does not bend for schedule: no safety-critical delegated design gets built from an unapproved submittal, and no component gets loaded before the engineer's approval is in hand and on site. If the schedule cannot accommodate that, the schedule is wrong. Not the rule.

There is a documentation corollary that is easy to miss. On the day a fall-protection anchorage or a shoring system is installed, you want three records: the approved sealed design, the installation inspection by the competent person identified in your safety plan, and photographs. If someone is hurt later, those three records are the difference between an event you can explain and an event you cannot. See Chapter 22 and Chapter 24.

25.7 RFIs That Get Answered in One Round

What an RFI is for, and what it is not for

A Request for Information is a written request for clarification of the contract documents where they are genuinely ambiguous, internally conflicting, incomplete, or in conflict with an actual field condition.

That is the whole legitimate scope. Here is what an RFI is not for, and each of these is common enough to have a name:

Volume matters, and it matters reputationally. Industry surveys and consultant studies generally find that large commercial projects generate RFIs somewhere in the range of one per a few hundred thousand dollars of construction value, that median response times commonly run one to two weeks, and that a substantial fraction — often estimated at a third or more — are avoidable, in the sense that the documents already contained the answer or the contractor could have resolved it. Treat all three of those as magnitudes, not measurements. They vary enormously by project type, document completeness, delivery method, and who is doing the counting.

But the underlying finding is robust enough to act on: RFI volume and RFI response time are among the most reliable early indicators that a project is in trouble. A high RFI count usually means incomplete documents, an inexperienced team, or both. A long response time usually means an overloaded design team, an under-resourced fee, or a contractor who is drowning the reviewer in noise.

Northgate closed with 287 RFIs on $47,500,000 of work and a mean response of 9.4 calendar days. Curtis Boone's Rivermont Elementary School job closed with 496 RFIs on $22,400,000 — roughly one RFI per $45,000 of construction — a mean response of 21 days, and sixty-three items open longer than thirty days at substantial completion. Curtis will tell you, and he is partly right, that the school's documents were thin and hard-bid pressure meant nobody had time in preconstruction. He is also describing the mechanism by which his project lost its credibility: by RFI 300, the architect had stopped reading them carefully, because most of them did not deserve careful reading.

💡 Aha moment. Your RFI log is a reputation. Every unnecessary RFI you send spends a small amount of the design team's willingness to take you seriously, and you will need that willingness intact on the day you send the one that matters. This is not a courtesy argument. It is a resource-management argument about the scarcest resource on the project, which is a reviewer's attention.

How to write an RFI that gets a usable answer in one round

There is a structure, it is six parts, and it is not optional if you want one round.

Element 4 is the one people leave out and the one that does most of the work. An architect who receives a question has to design an answer, which takes hours they do not have. An architect who receives a proposed answer has to evaluate one, which takes minutes. Same information, tenth of the effort, and — this is the part that surprises people — a materially better outcome for you, because the answer that gets approved is usually the one you proposed.

The same problem, written two ways

Here is a real condition from Northgate, level 2, Area B. The framing crew is laying out the clinic module and the dimension string on the floor plan does not close: the sum of the dimensioned room widths is 4 1/2 inches short of the overall grid dimension, and no dimension is marked HOLD.

The bad RFI. This is close to verbatim from a project I inherited, with the names changed.

RFI 112 Subject: Level 2 dimensions Date: [date]

The dimensions on level 2 don't work. Please advise ASAP as this is holding up framing and we are incurring costs.

That RFI will come back in eleven days with a response reading "Contractor to verify field conditions and coordinate." Which is not an answer, which starts round two, and which is entirely the fault of the person who wrote the RFI. Count what is missing: no sheet, no revision, no grid line, no dimension, no statement of the actual discrepancy, no proposed resolution, no quantified impact, and no date. The only concrete content is a threat.

The good RFI. Same condition, same day, same author with a method.

