How to Test a Service Business With a Paid Pilot
Short answer: Sell one fixed pilot lasting two to four weeks to one defined buyer, using a baseline the customer already accepts and one result you can measure by the end.
Charge at least all direct costs plus half your target pay for delivery time as a working floor, and state the normal future price before the pilot starts. Define customer value, your contribution and the decision at completion. If those cannot be measured, the pilot is discounted work, not a test.
A pilot should reduce uncertainty for both parties. The customer learns whether the result works in their operation. You learn whether someone will pay, whether delivery is repeatable and whether the economics survive real exceptions.
Many founders make pilots too generous and too vague. They promise broad help, absorb every change and call customer happiness validation. That proves only that flexible labour is welcome when underpriced.
Use the Six-Boundary Pilot Design
The Six-Boundary Pilot Design prevents a small test from becoming an open-ended engagement.
| Boundary | Define before payment | Reason |
| Buyer | One organisation, decision owner and participating users | Prevents feedback from people who cannot continue |
| Baseline | Gives improvement a credible comparison | |
| Intervention | Exact work, inputs and exclusions | Protects delivery time and interpretation |
| Measure | Stops success being redefined later | |
| Money | Fee, payment date and customer-side costs | Tests commercial commitment and contribution |
| Exit | End date, handover and next decision | Prevents implied renewal or unpaid continuation |
Write each boundary in plain language. If the customer cannot provide the baseline or inputs, that is part of the test result. Do not quietly do extra work to preserve an attractive case study.
Choose one uncertainty worth testing
A pilot cannot test the whole business. Decide which uncertainty could kill the offer:
- The customer may not value the result enough to pay.
- The service may not change the chosen measure.
- Delivery may take too long at a viable price.
- Customer participation may be too demanding.
- Important exceptions may prevent standardisation.
Choose one primary question and record secondary observations. If you change buyer, scope and method halfway through, the final result cannot tell you which assumption worked.
My view is that willingness to pay should be tested in the first pilot, not postponed until you have “proved value”. Free delivery changes customer attention, approval and expectations. It can provide operational learning, but it does not test a business.
Set the baseline with the customer
Use a recent period and data the buyer considers credible. A four-week pilot compared with one unusually bad week will exaggerate improvement. Where demand is seasonal, use comparable days or previous periods and state limitations.
Define the measure precisely. “Fewer delivery problems” is vague. “Number of paid driver overtime hours and repeat journeys per completed order” can be counted. Agree who records it and how disputes are handled.
Do not choose a measure simply because it is easy for you to improve. It must connect to the customer's buying reason. Activity such as reports produced or calls made is evidence of delivery, not necessarily value.
Price the pilot to expose real economics
Calculate direct cash, your delivery hours, selling time and customer support. The working floor of direct costs plus half your target pay recognises that an early pilot may include learning while refusing to treat your labour as free.
For example, if direct costs are £80, expected delivery is 20 hours and your target is £30 an hour, the floor is: £80 + (20 × £30 × 50%) = £380.
This is not a universal pricing rule. High-risk, regulated or specialist work may need full economic pricing from the start. A buyer may also refuse a pilot that is too small to matter. Check your market and responsibility.
State the likely ongoing price and scope. A £300 pilot that becomes £2,000 a month can produce false demand if the buyer was never prepared for that step.
Give both parties pass and fail rules
The customer needs a value threshold. You need an economic and delivery threshold. Write all three before work starts:
| Decision | Example rule |
| Customer pass | Verified monthly benefit exceeds full ongoing cost by an agreed margin |
| Founder pass | |
| Operational pass | Customer inputs arrive on time and exceptions remain within scope |
A customer win and founder loss is not success. Nor is your efficient delivery valuable if the customer's measure does not improve.
Practitioners disagree about guaranteeing pilot outcomes. A guarantee can reduce customer risk, but it can also make you responsible for factors the customer controls. My view is to guarantee your defined work and remedy delivery failures, not an operational result you cannot control.
Contract and consumer requirements vary, so use qualified local advice for material commitments.
