Hardware founders have a name for it: pilot purgatory. The pilot ran, the data came back clean, the champion is enthusiastic, and the production order never arrives. The scale of the problem is documented. Median pilot-to-production conversion is 12 percent (IDC 2025). Roughly seven pilots in eight produce no production revenue, and the failure is common enough that it has stopped embarrassing anyone. It should. This page is about the seam where pilots actually die, the pricing that pre-commits a production path, the ninety-day review almost nobody runs, and the buying committee that approves production while the vendor is still selling to one person.
SignalForge works with technical founders at $1M to $20M annual revenue in hardware, deep tech, robotics, energy infrastructure, agtech, and advanced manufacturing. In that segment the pilot is the modal sales motion: almost every significant deal passes through one. Which makes the 12 percent a pipeline-conversion problem, not a product problem. The product usually performs. The conversion fails.
The core claim of this page: pilots convert because the conversion was designed at signing, not chased after the data lands. A pilot is not a free trial, the divergence Pillar 01 lays out in detail. It is a rehearsal for a capital decision, and a rehearsal only matters if the performance is on the calendar.
Why do pilots die at the innovation-to-business-unit budget handoff?
Follow the money inside the buyer and the purgatory explains itself. The budget that funds the pilot and the budget that funds production are different budgets, owned by different people, scored on different outcomes.
The pilot is usually paid for by an innovation department, a digital transformation group, or an engineering evaluation budget. That budget exists to run evaluations. Its owner is scored on pilots launched, technologies assessed, learnings reported. A completed pilot is, for that owner, already a success. Nothing in their incentive structure requires anything to be purchased afterward.
Production is paid from a business unit's profit-and-loss statement (P&L), owned by an operator scored on output, uptime, and cost. For that owner, adopting your hardware is not an innovation win. It is personal risk: integration downtime, retraining, a vendor bet they will carry for years. They were often not in the room when the pilot was scoped, and the pilot's success criteria were not written in their language.
The pilot did not fail. It succeeded against a budget that was never going to buy it.
That is the seam, and deals die there in volume: a pilot that succeeds against the innovation budget's criteria still has to cross to a P&L owner who never agreed to anything. The fix is deal design, the discipline Stage 03 of Proof to Pipeline builds as a layer of the engine: the production path gets written into the pilot agreement itself. Three clauses do most of the work. The business-unit budget owner is named in the pilot agreement, as a signatory or a named reviewer. The success criteria are co-signed by that owner, in operating language, cost, uptime, output, not in evaluation language. And production pricing and timeline ride as an appendix, so the production decision is a pre-negotiated next step instead of a brand-new deal.
If the business-unit owner will not put their name on the pilot document, that is not a lost deal. That is the deal's truth, surfaced for the price of a pilot instead of the price of a year. Walk, or fix the sponsorship before the hardware ships.
How should you price a hardware pilot?
Never free, and never as cost recovery. Both answer the wrong question. The right question is what the pilot has to accomplish commercially, and the answer is: pre-commit the production budget path.
The free pilot fails twice. It signals that the pilot is not serious, so the buyer never resources it to convert, the vendor-side pattern Pillar 01 names. And it is a fiction anyway: the buyer pays for every pilot in integration hours, floor space, staff time, and risk, whatever the invoice says. A free pilot does not lower the buyer's cost. It removes the one signal that forces the buyer's organization to process the evaluation as a purchase.
Priced as a purchase, the pilot starts selling for you. A pilot fee sized as a meaningful fraction of a production unit cannot be approved on a champion's discretionary budget. It pulls procurement and finance into the deal months early, which feels like friction and is actually the conversion mechanism: the people who must approve production have already approved you once, at a smaller number.
Three structures pre-commit the path. Credit: the pilot fee credits against the production contract, so walking away from a successful pilot has a named cost. Locked production pricing: unit pricing and volume tiers ride in the pilot agreement, so the production negotiation happens while goodwill is at its peak instead of after the leverage is gone. A dated decision: the agreement names the review meeting where the production decision gets made, which is the next section.
The pilot is free or token-priced. Nobody on the buyer side had to defend the spend, so nobody on the buyer side owns the outcome.
No business-unit budget owner is named in the agreement. The deal will have to find a new sponsor after the data lands, which is the seam where pilots die.
No production pricing exists in writing. A successful pilot leads to a negotiation, not an order.