RFI 112 — Level 2 Area B clinic module: dimension string does not close, grids 6–9 Project: Northgate Outpatient Pavilion · To: Dale Whitcomb, Halvorsen + Pike · From: D. Okonkwo, Kestrel Construction Group Date issued: [date] · Required response date: [date + 7 CD] · Priority: Schedule-critical

References Sheet A-103 (Level 2 Floor Plan), Rev. 2 · Sheet A-421 (Enlarged Plan — Level 2 Clinic Module), Rev. 2 · Specification Section 09 22 16 (Non-Structural Metal Framing) · Grids 6 through 9, column line C.

Observed condition 1. Sheet A-103 dimensions the overall distance from grid 6 to grid 9 as 72'-0". 2. Sheet A-421 dimensions the same run as a string of seven exam rooms and one corridor: 11'-4", 11'-4", 11'-4", 11'-4", 5'-8" (corridor), 11'-4", 11'-4", 11'-4". These dimensions sum to 71'-7 1/2". 3. The difference is 4 1/2 inches. No dimension on either sheet is designated HOLD or CRITICAL. 4. Exam rooms 2-118 through 2-124 have a 10'-6" clear requirement noted on A-421 for equipment clearance; casework shop drawings (Submittal 087, approved as noted) are dimensioned to the A-421 room widths.

Question Where should the 4 1/2 inches be absorbed: (a) in the corridor width, (b) distributed among the seven exam rooms, or (c) at the grid-9 end wall?

Contractor's proposed solution Absorb the full 4 1/2 inches at the grid-9 end wall (option c), holding all exam-room dimensions and the corridor dimension as shown on A-421. This preserves the 10'-6" clear equipment dimension in all seven exam rooms, preserves the approved casework dimensions in Submittal 087, and places the variance in a wall with no fixed equipment, no casework, and no accessibility clearance governed by the ADA Standards for Accessible Design. Kestrel proposes to proceed on this basis unless directed otherwise. This proposal carries no cost or time impact.

Impact if not answered by the required response date Metal stud layout for Area B level 2 is scheduled to start [date + 8 CD] with a crew of six. If layout cannot start on that date, the crew has no alternate work available in Area B, since Area C in-wall inspection is not released. Estimated idle exposure: 6 workers × 8 hours × [rate] per day, plus a corresponding slip in the Area B in-wall inspection date, which is a predecessor to drywall in the level-2 finish sequence.

Attachments: Marked-up excerpts of A-103 and A-421 with the dimension string highlighted; photograph of the as-built grid-6 to grid-9 slab dimension taken [date], measured at 72'-0 1/8".

Copies: P. Sethi (Meridian) · R. Caldwell (Caldwell Structural) · M. Deacon (Kestrel) · Submittal 087 file

Look at what that document does that the first one does not.

It removes the reviewer's work. Dale does not have to find the sheets, re-add the dimension string, figure out which rooms have clearance requirements, or check the casework. All of that is done. He has to say yes, no, or "not option c, use option a."

It states the actual measurement. Somebody went out with a tape and measured the slab. 72'-0 1/8". That single fact converts a drawing question into a field fact and forecloses an entire round of "verify field conditions."

It proposes a resolution with a reason — and the reason is expressed in terms the architect cares about (equipment clearance, approved casework, accessibility), not in terms the contractor cares about (it is easier).

It quantifies the impact without threatening. There is no "we will be seeking costs." There is a crew size, a date, an absence of alternate work, and a downstream milestone. That paragraph is doing two jobs: it is telling Dale why the date is real, and it is creating a contemporaneous record of the exposure that will be worth something later if the answer is late. Notice that the record is created before anyone knows whether it will be needed. That is what makes it evidence rather than argument, and it is the point of §25.9.

It says what the contractor will do absent direction. "Kestrel proposes to proceed on this basis unless directed otherwise" is a powerful sentence and it must be used carefully. It is appropriate when your proposal is genuinely no-cost, low-risk, and reversible. It is not appropriate when your proposal involves cost, changes design intent, or affects a life-safety or code-governed condition.

RFIs and change orders: the question you ask on every response

Here is where contractors donate work by the millions of dollars per year.

