Case Study 2 — Five Hundred Sixty-Five Thousand Dollars: Three Years of Kestrel's Technology Budget, Reviewed Honestly

All people, companies, and projects in this book are illustrative composites. Every dollar figure is teaching data, consistent within the book and not drawn from any actual project.

Setup

Case Study 1 looked at one decision in depth. This one does the opposite: it looks at every decision Grace Lindqvist made with Kestrel's discretionary technology money over three years, and asks a question almost nobody asks out loud.

Which of these are we still doing?

The occasion. Kestrel Construction Group runs a three-year capital and overhead planning cycle. Nadia Haddad, vice president of operations, asked Grace to bring the whole ledger to the fourth-quarter review — not the plan for next year, the history. Every line, what it cost, and whether it is still running.

The people.

• Grace Lindqvist — VDC/BIM manager; owns the technology budget and kills pilots quarterly
• Nadia Haddad — vice president of operations; the one who asks the question
• Margo Deacon — general superintendent; owns the layout crew, which turns out to matter
• Bea Salgado — corporate safety director
• Jamal Foster — self-perform concrete superintendent
• Ray Alvarez — senior project manager; in the room because he has to use whatever survives
• Owen Baptiste — chief financial officer; wants one number and gets a worse one

The scope. Discretionary technology spend only. Kestrel's field management platform licenses are excluded, because that is not a decision anymore — it is plumbing, like phone service. What is left is nine line items over three years.

What Happens

The ledger

Grace put it on one page, which was itself the bravest thing that happened in the meeting.

Owen Baptiste did the arithmetic before anybody else in the room.

"Four of nine still running. Two hundred seventy-three thousand six hundred on things that survived, two hundred ninety-two thousand on things that didn't. You spent slightly more than half of it on things we stopped doing."

"Yes," Grace said.

Nadia let that sit for a second and then said the thing that made the meeting useful instead of humiliating.

"That's the right ratio. A technology program where everything works is a program that never tried anything, or a program that isn't telling us the truth about the results. What I want to know is whether the six failures rhyme."

They did.

The Two That Quietly Became Indispensable

Line 1 — Reality capture: $186,000, and nobody has an opinion about it

This is now the single most boring item on Grace's agenda. Last quarter's renewal took one line and forty seconds because nobody in the room had thought about it in eighteen months.

Here is what it did in the same period. Eleven scanned scopes across four projects. Six of those scans caught something — a beam that was not where the model said it was, an existing chase two inches narrower than the survey drawing, a run of abandoned conduit somebody was about to hang a duct through. Aggregate avoided rework, estimated conservatively: $214,000. On Northgate specifically, the 360-degree walkthrough program cost $13,600 over nineteen months, in a $47.5M job, and the record was used substantively eleven times.

Why it survived, mechanically. Kestrel bought it, Kestrel operated it, and Kestrel captured the benefit. No subcontractor had to change anything. No foreman typed anything for somebody else's report. The deviation map went back to the same team that produced the scan. And there was always a named person who looked at the output — which is the only place this technology is exposed, and Grace protected it deliberately.

Why it is invisible. Because avoided rework never appears in a cost report. There is no line item called the duct we did not have to cut down and rehang. A technology whose entire value is a subtraction from an event that never happened will be undervalued by every accounting system you will ever work with, and you have to defend it with a log instead of a ledger. Grace keeps that log. That is why it renews in forty seconds.

Line 8 — Equipment telematics: $14,400, nearly cancelled by accident

Grace put this on the kill list in year two. It was a small aggregation seat that pulled engine hours, idle time, fuel burn, location, and fault codes from Kestrel's owned fleet and from three rental houses into one place. Nobody in VDC used it. Grace's reasoning was sound and her information was incomplete.

Kestrel's equipment manager found out the week before the review and came to the meeting.

The idle-time report had become the basis for the entire equipment budget. Actual fuel burn per hour had replaced the spec-sheet figure that produced a $41,000 miss on a previous job. Rental invoices were being checked against engine hours. And fault codes were driving scheduled maintenance instead of breakdowns — which matters because a machine down is one line item and the crew waiting on it is four. On Northgate, an idle crew of six at a burdened rate near $54/MH costs $324 an hour to watch.

"It's fourteen thousand dollars," the equipment manager said. "It saved more than that on one excavator."

Why it survived, mechanically. It removes work rather than adding it. Nobody types anything; the machine reports itself. The party that owns the machine is the party that gets the benefit. And every output landed on a decision that already existed and already had an owner.

