How do you price a pilot, and should you charge at all?

How to answer this, stage by stage

this is a should-you-charge tradeoff, not a range of options, so PICK carries it. commit to a position, then earn it.

Let's learn

What does a general contractor actually owe you for six weeks with a tool they haven't decided to trust yet?

Bidframe is a tool a construction-tech company called Trussgate sells to general contractors. It reads a project's plans and spec sheets, checks them against thousands of past bids, and hands an estimator a starting price for every line item, along with a note on which numbers it's sure about and which ones need a second look before anyone submits them.

For its first two years, Trussgate ran every pilot free. No fee, no contract, just a shared folder and a kickoff call. It worked, because in a market with no track record yet, free was the only way to get a contractor to open the folder at all.

It stopped working the way it used to. Of the last nine free pilots, six never had a single real bid run through the tool. A contractor's estimating lead would agree on the call, sound genuinely interested, and then go back to the bids that were actually due that week. The free tool sat there, untouched, until the six weeks ran out.

Here is the turn. The wasted sales calls are not the real problem. The real problem is what happens when a team reacts to that by protecting itself the wrong way: charging everyone the same fixed price with no room to think, and nearly losing a contractor whose bid mix was exactly the kind of project the model had almost never seen.

At its worst, this costs more than thirty five wasted hours here and there. A team that reacts to free-pilot fatigue by charging a flat, uniform fee with no lever to pull can end up shipping Bidframe to more contractors with a tool that sounds confident everywhere, while quietly guessing wrong on any project type it was never actually trained to see.

What I would leave alone. Support for a contractor who has already signed should still be fast and free, this is only about the unsigned trial window. A tiny, self-serve trial, upload one project, no calls, no integration, can also stay free forever, because it costs almost nothing to run and it was never going to fill the golden set anyway.

The lesson: a price on a pilot was never there to make money on the pilot. It is there to sort the contractors who will actually run real bids through the tool from the ones who signed up because it was free and it was Tuesday.

One side of this costs a timesheet entry. The other side costs a whole project type nobody ever measures the model against.

The week Stonebrace almost said no

the short version sits above. read on for what it actually sounds like when a contractor pushes back on a fee, and what almost got missed.

Every Tuesday morning for two years, Obiora Uzoma ran the same call at Trussgate: a kickoff, a shared folder link, a promise to check back in six weeks. He runs pilots for Bidframe, and for most of that time the process barely changed, because it barely needed to.

In the early years, free pilots were the entire way Trussgate got anyone to try Bidframe at all. The first dozen contractors who tried it free went on to sign real contracts, because in a market with no track record, free was the only door anyone would walk through.

As Bidframe's name spread, the pilots stopped meaning the same thing. Some contractors wanted a free peek and nothing more. Some had an in-house estimator quietly benchmarking their own numbers against Bidframe's, with no plan to ever pay for it. And some simply meant to try it and then got busy the day the pilot started, the way anyone does with something that cost them nothing to say yes to.

The sixth of nine, a general contractor called Colbeck Builders, agreed enthusiastically on the kickoff call and then never opened Bidframe once in six weeks. At week six, Obiora pulled the usage log. Zero projects uploaded. Nothing to point to, good or bad.

That week, Obiora set a flat fee: nine thousand dollars for any new pilot, applied the same way to everyone, no exceptions. It worked exactly as intended. Pilot volume dropped by more than half. Of the pilots that did sign up and pay, seven of eight ran at least one real bid through the tool inside six weeks, and three converted to paying contracts.

Then Stonebrace called.

Stonebrace Construction is a mid-size general contractor with a real specialty: nearly half its work is modular and prefabricated building, an unusual mix for a company its size. Radek Marchetta, Stonebrace's chief estimator, had heard about Bidframe from a competitor and sounded interested from the first minute of the call. He was also plain about the fee. "We don't pay to test-drive software," he said. "Prove it works, then we'll talk about money."

Following the flat rule exactly as written, Obiora held the line. Radek said he'd think about it. Two weeks went by with nothing back.

What broke the silence was a routine check, not a change of heart. Trussgate ran a quarterly audit of its own golden set, the collection of hand-checked bids used to grade Bidframe's accuracy. Modular and prefabricated construction made up about 46 of roughly 1,200 examples in it, under four percent, next to nothing next to the amount of ordinary stick-built commercial work in there. Obiora already had the bid list Radek had shared during the sales call sitting in his notes. Stonebrace's own annual bid volume ran about 45 percent modular.

Obiora went back to Radek. Not with a quiet discount dressed up as a favor, but with the actual reason named out loud: "Your bid mix is exactly the thing our model is thinnest on. I'll waive the fee for this pilot, but I need your estimators to flag every place Bidframe's number looks wrong, especially on the modular jobs." Radek agreed, mostly because it read as a real trade, not a favor he had to feel grateful for.