An RFI response that adds scope is a change. The architect answers your question, the answer requires something the contract documents did not require, and — because it arrived as a response to your own question, on a form with your own number on it, in a paragraph that sounds like a clarification — the crew builds it, and nobody prices it. You have just performed extra work for free and, worse, established a course of dealing that makes the next one harder to charge for.

So train yourself and your team to read every single RFI response with one question in the front of your mind:

Did the answer change the contract?

Three ways to sort responses:

⚖️ What the contract says. An RFI response is generally not a change order. On most standard forms, an architect's response to an RFI is an interpretation, and the mechanism for authorizing changed work is a change order, a construction change directive, or an architect's supplemental instruction — depending on your form and whether price and time are agreed. If a response directs added work, your obligations are to give written notice within the contractual period and to seek a written directive. Performing directed work without one is possible in an emergency and is a bad idea otherwise. Check your contract for the exact instruments, the notice period, and who is authorized to direct changes — those vary, and on public work they are often stricter and less forgiving.

📋 Try it: write the RFI.

You are the project engineer. Here is the situation, on a Tuesday.

Cardinal Mechanical's foreman calls you from the roof. The specified rooftop unit for RTU-4 has an 18-inch curb height measured from the roof deck. The parapet detail at the north edge shows the parapet at 14 inches above the finished roof membrane at that location. The mechanical plan places RTU-4 with its service side 3 feet from the north parapet. Sheet M-104 (Roof Mechanical Plan) shows RTU-4 in that position; sheet A-107 (Roof Plan) shows the parapet; detail 4/A-514 shows the parapet section. Specification section 23 74 13 covers the packaged rooftop units and 07 54 00 covers the TPO roofing.

The condition: with an 18-inch curb and a 14-inch parapet, the unit's condensate drain and the service access on the north side conflict with the code-required and manufacturer-required clearances, and the roofer cannot complete the parapet flashing detail with the unit in that position.

The roofer, Summit Roofing, is scheduled to start the north roof area in nine calendar days. TPO roofing is 34,000 SF total, and the north area is on the critical path to dried-in.

Two solutions exist: - Option A: Relocate RTU-4 six feet south, away from the parapet. Cost: roughly $8,400 in additional ductwork, revised structural support framing, and a revised roof penetration. No schedule impact if directed within nine days. - Option B: Raise the parapet at the north edge by 8 inches for a 24-foot length. Cost: roughly $19,700 in framing, sheathing, membrane, coping, and a revised coping shop drawing — and the coping is already fabricated, adding a 4-week resubmittal and refabrication cycle for the affected length.

Your task: 1. Write the RFI in full — all six elements. 2. State what you enter in the RFI log. 3. State who you copy, and why each. 4. State what you do if no answer comes back in nine days.

Write yours before you open the model answer.

25.8 Document Control: There Is Exactly One Current Set

The current set

Every project needs one rule, stated out loud at the kickoff meeting and enforced without exception:

There is exactly one current set. Everything else is marked REFERENCE ONLY or is destroyed.

The set changes constantly, and it changes through named instruments. Know them, because the instrument tells you the contractual character of the change:

The failure mode is not exotic and it is not rare. On Northgate it cost $340,282: an assistant superintendent taped a level-3 partition plan to the field-office wall on a Tuesday; Bulletin 12 landed on Thursday; the taped copy was never swapped; and a drywall crew framed 118 partitions from a superseded sheet over nine working days. Kestrel had received the bulletin. Kestrel had acknowledged it in writing. Kestrel built the wrong thing anyway. Chapter 7 walks that failure in detail and it is worth re-reading.

The lesson is narrower than "control your documents." It is this: a document is not controlled until the last person who acts on it knows its revision. Transmittals prove you sent it. Only a layout-verification habit — the foreman checks the revision block against the current drawing log before laying out — proves somebody used the right one.

The transmittal

A transmittal is a cover record: what was sent, to whom, on what date, in what form, by whom, and what response is expected. It is not bureaucracy; it is proof.

