Most pilots do not fail. They expire. Nobody cancels them, nobody renews them, the champion changes jobs, the budget cycle turns over, and eight months later you are explaining to someone why there is a login for a tool nobody remembers buying.
That is the actual failure mode, and it is not a product problem. A pilot that dies quietly almost always died because nobody wrote down, in advance, what would have to be true for it to become a contract. This playbook fixes that. It works whether you are a freelancer running a paid trial, a startup running a 90-day proof of concept, or an internal team piloting a tool you want budget for next year.
The mental model: a pilot is a different product
The mistake is treating the pilot as a small version of the contract. It is not. They have different buyers, different success conditions, and different failure risks.
- The pilot is bought by a champion. One person who is curious, has discretionary budget or discretionary time, and wants to look smart. Their success condition is "this was interesting and did not embarrass me."
- The contract is bought by a committee. Finance, procurement, security, and the champion’s boss. Their success condition is "this is a defensible line item next year."
Those two things are almost unrelated. A pilot can delight the champion completely and still produce nothing a committee can approve. Your job during the pilot is not only to make the thing work. It is to manufacture the evidence the second buyer will need — while you still have access to the first one.
Part 1 — Qualify before you pilot: the four questions
Half of dead pilots should never have started. Ask these four before you agree to one. If you cannot get a clean answer to all four, you are running a demo, not a pilot, and you should price it as such.
- What is the budget line this comes out of if it works? Not "we'd find budget." The name of the line. If nobody knows, there is no contract at the end of this, only a fond memory.
- Who signs? Get the title. If your champion says "me," ask what the approval threshold is above which it stops being them.
- What is the decision date? Anchor to their calendar, not yours — budget cycle, renewal date of the incumbent tool, a board meeting. "When it's ready" is not a date.
- What happens to the current way of doing this? If the answer is "it keeps running alongside," you are not replacing anything and your value will always be additive-nice-to-have, which is the first thing cut.
Script: "Before we scope this, I want to make sure I'm building toward something real. If this works exactly as we both hope, what's the path to it becoming a normal line item — who signs, out of what budget, and roughly when?"
Nobody has ever been offended by that question. Plenty of people have been unable to answer it, which is the information you wanted.
Part 2 — Write the exit criteria before day one
This is the whole playbook in one move. Before any work starts, you and the champion co-write a single page that says what would make this a yes. Not a proposal. A page. Send it as an email so it is timestamped and quotable.
The template, verbatim:
- What we are testing: one sentence, in their words, not yours.
- The number: one primary metric with a baseline and a target. "Ticket resolution time drops from 34 hours to under 20."
- The guardrail: one thing that must not get worse. "No increase in reopen rate."
- The window: start date, end date, and the review date — which is at least two weeks before the budget deadline, not after.
- Who is in the room at review: names. Include at least one person who is not the champion.
- What happens if we hit it: the specific next step. "We move to a 12-month agreement at the scoped price." Say the number.
- What happens if we don't: also specific. "We stop, and I send a written summary of what we learned." Having a clean no makes the yes credible.
Two rules. The metric must be one they already measure — if you have to build the measurement, you have added a project to your project and you will be arguing about the instrument instead of the result. And the target must be a number the champion said out loud. Numbers you invent are your numbers. Numbers they say are theirs, and people defend their own numbers in committee meetings.
Part 3 — Instrument on day one, not day sixty
Capture the baseline before you change anything. This sounds obvious and is skipped constantly, because on day one everyone is excited about setup and nobody wants to spend a session pulling boring historical data.
You need three things in a shared document before the first change ships:
- The baseline number, with the date range it covers and where it came from. Screenshot it.
- The comparison period you will use at review — and make it apples to apples. Comparing a December pilot to a November baseline will get you destroyed by anyone in finance.
- Two or three named humans who will use the thing and can be quoted at the review. A committee discounts your metrics and believes their own people.
If your champion resists the baseline work, that is your first real signal. It usually means the number is embarrassing, which is good news for you, or that nobody actually tracks it, which means you are back to Part 1.
