Paid Pilot vs PoC vs Design Partner: Which One Proves Demand?

A paid pilot provides the strongest evidence of commercial demand because a real buyer commits budget and tests the product in a real workflow. A proof of concept proves technical feasibility. A design partner proves that a representative customer has the problem and will help shape the solution. None of the three, on its own, proves repeatable market demand.

The label on the statement of work does not decide what you learned. A buyer can pay for a PoC because it wants custom research, not because it wants your future product. A design partner can sign a glowing agreement and never use the software. A pilot can look commercial while the founder performs so much manual work that no second customer could receive the same product.

This comparison focuses on choosing the right commercial test. If you only need the definition of a proof of concept, use the glossary. If you are deciding whether to build an MVP at all, start with the difference between an MVP and throwaway feasibility work.

Paid pilot vs proof of concept vs design partner at a glance

The distinction is consistent across public procurement and product practice. The European Innovation Council procurement toolkit defines a PoC around practical feasibility and a pilot as limited deployment in the environment where the solution will be used. The Australian Government's PoC-to-scale guidance makes the evidence boundary even clearer: PoC metrics concern technical feasibility, while pilot metrics concern users and business impact.

Which one proves demand?

Choose a paid pilot when the main uncertainty is commercial demand. It creates a stronger signal than an interview, letter of intent or free beta because the buyer has to pass a budget decision. It becomes credible only when the payment is for access to a repeatable product outcome, not for founder hours or a bespoke build.

Use this evidence ladder to state the result honestly:

1. Problem evidence: target users describe the same painful workflow and an active workaround.
2. Commitment evidence: a buyer provides data, staff time, security review or implementation access.
3. Payment evidence: a budget owner pays a meaningful, non-refundable fee.
4. Value evidence: real use moves a pre-agreed operational or financial baseline.
5. Conversion evidence: the buyer signs production terms or continues at the intended price.
6. Repeatability evidence: multiple independent customers in the same ideal customer profile buy for the same reason without a new product each time.

A paid pilot can reach steps three to five. One pilot cannot reach step six. That is why “we have a paid pilot” means one account has demonstrated demand under these conditions, not the market is validated. Product-market fit needs repeated pull, continued use and economics that survive delivery. Our minimum credible product guide explains how to define the smallest release that earns that stronger evidence.

What does a proof of concept actually prove?

A PoC should retire one technical uncertainty before you spend product money. Examples include whether a legacy interface can sustain the required throughput, whether a model reaches an agreed result on representative data, or whether a cryptographic construction fits the latency budget. The answer is yes, no or inconclusive against a written test.

A paid PoC still does not automatically prove product demand. Payment may prove that the customer values research, integration work or your team's expertise. It does not prove that the same buyer will purchase the resulting product at its production price. Keep two invoices conceptually separate:

• Feasibility fee: payment for answering a technical question.
• Product spend: payment for using an outcome repeatedly in the business.

If the only uncertainty is requirements, ownership or scope, a software discovery phase is usually the better instrument. Do not manufacture a technical experiment to make ordinary product decisions feel scientific.

What does a design partner prove?

A design partner gives an early product team recurring access to a real workflow, actual users and a buyer who can explain how a purchase happens. The best partner is representative of the intended segment, feels the problem urgently and has enough capacity to test. Those are also the three selection criteria in Andreessen Horowitz's design-partner framework.

A free design partner can produce excellent problem and usability evidence. It produces weak willingness-to-pay evidence. Praise, feature requests, executive enthusiasm and permission to use a logo are not substitutes for a purchase. The commercial question remains open until a budget owner accepts a price and a route to production.

A paid design partner can be the strongest early hybrid. The customer pays, supplies weekly feedback and helps shape a product that is not fully formed. Sierra used a high-commitment version of this model with payment, system access and recurring participation, as described in First Round's account of its design-partner strategy. The risk is overfitting. Every request must be tested against other target buyers before it becomes core roadmap.

What makes a paid pilot valid demand evidence?