The lesson Grace took from nearly killing it: the technologies that work are the ones you stop noticing, and the ones you stop noticing are the ones you are most likely to cancel. She now keeps a one-line "what decision does this feed, and whose" note against every running item, refreshed annually, in the technology budget file. It costs her an afternoon a year.

The Two That Consumed Budget and Produced Dashboards

Line 3 — Portfolio risk analytics: $142,000 over two years

A platform that ingested schedule, cost, and RFI data from fourteen active projects and produced a weekly red / amber / green score per job, with drill-downs, trend lines, and a portfolio heat map. It worked. It was not buggy. The integrations held.

After a year, every project executive said the same three things: the reds were projects they already knew were red, the greens told them nothing, and nobody had ever changed a decision on the strength of a score.

Nadia asked the question she always asks. "Name one thing we did differently."

Nobody could.

Why it failed: the Unread Dashboard. It produced information, and there was no decision waiting for that information. A project executive with fourteen jobs already knows which three are in trouble — she knows because Ray called her, because a schedule update slipped, because a superintendent's voice changed on the phone. The score was a second, slower, prettier copy of knowledge that already existed and had already been acted on.

The subtler failure underneath it. The score was computed from schedule and cost data whose quality varied by project. On a job where the schedule was updated honestly and the cost codes were disciplined, the score reflected the job. On a job where the schedule had not been meaningfully updated in six weeks, the score reflected the schedule update habit of the project team, and it reported green. The system did not distinguish between a healthy project and an unmeasured one, and it was the unmeasured ones Nadia actually needed to find.

Line 6 — Wearable safety sensors: $32,000 for the pilot, $96,000 a year to deploy

One hundred eighty devices on belts, two projects, four months. Location, proximity to equipment, fall detection, an analytics platform. The dashboard was genuinely beautiful — heat maps of worker density by hour, a proximity-event counter, trend lines.

2,100 proximity events logged. Zero changes to a work plan, a job hazard analysis, a route, a gate, a laydown area, or a crew size.

Bea Salgado, who had been in every review of it, did not hedge: "It logged twenty-one hundred proximity events and I could not tell you which ones were hazards and which ones were two guys walking past a skid steer. It is a very good record of a job I already had eyes on."

Nadia: "Then it's a camera pointed at your own memory. Grace, kill it."

Why it failed: the Unread Dashboard again, and it is worth being precise about the mechanism, because "it didn't work" is a useless conclusion. Bea did not need a count of proximity events. She needed a route change — and a route change requires a superintendent, a laydown decision, and two subcontractor conversations. The device measured a problem that was already known and did not touch the thing that would have fixed it.

And it carried a cost that was not on the invoice. Most of the people wearing those devices did not work for Kestrel. Kestrel was proposing to monitor another employer's employees, all day, including breaks. Margo Deacon spent real credibility on that pilot explaining to foremen what the data was and was not for. That trust does not appear anywhere in the $32,000, and it was the most expensive thing the pilot spent.

The One That Failed for a Purely Organizational Reason

Line 2 — Model-based layout: $67,000, idle for five months, then indispensable

This is the important one, and it is the one that changed how Grace evaluates everything.

The instrument takes coordinates directly from the coordinated model and marks them on the deck — partition layout, sleeve locations, hanger and anchor points, floor penetrations, equipment pads. Instead of two people pulling tape from a control line, one person walks a prism to each point. It is faster, it uses fewer people, and — the part that matters most — the marks agree with the model, which means the hangers the mechanical subcontractor installs land where the coordination said they would.

Nothing about that is in dispute. It is arguably the most reliable robotics win in building construction.

Kestrel bought one and it barely got used for five months.

Not because it did not work. Because the layout crew reported to Margo Deacon and the model was owned by Grace, and nobody owned the daily handoff of the layout file. The crew needed tomorrow's deck file by the end of today. Grace's technician spent his afternoons answering clash questions from three project managers with live coordination problems. There was no standing time, no named person, and no deadline. So the layout file was late, or partial, or arrived after the crew had already planned its morning — and the crew pulled tape, because tape is always available and never late.

Look at what that actually is. The cost of producing the file landed on VDC. The benefit of having it landed on the field. Neither department was behaving badly; each was rationally prioritizing its own queue. There was no contract between them, no price, and no clock.

That is the same misalignment that kills technology between companies, occurring between two departments of one company. It does not require a fragmented supply chain. It only requires that the person doing the work and the person getting the value are not the same person and have no mechanism connecting them.

The fix cost nothing. One named person responsible for layout files. A standing Wednesday deadline for the following week's decks. A two-line item on the coordination meeting agenda so a miss was visible to Margo and Grace in the same room. Utilization went from occasional to daily inside a month.

