"Let's Start With a Pilot" Sounds Like a Yes
It arrives near the end of a good meeting. They like the idea, the room is warm, and somebody says: why don't we start with a small pilot and see how it goes.
You leave feeling like you won something. What you have actually agreed to, in most cases, is to do a portion of the work before anyone has committed to buying anything, on a timeline nobody set, against a definition of success nobody wrote down.
That is not a sale. It is a second sales process, and you just volunteered to fund it. The same way a good meeting followed by a brochure is not a deal either.
None of which means refusing. Pilots are genuinely useful, and for a cautious buyer they are often the only honest way through. But a pilot that converts and a pilot that quietly dies look identical on the day you agree to them. The difference is decided entirely by what gets settled before the work starts.
Three Words, Used Interchangeably, Meaning Different Things
Before anything else, work out which one is on the table, because most of the confusion in these conversations comes from the two sides using the same word for different arrangements.
A trial is the loosest. The customer uses the thing in their own environment, with or without your help, to get hands-on with it. There is no particular claim being tested. They are having a look.
A proof of concept adds a hypothesis. Before it starts, both sides agree on what is being proved. That single addition is what separates a POC from a trial, and it is the reason a POC can end in a decision rather than a shrug.
A pilot, properly used, adds purchase intent. The customer agrees in advance that if the thing works as specified, they will buy. That is what makes it a pilot rather than a long free look.
The reason this matters is not vocabulary policing. It is that buyers routinely ask for a pilot and mean a trial. They get the seriousness of the word without the commitment behind it, and you do the work of the third while being paid like the first.
So the useful move when somebody proposes a pilot is to ask, warmly and directly, which of these they mean. If it works, are we buying? The answer tells you what the next month is actually for.
If Nobody Wrote Down What Counts as Success, It Cannot Succeed
This is the failure that swallows most pilots, and it is entirely preventable.
Ask a firm mid-pilot what success looks like and you tend to get a muddle. It went well. They seemed happy. We are getting good feedback. None of that converts, because none of it can be presented to whoever signs off on spending.
Four things need to be agreed and written down before the work starts.
Success criteria that both sides accept. Not your criteria, presented afterwards. Theirs too, agreed in advance, so nobody can move the target once results exist.
Quantitative, not qualitative. Not "we want faster response times." Instead: we want to cut response time by half. Not "we want better lead quality." Instead: we want a third of leads to reach a first meeting. A qualitative goal cannot be passed or failed, which means the pilot cannot end.
One metric that matters most. Keep the list short and name the single number they would most love to see move. A pilot with nine success criteria has none. This is also the number your champion repeats when you are not in the room, and you will usually not be in the room.
A time frame. One month, six weeks, a quarter. Say it out loud, because a pilot without an end date does not have an end.
Charge for It
The instinct is to run it free, because charging feels like it might spook a buyer who is already hesitant. In practice the opposite is true, and the fear of raising money is mostly the seller's, not the buyer's.
Charging does three things at once.
It qualifies. Somebody unwilling to spend a small amount to test a solution is usually not going to spend a large amount to adopt it, and it is far better to learn that in week one than in month four.
It changes how they behave. A customer who has paid for your time turns up to the check-ins, chases their own colleagues, and does not let the thing drift. A free pilot is nobody's priority.
It sets a floor. Once money has changed hands, the conversation about the full price is a continuation rather than a fresh negotiation.
Two refinements worth stealing. Offer to credit the pilot fee against the full purchase if they go ahead, which removes the sting without removing the qualification. And consider deliberately pricing the pilot beneath whatever threshold your buyer can approve personally, because a smaller number that one person can sign off often moves faster than a larger one that needs a committee. Trading some fee for that speed is usually a good trade.
While you are having the money conversation, have the whole of it. Ask what the full engagement would cost them if this works, and check that number is inside their budget. Discovering afterwards that they were never able to afford the real thing is the most expensive way to run a successful pilot.
Prove a Number They Already Care About
The strongest pilots do not prove that your thing works. They prove a number the buyer had already decided was worth money.