Score one point for each condition below. A pilot scoring six or seven creates useful commercial evidence. Four or five is mixed. Three or fewer is closer to a demo, consultancy or subsidised experiment.

The European Space Agency describes pilots as customer testing in the primary target market and in a real operational setting, where the value proposition can be confirmed. That is the right standard. Payment raises the evidence quality, but operating truth is what distinguishes a paid pilot from paid theatre. See the ESA distinction between PoC studies and pilot projects.

How should you choose?

Do not run all three by habit. Start with the smallest instrument that can resolve the riskiest unknown. If proven technology can support the workflow, skip the PoC. If the product already serves the workflow, skip unpaid co-creation and ask for a paid pilot or production contract.

What belongs in the agreement?

The following commercial brief is useful before legal drafting. It is not legal advice, and counsel should adapt IP, data protection, liability and procurement terms to the parties and jurisdiction.

1. Decision question: one sentence naming what this engagement must prove.
2. Scope and exclusions: workflow, users, data, integrations, volume and what is deliberately out.
3. Customer commitments: executive sponsor, operational owner, user time, data access, security review and feedback cadence.
4. Evidence plan: baseline, success metrics, event instrumentation, sample size and owner of the final readout.
5. Commercial terms: fee, payment timing, expenses, production price or pricing mechanism and any credit toward rollout.
6. Product boundary: standard product, configurable work and bespoke deliverables listed separately.
7. Rights: pre-existing IP, new IP, customer data, aggregated learning, confidentiality and reference rights.
8. Exit gate: fixed end date, decision meeting and the exact options: convert, one bounded iteration or stop.

For a software build, pair the brief with a clear statement of work. If the pilot will use personal or regulated data, do not postpone security and data responsibilities until rollout. Real operating conditions include real governance constraints.

Three examples, three different verdicts

1. The bank pays €20,000 for an integration PoC

The vendor proves that its engine can read one legacy message format at the required speed. The bank pays for specialist engineering. Verdict: technical feasibility and services demand are proven; product demand is not. The next gate should expose real users to the standard product and state production pricing.

2. Five design partners attend weekly sessions for free

Four use the prototype repeatedly and describe the same painful workaround. Verdict: problem urgency and product direction are supported. Willingness to pay is still unknown. Offer the same bounded paid pilot to the representative partners, not five different custom roadmaps.

3. Three operations teams buy the same six-week pilot

Each pays from the intended budget, uses the same core product and evaluates the same business outcome. Two convert at the pre-agreed production range, one stops because volume is too low. Verdict: there is early repeatable demand in the high-volume segment. The loss improves the ICP instead of invalidating the whole test.

"A PoC proves that the machine can work. A design partner proves that the problem deserves attention. A paid pilot proves that one buyer will spend money to change a real workflow. Repeatable demand begins when another buyer does the same without asking you to become a different company."

How much demand is enough?

There is no universal customer count. Enterprise products with high contract values and narrow markets need fewer observations than low-price horizontal tools. Use a stronger rule: seek independent replication across the segment you plan to sell.

• One paid pilot: account-level demand evidence.
• Several similar paid pilots: segment-level demand hypothesis.
• Conversions at the intended price: early commercial validation.
• Continued use, renewal or expansion: durable value evidence.
• A repeatable sales and delivery motion: a credible path toward product-market fit.

Willingness to pay matters because demand without viable economics is not a business. Andreessen Horowitz's pricing and packaging guidance connects willingness to pay with both demand and expected return. Test the intended commercial model early enough that a discounted pilot does not conceal a production price nobody will approve.

After a live pilot has accumulated representative operating data, use the AI pilot kill-or-scale scorecard for AI workflows, or adapt its baseline, cost, adoption, reliability and payback logic to conventional software.

Sources and methodology

This comparison is an original Wavect decision model based on the type of uncertainty each engagement reduces. Definitions and procurement boundaries were checked on 22 July 2026 against the European Innovation Council, Australian Government and European Space Agency. Design-partner and willingness-to-pay claims were checked against Andreessen Horowitz and First Round. The evidence ladder and seven-condition pilot test are Wavect frameworks, not universal standards.