Part 4 — The weekly rhythm
Fifteen minutes a week, same slot, non-negotiable. Not a status meeting — a evidence-building meeting. The agenda never changes:
- Where the number is (30 seconds, one line, sent in advance).
- One thing that worked, with a name attached. "Priya cleared her queue by 2pm Thursday for the first time."
- One thing that is friction, with what you are doing about it. Volunteering problems is the single fastest way to be trusted.
- One question for them. Rotate through: who else should see this, what will procurement ask, what would your boss push back on.
That fourth item is the one that converts. Over eight weeks it quietly recruits the second buyer, surfaces the objections early enough to answer them, and turns your champion into someone who has been rehearsing the internal pitch for two months. By review day they should not be hearing your case. They should be delivering it.
Part 5 — The conversion conversation
Hold it two weeks before the deadline, with the non-champion in the room. Do not open with the demo. Open with the page you wrote in Part 2.
Script: "In March we agreed this was a yes if resolution time got under 20 hours without reopens going up. It's at 16 and reopens are flat. Priya and Marcus both said they'd push back if we turned it off. So I think the honest read is that we hit it. What I'd like to agree today is the shape of the next twelve months — and I'd rather hear the objections now than in procurement."
Then stop talking. The most common mistake here is filling silence with more features. You are not selling any more. You are asking them to confirm a decision they already made in Part 2.
If they hedge, the hedge is information. There are only three real hedges:
- "We need to see more." Means the metric did not land emotionally. Ask which single additional data point would settle it, get that one, and re-book. Do not extend the pilot open-endedly — extend it by a specific number of days with a specific new criterion.
- "Timing is bad." Usually true and usually solvable. Ask for the paperwork now with a start date later. A signed agreement with a January start beats an enthusiastic verbal in September.
- "I need to check with X." You missed a stakeholder. Ask to be in that conversation, or at minimum to write the one-pager X will actually read.
Part 6 — The six ways pilots die
- Free. Unpaid pilots convert far worse than cheap ones, because nothing that costs nothing has an owner. Charge something, even if it is nominal, and even if you credit it against year one.
- No end date. An open-ended pilot has no forcing function, so it never reaches a decision. It just fades.
- Scope creep as flattery. "Could it also do X?" feels like buying signal. It is usually the pilot expanding past its own success criteria so that nothing ever gets proven. Say yes to X after the contract.
- Single-threaded. One champion is one job change away from zero. If by week four you have not been introduced to a second person, make that the ask.
- Vanity metric. Usage is not value. "Logins are up" survives no committee. Time, money, or error rate.
- The silent extension. The review slips a week, then a month, then the quarter ends. Treat the first slipped review as a red alert, not an inconvenience.
A worked example
A two-person consultancy pilots a document-review workflow with a mid-size insurer. Old approach: three analysts, average 11 days per file. They write the page in Part 2 — target under 6 days, guardrail no rise in exception rate, review on 14 October because budget locks on 1 November, and the room includes the operations lead as well as the champion.
They screenshot the baseline on day one. Week three, an analyst named Dan says out loud that he stopped dreading Mondays; they write that down verbatim. Week five, the champion says procurement will ask about data residency, so they answer it in writing five weeks before anyone formally asks. Week eight, files are averaging 5.2 days and exceptions are down slightly.
The review takes eleven minutes. Nobody is surprised by anything, because nothing has been saved for the ending. That is the goal state: a conversion conversation should be boring.
Your first week
- Day 1. List every pilot, trial, or proof of concept you currently have running. For each, write the decision date and the signer. Any blanks are your priority.
- Day 2. Send the Part 2 page to every pilot missing one — yes, even mid-flight. "I want to make sure we're measuring the right thing" is a perfectly good reason to send it in week six.
- Day 3. Capture or reconstruct baselines. If a baseline is genuinely unrecoverable, say so now rather than at review.
- Day 4. Book the weekly 15 minutes for the remaining duration of every pilot. All of them, in one sitting.
- Day 5. For every single-threaded pilot, ask for one introduction.
None of this makes a mediocre product good. What it does is make sure that a good result actually turns into a signature — which, judging by the number of pilots that quietly expire every quarter, is the harder half of the job.