The move is to build the arithmetic with them before you start. What does this problem currently cost you, and what would a fix be worth? A support team of a hundred people costs what it costs; if the claim is that a fifth of the volume can be handled without them, that is a number with a salary line attached, and now the pilot has an obvious job: find out whether it is really a fifth, or a tenth, or a quarter.
Once that equation exists, your fee stops being a price and becomes a fraction. Charging a sum that recovers a multiple of itself is a much easier conversation than charging a sum on its own, and it is a conversation your champion can repeat accurately to a finance director who was never in the room.
It also gives the pilot an honest way to fail. If the number comes in at a tenth rather than a fifth, you have learned something real, and so have they. A pilot that can only succeed was never a test.
The Trap Both Sides Walk Into: Integration
Here is the one point where two very different sources say exactly the same thing, which is usually a sign it is true.
Do not integrate during a pilot if you can possibly avoid it.
The moment your pilot depends on connecting to their systems, it stops being your project. It now needs their technical people, who are busy. It may need a security review, which has its own queue. Each dependency adds weeks, and the weeks are what kill pilots, because momentum is the thing you are actually running on.
The workaround is to be deliberately unglamorous. Take an export instead of a connection. Let them email you the data and email the finished work back. Run alongside their existing process rather than inside it. Work from a sample rather than the whole system. None of that is how the finished engagement will look, and it does not need to be. The pilot exists to prove a number, not to demonstrate the architecture.
A related kindness worth offering: run it somewhere the stakes are survivable. One region, one team, one product line. Your champion is taking a personal risk by backing you, and a pilot confined to a corner of the business is one they can defend if it does not work. Making it safe for the person betting on you is not a soft skill, it is deal mechanics, and whoever absorbs the risk tends to win the customer.
The related version of this is back-testing. If you can prove your claim against data they already have, from a period they already understand, you can sometimes skip the live run altogether and still produce the evidence.
Book the Ending Before the Beginning
Two appointments should exist before day one.
The first is the start, and it is worth being picky about it. If your work needs a live project, a busy season or a real campaign to be judged fairly, wait for one. Starting a pilot into a quiet month, against nothing, produces a result nobody can read. Delaying two weeks for the right conditions is not lost time.
The second is the ending. Put a post-pilot meeting in the calendar before the pilot starts, with the decision-makers in it, and say plainly what that meeting is for: to look at the agreed numbers and decide.
That single booked meeting does more work than almost anything else on this list. It converts a pilot from an open-ended arrangement into something with a scheduled verdict, and it makes it socially awkward for the whole thing to trail off into silence, which is how most of them actually end. When one does die quietly, go and ask why rather than guessing.
Short Beats Thorough
Long pilots feel generous. They are mostly a way of dying slowly.
Two or three months is long enough for the sponsor to change roles, the priority to shift, the budget cycle to close, and the person who championed you to stop thinking about it. Keep it just long enough for them to genuinely experience the benefit, and no longer.
The number worth watching inside that window is how quickly they reach something valuable. Time to first value is the lever with the most leverage on whether a pilot converts, and it is why the unglamorous shortcuts matter so much. Every week you shave off the setup is a week of actual proof.
Even a Successful Pilot Is Not a Signed Contract
Worth knowing before you celebrate: the standard paid pilot still ends with a full negotiation. You proved the value, everyone is pleased, and now you begin a second round of procurement, pricing and legal, at the precise moment you believed the deal was done.
The structure that avoids this is to agree, up front, that the pilot rolls into an ongoing engagement unless they opt out. A monthly or annual arrangement with a genuine opt-out window at the start. If the work performs and they do nothing, it simply continues.
That is a bigger ask, and it will not be available to you on your first pilot with a nervous buyer. But it is worth knowing it exists, because it names what the ordinary version costs you: a whole second sale, run at your expense, after you already won.
None of this makes you difficult to work with. Every item here is something a serious buyer will happily agree to, and the ones who will not agree to any of it are telling you something useful about how the deal was always going to end.
We spend a good deal of our time helping small B2B firms turn interest into something with a date and a number attached, which is most of what B2B marketing actually is. If you would rather work it out yourself, start with the shortest possible version of this: what are we proving, by when, and what happens if it works.