The same picture from Section 1, worth holding onto here. Stonebrace was standing in the box on the right until somebody checked the golden set.

Stonebrace ran 22 real bids through Bidframe over the six weeks. Radek's team corrected nine of them by hand. Seven of those nine corrections were modular jobs, where Bidframe's line item price had come in more than 15 percent off. Every one of those nine corrections became a new golden-set example. Stonebrace signed a contract after the pilot ended. For the next twelve prospects who came in without that same kind of data gap, the nine thousand dollar fee stood exactly where it had, and the no-show rate on paid pilots held close to what it had already been.

The old decision, told as a memory of an actual afternoon: the flat, no-exceptions fee was set in a fifteen-minute call with Trussgate's head of sales, right after the Colbeck usage log came back at zero. "No exceptions" was the exact phrase used, because the team had just watched a looser policy quietly turn into forty free pilots a quarter. It made complete sense in that room. It stopped working the moment it also blocked the one contractor whose bids the model actually needed.

One design protects you from every prospect who wants a favor. It cannot, on its own, tell the difference between a favor and the one contractor you actually need to pay to keep.

What I would tell myself, back in that fifteen-minute call: you built a filter for commitment. You forgot to build a lever for the one thing you were actually pricing the pilot to get.

PICK, run against the Stonebrace fee

this is a commit-then-defend question, so PICK carries the weight, not a list of pros and cons.

The alternative I ruled out: keep every pilot free forever, on the idea that more trial volume beats a filtered, paid funnel. That was Trussgate's actual first two years, and nine tries at it already proved the problem: six of nine free pilots produced zero real bids and zero usable data, only a longer list of contacts who'd once said yes on a call.

This is also the exact spot where the answer has to stay AI shaped, not just a pricing story. The real risk sitting behind all of this is Bidframe recommending a confident number on a job type it has barely seen, and an estimator trusting it without a second look. The guardrail already lives in the product, not just the pilot terms: Bidframe never auto-fills a line item below a calibrated match-confidence bar against the golden set, it routes that number to the estimator for manual review instead. During a pilot, how often that flag fires on a new project type is itself a live reading of a coverage gap, which is exactly what happened with Stonebrace's modular jobs.

Try it somewhere else, at an insurance MGA

Havelock Analytics builds Riskline, a copilot that reads a commercial insurance submission package and recommends whether to quote the policy, and at what price, for the managing general agents who write that business.

Solvei Kettlewell runs pilots for Riskline the way Obiora runs them for Bidframe. Every new MGA pays a modest monthly fee to pilot Riskline against its own submission flow, by default, no exceptions until the golden set says otherwise.

P. Charge a flat monthly pilot fee for every MGA, by default. Waive it only for an MGA whose book leans heavily into a coverage class Havelock's golden set is thin on.
I. A free pilot that fizzles costs Havelock's underwriting reviewers time on setup calls that go nowhere. A paid MGA lost over price costs an entire risk class's worth of real accept-or-decline examples, the kind that only shows up when someone actually underwrites that class.
C. The wasted setup hours are cheap and visible, caught within weeks. A missing risk class is quiet and expensive: it surfaces later as a bad quote recommendation on a real policy, in a class nobody had enough examples of to catch it sooner.
K. Waive the fee when an MGA's book sits above 25 percent in a coverage class where Havelock's golden set has under 40 examples. Fenmore Underwriters writes about 30 percent environmental liability against a golden set with only 31 examples of that class, clearing both numbers, so its pilot runs free.

Swap the trigger and it still runs.
Speed: an interviewer cuts you off after ninety seconds. Skip straight to the position and the kill number: charge by default, waive only above the coverage-gap threshold.
Cost: say the solutions-engineering team can't staff the extra onboarding calls a paid pilot needs without cutting support elsewhere. The pick doesn't move, the fee is exactly what buys that staffing time back, by filtering out the pilots that would have wasted it anyway.
The model got better, for real: suppose Bidframe's accuracy on modular jobs already matched its accuracy everywhere else. That's the actual evidence that flips the kill criteria off, once modular coverage clears the bar, Stonebrace's bid mix stops being scarce data and the fee goes back to standing like anyone else's.

Where people run it wrong.
Treating a waived fee as a favor for being nice, instead of naming the specific data gap it's buying, so every later prospect assumes they can ask nicely too.
Keeping every pilot free forever because a slower sales cycle looks bad on a pipeline chart, while the real cost is a folder full of pilots nobody ever opened.
Setting one fee at launch and never revisiting it, so a number that once filtered for commitment quietly becomes too cheap to filter anyone, or too high for the exact segment you need.

How to use it live. Say the reframe out loud before answering with a rule of thumb. "Give me a second, I want to check whether this prospect's data is scarce for us or just more of what we already have." That's true, it's already the kill-criteria step, and it buys you the time to find the real number instead of reciting charge always or never charge.

Flashcards (click a card to flip it)

Check yourself Score: 0 / 0