It exists to answer, months later, a question that will absolutely be asked: did they have it? Whether Bulletin 12 was received on Thursday or the following Tuesday is the difference between a rework claim you can defend and one you cannot. Modern platforms generate transmittals automatically with a timestamped audit trail, which is one of the genuinely large improvements of the last twenty years. Use it. And note that a transmittal proves transmission, not comprehension — see the paragraph above.

The distribution matrix

Decide, once, in writing, who receives what. A simple grid: document type down the side, party across the top, and one of three entries in each cell — action, information, or nothing.

Then enforce the matrix, because the alternative is the most common distribution strategy in the industry: reply all.

Reply-all fails in a specific and predictable way. It does not fail by withholding information; it fails by burying it. When every party receives every document, no party can distinguish the message that requires action from the eleven that do not, and within about six weeks people stop opening any of them. You then have a project where everyone has technically been informed of everything and nobody knows anything — which is, incidentally, a terrible position in a dispute, because "we sent it to you" is met with "you sent us four thousand things."

As-builts are a daily habit, not a closeout task

Record drawings — commonly called as-builts — document what was actually built: routing changes, buried utility locations, actual depths and inverts, dimensional deviations, and every field change.

There are two ways to produce them.

The way that works: a marked-up set lives in the field office. Each subcontractor is contractually required to keep and mark its own set. The superintendent or a field engineer walks the marked sets monthly, verifies them against the work, and — the enforcement mechanism that actually matters — the monthly progress payment is conditioned on the as-built markups being current. Not a threat. A line in the subcontract, applied consistently from pay application one.

The way that does not: at month twenty, somebody is assigned to "do the as-builts," and they sit in a trailer with a set of drawings and a stack of RFI responses trying to reconstruct where a 6-inch sanitary line actually runs under a slab that was poured fourteen months ago, and the honest answer is nobody knows, and what gets delivered to the owner is a set of drawings that is confidently wrong. Which is worse than no drawings, because in five years somebody will core a slab based on it.

The cost differential is enormous and almost entirely invisible on a cost report, because the expensive version is paid by the owner, in the fifth year, when a renovation contractor hits something. But the reputational cost is yours. More in Chapter 40.

Cloud document platforms

The category is mature. Procore, Autodesk Construction Cloud, Bluebeam, Newforma, and a number of comparable products handle document control on most commercial projects now, and what they solve is real:

And now the honest warning, which matters more than the list.

A platform does not create discipline. It makes the absence of discipline visible faster. If your submittal log is sorted by specification section, moving it into a cloud platform gives you a beautifully rendered, mobile-accessible, permission-controlled log that is sorted by specification section. If nobody has computed a required-approval date, the software will not compute one for you — it will faithfully display the empty field, in color, on every device your team owns. If your team does not read the ball-in-court report on paper, they will not read it on a phone.

I have been on jobs that spent real money on a platform and got worse, because the software created an illusion of control that replaced the eleven-minute Monday meeting. The tool is genuinely good and I would not run a job without one. It is an amplifier, and it amplifies in both directions.

Retention and archiving

How long you must keep project records depends on several clocks running at once:

• What your contract says. Many contracts specify a retention period for records, particularly cost records on cost-reimbursable and GMP work, along with audit rights. Northgate's GMP carries an owner audit right; the records that support it must survive it.
• The statute of limitations for contract and construction claims in your jurisdiction.
• The statute of repose — a separate, generally longer outer limit on construction-defect claims measured from completion rather than from discovery. This one surprises people, and it is the reason record retention on buildings runs in years rather than months.
• Regulatory requirements — OSHA record retention, certified payroll on prevailing-wage work, environmental permits and stormwater records, tax records.

All of these vary by state and change over time, and some vary by project type and by funding source. I am not going to give you a number, because a number would be wrong somewhere. Get the retention period for your jurisdiction and your contract from your company's counsel, put it in the document-control procedure at the start of the job, and follow it. The general shape of the answer is longer than you think, measured in years, and measured from completion.