No decision date is on a calendar. The pilot ends with a report and a silence.
None of this requires aggressive pricing. It requires honest pricing: the pilot is the first stage of a capital purchase, and every line of the agreement should behave like it.
What is the 90-day post-pilot review, and what gets measured?
The highest-conversion meeting in hardware sales is one most vendors never schedule: a formal review, roughly ninety days after pilot deployment, where the buying committee decides production in a room with a document in front of it. It gets booked at pilot signing, not requested after the pilot ends, because by then the buyer's attention has moved on and the meeting you ask for is a favor instead of a clause.
Who attends matters as much as the date: the champion, the business-unit budget owner, procurement, a finance delegate, and the vendor. The same seats that approve production, which is the point. The review is the production decision, staged as a meeting.
What gets measured is deliberately narrow. Performance against the co-signed success criteria, those and only those, because criteria introduced after the fact are how committees defer. The operating cost delta the business-unit owner will defend upward in their own budget cycle. Integration cost actually incurred against plan, because the committee will price the production rollout from it. And the incident record: what broke, what the response time was, and what that proves about the support story, the front-loaded evidence a physical, high-switching-cost purchase demands.
The meeting ends in one of three outcomes, written down. A production order with a date. A scoped second deployment with a date. Or a documented no with reasons. All three beat the default, which is an enthusiastic champion, a circulated PDF, and a deal that ages in the CRM until everyone stops mentioning it.
A pilot that ends without a decision meeting was a demonstration, not a deal stage.
Most reviews land in the middle: the data is good but not unambiguous, one criterion missed, one stakeholder unconvinced. Run the meeting anyway. The mixed case is exactly what the co-signed criteria exist for, because they convert a feeling of almost into a short, named list of what stands between the pilot and the order. A scoped second deployment with a date is not a stall. It is conversion in two steps instead of one, and it keeps the deal on a calendar instead of in a mood.
The no is the underrated outcome. Forced early, it releases engineering and calendar capacity for deals that can close, and it usually names a fixable reason. The yes converts pilot evidence into the internal memo the committee actually reads, which almost never happens on its own. Proof does not circulate itself, the failure pattern the Signal Notes piece The milestone is not the market dissects in detail.
Who is in the room when production gets approved?
More people than the vendor has met. The modal B2B buying committee runs 6.3 to 6.8 people (Gartner and 6sense composite), and a hardware capital purchase skews above that, pulling in procurement, operations, plant or facilities, and on regulated products, compliance and safety.
Multi-stakeholder selling is the discipline of treating every one of those seats as a buyer with its own question. Finance asks payback and capital-expenditure treatment. Procurement asks vendor risk and contract structure. Operations asks integration, downtime, and training load. Compliance asks the certification path. The champion's enthusiasm answers none of those questions, and selling to the champion alone, the strongest instinct software training installs, is how hardware deals die politely, in rooms the vendor was never in, the pattern Pillar 01 documents.
The pilot phase is when the map gets built, because the pilot is the vendor's one legitimate season of access to the account. Every committee seat gets its artifact: the finance memo, the procurement risk one-pager, the operations integration plan, the compliance path. Each artifact carries the sentence that seat can repeat upward, the Stage 02 discipline of Proof to Pipeline: hand the buyer the exact narrative they will repeat in the meetings you will never attend.
Because that is the structural fact of hardware selling: the production decision is made in a meeting the vendor does not attend. The committee that walks into that room either carries your artifacts and your sentences, or improvises. The 12 percent are the deals where nobody had to improvise.
The conversion, in one place.
Pilot purgatory is not a market condition. It is a deal-design outcome, and it reverses deal by deal: name the business-unit budget owner at signing, price the pilot as the first stage of a capital purchase, book the ninety-day review before the hardware ships, and arm every committee seat for the meeting you will not be in. Each move is unglamorous. Together they are the difference between the 12 percent and the rest, and they are why pipeline conversion in hardware is an architecture question before it is an effort question.
The data anchors cited on this page are sourced, with their primary references, on the Hardware GTM Benchmarks 2026 page.
Who runs this motion matters, which is why the conversion problem and the hiring problem are the same problem two pages apart. Pillar 03 covers who should be running it, and when the founder hands it off. And if you want to know where your own engine leaks before the next pilot signs, the Diagnostic scores it across twelve constructs, deal design and pipeline math among them.