Ray's assessment, and he includes himself in it: "I was one of the people who said that thing doesn't work. I said it in a meeting. I was wrong in a specific way that is worth naming — I evaluated the machine and the problem was the org chart."

Analysis: The Pattern

Scored on the framework, after the fact

Grace went back and scored all nine on §39.9.2's four dimensions — A: value concentration, B: adoption surface, C: work displacement, D: decision linkage.

Read the bottom four rows first. Every one of them failed exactly where the framework predicted. The materials-tracking system scored 1 on adoption surface because it required fourteen subcontractors' suppliers' truck drivers — people with no contract with Kestrel and no reason to stop at a gate podium — to change their behavior for Kestrel's benefit. After eight months, roughly a fifth of deliveries were being scanned. The field productivity module scored 1 on work displacement because foremen typed for twenty minutes at the end of a ten-hour day and got nothing back; entries stopped after six weeks. Both dashboards scored 1 on decision linkage, which is a gate, not a score — and both were killed on the gate rather than on the total.

Mapped onto the four failure modes

The three findings

Finding one: the survivors all have the same shape. One party adopts alone. That party captures the benefit. The technology removes work at the point of use rather than adding it. And a named person makes a named decision differently because of the output. Reality capture, telematics, and maturity sensors are three completely unrelated technologies with identical structural profiles, and that is not a coincidence — it is the only profile that survives a fragmented, one-off, low-margin industry.

Finding two: the two most expensive failures were the two best-looking ones. The wearable dashboard and the portfolio risk platform together cost $174,000 and were, by a wide margin, the most impressive things anyone demonstrated to Kestrel in three years. The two cheapest survivors — telematics at $14,400 and maturity loggers at $6,200 — produce no picture at all. A technology that generates a beautiful visual is optimized to be bought. That is not a conspiracy; it is what the sales process selects for. Be suspicious of the demo that looks best.

Finding three, and this is the one Grace wrote on the front of the file: the two highest-scoring failures both failed the same gate, and it is the gate everybody skips. The rebar robot scored 17 and could not address enough of a building's scope to matter. The layout robot scored 18 and sat idle because no one owned a daily handoff. Both are gate question 5 — what has to be true here before it works at all, and who owns it on a Tuesday when the champion is on vacation.

Questions 1 through 4 are about the product. Question 5 is the only one that is about you, which is exactly why it is the one that gets waved through in the meeting. The vendor cannot answer it. The demo cannot answer it. Nobody in the room is embarrassed by leaving it blank, because it looks like a formality.

It is not a formality. It is where two-thirds of this book's technology money went.

Discussion Questions

1. Grace almost cancelled a $14,400 line that the equipment manager had built an entire budgeting practice on. Design the specific mechanism that prevents that mistake without adding a bureaucratic review to every renewal. What does your mechanism cost, per year, in hours?
2. Nadia called a 4-of-9 survival rate "the right ratio." Argue the opposite: at what survival rate would you conclude the evaluation process is broken rather than healthy, and in which direction? What would a 9-of-9 rate tell you?
3. The portfolio risk platform reported green on projects whose schedules had not been meaningfully updated in six weeks. Is that a flaw in the platform, a flaw in Kestrel, or both? Write the one sentence you would have added to the evaluation before purchase that would have surfaced it.
4. The layout robot's failure was that the cost landed on VDC and the benefit landed on the field. Kestrel fixed it with a named owner and a Wednesday deadline. Name two other places in a construction company where the same internal misalignment exists, and say what the equivalent fix would be.
5. Bea Salgado's wearable pilot was well run and produced a clear negative result. Grace killed it and Nadia credited Bea publicly. What happens to a technology program at a company where a kill lands on the champion's performance review? Describe the failure mode in terms of the data you would subsequently receive from pilots.

Your Turn

Build your own version of Grace's page.

Pick any organization you can actually see inside — your employer, a previous employer, a school department, a volunteer organization, a family business. List every tool, subscription, device, or system it has adopted in the last three years that cost real money. You will probably find between four and twelve.

For each one, fill in five columns: what it cost, is it still genuinely used, who pays, who benefits, and name the person and the decision the output feeds. Where you cannot fill in the last column, write "none" — do not soften it.

Then score each on A/B/C/D and answer two questions in writing, in under 200 words total:

1. What fraction of the money went to things that are no longer used, and does the survivors' shape match Finding One?
2. For the single most expensive item still running, write the gate-5 answer: who owns it on a Tuesday when the person who championed it is on vacation? If you cannot name that person, you have found the next failure before it happens, which is the entire point of the exercise.