A well-organized archive, at the end, looks like this — and it is assembled continuously, not at closeout:

  NORTHGATE-ARCHIVE/
    01-CONTRACT/         agreement, GCs, SCs, exhibits, bonds, insurance certificates
    02-DRAWINGS/         bid set / permit set / conformed set / all revisions by date
                         + AS-BUILT/ (record set, signed)
    03-SPECIFICATIONS/   project manual + all addenda
    04-SUBMITTALS/       by number: transmittal, submitted content, returned stamped copy
    05-RFI/              by number: request, response, attachments, linked change
    06-CHANGES/          bulletins, ASIs, CCDs, proposals, executed change orders
    07-SCHEDULE/         baseline, every update, every TIA, narratives
    08-COST/             budget, cost reports, pay applications, lien waivers
    09-DAILY-REPORTS/    by date, complete, no gaps
    10-MEETINGS/         agendas and minutes, by meeting type, by date
    11-PHOTOS/           by date and location, with metadata intact
    12-QUALITY/          ITP, inspection reports, test results, NCRs and their closure
    13-SAFETY/           orientations, JHAs, toolbox talks, inspections, incidents
    14-CLOSEOUT/         O&M, warranties, training records, attic stock, CO
    15-CORRESPONDENCE/   by party, chronological

The test of an archive is not whether it is complete. It is whether a person who has never seen the project can find the answer to "what happened on the north elevation between April and June" in under an hour. If they can, you have an archive. If they cannot, you have a storage unit.

🔄 Check your understanding. Your platform's audit trail shows that the level-3 partition plan revision was downloaded by Kestrel's assistant superintendent at 4:12 p.m. on Thursday. Nine working days later a crew has framed 118 partitions from the old layout. Does the audit trail help you?

25.9 🚪 The Paper Trail Is the Project's Memory

🚪 Threshold concept. The paper trail is the project's memory, and contemporaneous records are worth roughly ten times reconstructed ones. A record created on the day, by someone who had no idea it would ever matter, is evidence. The same facts assembled eight months later by someone who very much wants them to matter is argument — and argument loses.

Before you understand this, documentation feels like overhead. It is the part of the day that produces nothing you can see. You write the daily report at 4:55 because somebody makes you, you file the RFI response because that is the process, and when a dispute comes you assume that being right is what matters — that if the work was real, and you can explain it clearly, you will be paid for it. You believe that the truth is in the world and the paperwork is a description of it.

After you understand this, you know that in a dispute the truth is not in the world. The building is finished; the crews have gone; nobody remembers a Tuesday in March. The truth is whatever the record can support, and the record was fixed months ago by people who were not thinking about disputes. You stop writing records to describe what happened and start writing them because the record is the only version of what happened that will survive. You start asking a different question at 4:55 p.m.: not "what do I have to write," but "if somebody reads this in eleven months, what will they be able to prove?"

Here is what that difference is worth in dollars.

CO #14: a documentation autopsy

You know the outline. Meridian's imaging vendor selected a different MRI unit after the GMP was set. The new unit required a deeper depressed slab, added structural framing, additional RF shielding, and a larger electrical feed. Pri Sethi gave a verbal go-ahead on a Thursday. Kestrel's assistant superintendent, under schedule pressure and trying to help, let the concrete crew build it Monday.

No written directive. No agreed price. No time-impact analysis. And no time-and-material tickets for the first four days.

Now the part that is usually left out of the telling, and that is the actual lesson. Here is every cost element, what record existed for it, and what it was worth.

Read the "Record that existed" column top to bottom and a pattern appears immediately. Every row with a zero gap is a row where a third party generated a document as a matter of its own routine. Batch tickets. Supplier invoices. A testing lab's report. A subcontractor's executed change order. Those documents exist because somebody else's business process created them, not because Kestrel remembered to.

Every row with a large gap is a row where the record had to be created by a Kestrel employee, on purpose, in the moment. Rows 6, 4, 9, and 10. Those are the four places where a human being on our payroll had to stop and write something down while the work was happening, and did not.

That is not a moral failing. It is a system design failure. If your documentation depends on people remembering to document under pressure, it will fail under pressure — which is precisely when you need it.

💰 Money check: what each missing record cost.

The settlement recovered $142,750 against $121,000 of substantiation — Pri Sethi paid roughly $21,750 above what Kestrel could prove, because she knew the work was real and because her own verbal go-ahead had started it. But she could not defend more than that to a hospital board without documents. So Kestrel recovered about 33 cents of every dollar of undocumented cost, and lost about 67 cents.

Apply that to the gaps:

Forty minutes of time-and-material tickets was worth $15,480. That is roughly $387 per minute.

And put the $43,650 in context. Kestrel's entire CM fee on Northgate is $1,804,800. One documentation failure on one change order consumed 2.4 percent of the fee for a $47.5 million project — money that was earned, spent, and never recovered, on work that everybody agreed had been performed.

⚖️ What the contract says. Note carefully what did not fail here. The work was real; nobody disputed that. The owner's representative genuinely had authorized it. Kestrel was, in the ordinary sense of the word, right.

Three contractual mechanisms existed to protect Kestrel and none were used:

1. Written notice of a change, within the period the contract specifies, before or immediately upon starting.
2. A construction change directive — the instrument that exists for exactly this situation, where the owner needs work to proceed before price and time are agreed. A CCD would have obligated Kestrel to proceed and established the cost-accounting basis from hour one.
3. The contract's own force-account provisions, which typically specify what records a contractor must keep for directed work performed on a time-and-material basis, and frequently require the owner's representative to sign the daily tickets.

Check your contract for the notice period, the authorized instruments, and who is empowered to direct changes — these vary substantially by form and by owner, and they are frequently stricter on public work. And note the trap in this story: Pri Sethi was the owner's representative, but whether her verbal go-ahead constituted authority to direct a change depends entirely on what the contract says about how changes are directed. On most forms, it did not.

🔄 Check your understanding. Kestrel could substantiate $121,000 and settled for $142,750. Why did Meridian pay $21,750 more than Kestrel could prove — and why does that number not comfort you?

25.10 What to Write, and What Not to Write

This section is the uncomfortable one, and I would rather you be uncomfortable now than surprised later.

The project record will be read by lawyers. Not necessarily yours. Not necessarily on this job. But on some job, some day, a person whose profession is finding the worst sentence in your file will read every daily report, every email, every meeting minute, and every text message you sent about a project, and they will read them in an order you did not choose, out of context, looking for the one that hurts you.

That is not a reason to write nothing. Writing nothing is what cost Kestrel $43,650 on CO #14. It is a reason to write well, which in this context has a precise and learnable meaning.

Three specific prohibitions, each of which I have seen cost real money.

Do not write legal conclusions. "This is a differing site condition." "This constitutes a breach." "We are entitled to an extension." You are not the person who decides those things, and a document in which a project engineer confidently characterizes a legal question is a gift to the other side — especially when the characterization turns out to be wrong, because now every other characterization in your file is questionable too. Write the facts that would support the conclusion and let the conclusion be drawn by whoever gets paid to draw it.

Do not write speculation or blame. "I think Trellis never coordinated this." "Typical Cardinal." Speculation is unreliable and blame is unpersuasive, and both make you look like a participant in a fight rather than a manager of a project.

Do not write sarcasm, and never write anything you would not want read aloud. Sarcasm does not survive the transition from a Tuesday afternoon email to a deposition exhibit. It reads as contempt, always, and it converts a document about a scheduling problem into a document about your attitude.

And one more, which I want to state plainly because the reasoning is not obvious: do not write "this is going to be a disaster." Even when you believe it. Even when you are right — especially when you are right. A contemporaneous document in which the contractor's project manager predicts the failure that later occurs is used to establish that you foresaw the problem and did not prevent it. Your genuine concern becomes evidence of your own notice. If you foresee a disaster, the correct response is not a memo predicting it. It is a written notice identifying the specific condition, the specific impact, and the specific action you are requesting — which protects the project and protects you, and does both better than the memo would have.

A field note, rewritten

Here is a real daily-report entry, cleaned of identifying details.

Before: "Framing in Area B still shut down waiting on RFI 112. Architect has had it forever and still nothing. Total mess up there, guys standing around all week. This is going to blow up the whole level 2 schedule and somebody needs to light a fire under H+P. At this rate we'll never make the drywall date."

Everything in that paragraph is probably true. It is also nearly worthless as evidence and actively harmful as a document. It contains no dates, no quantities, no crew size, and no reference numbers. "Forever" is not a duration. "Total mess" is not a condition. "Somebody needs to light a fire" is an opinion about another party's diligence, written by a person with no authority to hold it, into a document the other party will eventually read. And "this is going to blow up the whole level 2 schedule" is a prediction of failure in a document that will be exhibit 47.

Here is the same day, written by somebody who understands what the record is for.

After: "Area B, Level 2 — metal stud framing. Framing layout at grids 6–9 remains stopped pending response to RFI 112 (issued March 14; required response date March 21; no response received as of today).

Crew: 6 carpenters (Kestrel self-perform), on site 7:00 a.m. to 3:30 p.m. No production in Area B. Crew reassigned to Area D soffit framing 10:00 a.m.–3:30 p.m., approximately 32 of 48 available man-hours redirected; 16 man-hours non-productive.

Cumulative: Area B framing has been stopped since March 21. Six carpenters idled or partially redirected on March 21, 22, 23, 24, 26, and 27 — see prior daily reports.

Area B in-wall inspection remains a predecessor to Level 2 drywall (schedule activity A-2140). Current forecast start for Area B framing is contingent on RFI 112 response.

Actions taken: follow-up notice to H+P sent March 21 and March 25 (copies to file); telephone call with D. Whitcomb March 26, 9:15 a.m., who advised a response is expected March 30 — confirmed by email same day. Superintendent notified of Area D resequence."

Same day. Same frustration, presumably. Every fact preserved, and several added — crew size, hours, redirected hours, the specific schedule activity, the mitigation, the dates of every follow-up, the phone call and the email confirming it.

Now count what the second version can do that the first cannot. It can support a time-impact analysis. It can support a claim for idle labor, with a number: 16 non-productive man-hours that day, and a series of prior daily reports establishing the same pattern. It demonstrates mitigation, which most contracts require and every tribunal looks for. It documents notice, twice, in writing, plus a confirmed conversation. And it contains not one sentence that a lawyer can read aloud to make its author look unprofessional.

Here is the part I want you to notice most: the second version is not more work than the first. It is roughly the same number of words. It is the same amount of typing at 4:55 p.m. The entire difference is that the person writing it knew what the document was for.

🏗️ From the field. I keep one line taped inside the cover of every project notebook I own. Write it so a stranger can use it in a year. Not so your project manager understands it — they were there. Not so it is defensible — that comes for free if the facts are there. So a person who has never met you, never walked the job, and has no memory of March can pick it up, read it once, and know what happened.

Every good record I have ever written passed that test. Every bad one failed it in the same way: it assumed a reader who already knew.

Spaced Review

Answer these before reading on. Cover the answers.

1. From Chapter 23 — the submittal-to-installation chain. Most projects break the chain between an approved submittal and what actually gets installed. Where does it break, and what closes it?

It breaks at the work face. The submittal is approved in an office and filed, and the crew installing the product has never seen it — so the product gets installed per habit, per the sub's standard detail, or per whatever showed up on the truck. What closes it is the preinstallation conference plus a physical verification: the approved submittal is present at the work face, the material delivered matches the approved submittal, and the first installation is inspected against it before production continues. An approved submittal that nobody at the installation reads is a piece of paper that documents an intention.

2. From Chapter 16 — back-scheduling long leads. What is the one fixed point in a procurement chain, and what is the most common mistake in identifying the release gate?

The fixed point is the required-on-site date, and everything else is computed backward from it. The most common mistake is tracking your approval date rather than the fabricator's release date. On Northgate those were April 15 and April 18, three days apart, and only the second one was real. Ask every long-lead supplier, in writing, at buyout: what is the last date you can release, and what happens if you miss it?

3. Deep callback to Chapter 7 — order of precedence and how a set changes. Two questions. First: is there a universal order of precedence in construction? Second: name three instruments that change a set after the contract is signed, and say what each implies about cost.

First: no. There is no universal order of precedence. It is whatever your contract says, and contracts say materially different things — complementary with no hierarchy (the architect interprets), a stated ranked hierarchy, or "most stringent governs." Under the complementary regime, which is the most common, you cannot self-resolve a conflict. You write the RFI. Building your preferred reading and arguing about it later is how a $96,000 insulation conflict becomes a $96,000 loss.

Second: an addendum (pre-bid; priced in your number), an ASI (the architect's position that no cost or time is involved — evaluate it, do not accept it reflexively), a bulletin or proposal request (price it, do not build it), a CCD (build it, track every hour from hour one), and a change order (agreed and executed). Notice that three of those five carry cost consequences the instrument's own name understates. That is why §25.8 says read every revision block for its source.

Project Checkpoint: The Willow Street Submittal Log, RFIs, and Document-Control Procedure

In Chapter 24 you built the site-specific safety plan — the hazard analysis, five JHAs, the orientation, and four toolbox talks. This chapter builds the project's information spine. Four deliverables.

Part 1 — The full submittal log, back-scheduled. Go through the Willow Street project manual in Appendix K and extract every required submittal from Part 1 of each specification section. Expect somewhere between 90 and 150 items on a $6.8M, 24,000 SF building.

Then do the thing this chapter is about. For each item, pull the required-on-site date from the CPM schedule you built in Chapter 14, and back-schedule: subtract installation lead, delivery, fabrication from release, the approval cycle with one resubmittal allowance included, and the subcontractor's preparation time. Record the submittal due date and the award-by date.

Then sort the whole log by submittal due date and never look at it in specification order again.

Your columns, at minimum: number · spec section · description · type (shop drawing / product data / sample / mockup / mix design / certification / delegated design / O&M / warranty) · responsible party · reviewer · required on site · release gate · submittal due · award by · status code · ball-in-court · days aging · gates what.

Part 2 — The three longest-lead items. Identify them, and show the full back-schedule arithmetic for each as a table. On a building like Willow Street the candidates are the structural steel and open-web joists, the gymnasium maple flooring system, the rooftop units and their curbs, the electrical switchboard, the commercial kitchen equipment package including the grease interceptor, and the folding partitions. State the submittal due date and the award-by date for each, and — this is the point of the exercise — flag any whose award-by date has already passed in the schedule you built.

Part 3 — Three complete RFIs. Go back to the ten discrepancies you found in your Chapter 7 discrepancy log and pick three that are genuine ambiguities rather than questions the documents answer. Write each RFI in full, with all six elements: reference, observed condition stated factually, the specific question, your proposed solution with a reason, the quantified impact and the date it applies to, and the required response date. Then write, for each, who you would copy and why.

Part 4 — The document-control procedure. Two pages, no more, written as though a new field engineer starts Monday and this is what you hand them. It covers: numbering conventions for submittals, RFIs, and transmittals; the current-set rule and how superseded sheets are handled in the field office and at the work face; the transmittal process; the distribution matrix as a grid; the as-built markup routine and the pay-application condition that enforces it; the filing and archive structure; and the retention period — which you will state after actually checking your jurisdiction's requirements and your contract, not from memory.

Next chapter turns this record from documents into a routine: the meeting structure, agenda templates, five completed daily reports, and the photo-documentation protocol that makes the daily report worth reading.

Chapter Summary

A reference table. The left column is what you do; the right is what it is worth.

The one sentence, if you remember nothing else: a submittal log is a schedule, an RFI is a request for a decision you have already proposed, and a record made today is worth ten of the same record made in August.

What's Next

Chapter 26 takes the second information system: the meetings that actually run a job and the daily report, which is the single most valuable document a project produces and the one most often filled out badly at 4:55 p.m. by somebody who wants to go home. You have just seen what a well-written field note is worth in a dispute — the next chapter is about producing one every day, for eighteen months, on a job with a hundred and forty people on it.

Then Chapter 27 takes the third: the constraint log, and the reason a CPM schedule can be perfectly correct and still send a crew to a wall that is not ready.