Worked example: Nora's delivery-slot planning pilot
Nora offers delivery-slot planning to an independent florist. They agree a three-week paid pilot for £450. The baseline is 12 hours of driver overtime a week and nine repeat delivery journeys over three weeks.
During the pilot, overtime falls to seven hours a week. The florist values driver time at £18 an hour. Overtime value saved is:
- hours a week × 3 weeks × £18 = £270.
Repeat journeys fall from nine to three, so six are avoided. The florist calculates vehicle and paid-time cost at £24 per journey:
- × £24 = £144 saved.
Total measured benefit is £270 + £144 = £414. After Nora's £450 fee, the customer result is negative £36. The pilot misses the customer's financial pass rule.
Nora spends 16 hours delivering and values her time at £24 an hour. Travel costs £36.
Her economic cost is (16 × £24) + £36 = £420. Her surplus is £450 minus £420 = £30. It is positive but too thin to cover selling time or a correction.
The pilot is useful because it fails cleanly on both sides. Nora should not turn £414 of benefit into a success story. She could reduce the service to the highest-impact planning step, target florists with more deliveries or test a higher baseline-cost segment. Cutting the fee would improve customer value while making her economics worse.
Keep the pilot representative
Do not select only ideal cases, perform hidden overtime or rely on founder judgement that cannot be repeated. Record every exception and hour. If you create a manual workaround for the pilot, state what would happen at normal volume.
The customer should contribute the same inputs an ongoing service would require. Chasing data personally for three weeks may conceal a fatal dependency. Late inputs are evidence about delivery design.
Where you access personal, financial, health or commercially sensitive information, define permissions, security, retention and responsibility. Requirements differ by country and sector. Obtain qualified professional advice where the pilot creates legal, safety, data or employment obligations.
Close the pilot with a decision
On the final date, compare the agreed baseline and measure, then show customer value, your delivery economics and any limitations. Do not use a presentation to obscure a failed threshold.
Choose one of four outcomes:
- Continue at stated commercial terms because all pass rules were met.
- Run one materially different test because a specific assumption changed.
- Offer a narrower service supported by the evidence.
- Stop because value, delivery or participation cannot work.
Within the next seven days, define the Six-Boundary Pilot Design and make it to five qualified buyers. Sell one pilot, deliver it within four weeks and make the decision on the agreed end date. The purpose is an answer, not a permanent trial rate.
Frequently asked questions
How long should a paid service pilot last?
Use the shortest period that includes a representative operating cycle, usually two to four weeks for a recurring small-business service as an indicative range. A weekly workflow may need several repetitions, while a monthly close or seasonal event needs longer. Do not extend simply because the result is disappointing.
Set the end date and minimum data before payment. The exception is a high-value or regulated process where safe implementation and approval require more time. In that case, divide the pilot into clear stages and obtain qualified advice on responsibilities rather than compressing work to fit an arbitrary deadline.
Can I run paid pilots with two customers at once?
Yes, if you can deliver both without changing the defined process or weakening service. Use the same core scope, evidence fields and review date, while keeping customer information separate. Two pilots can reveal whether the result repeats, but they also double exceptions and make it easier to confuse a capacity failure with a weak offer.
Cap your weekly delivery hours and leave correction time before accepting the second buyer. The exception is a safety-critical, regulated or operationally complex service, where one controlled pilot may be the responsible limit until qualified review confirms the method.
Should I refund the fee if the pilot misses its target?
Only if your agreed terms make the result your responsibility and you can control the inputs. A fair structure may remedy work that failed the specification while leaving the fee payable when customer data, access or external events caused the outcome. Never invent the refund rule after seeing results. State deliverables, dependencies, measurement and remedy before payment.
Consumer, contract and professional rules vary by jurisdiction and may override your preferred terms. For material commitments or high-consequence work, obtain qualified local legal advice. A refund can reduce buying risk, but it should not turn an uncontrolled outcome into your guarantee.
Does one paid pilot prove the service business works?
No. It proves that one buyer paid and reveals one delivery case. Use it to correct scope, measure time and identify exceptions, then repeat with similar buyers. Three comparable paid engagements can show patterns, but no fixed count proves a market. Watch whether selling effort, customer participation and contribution remain acceptable without founder heroics.
The first customer may be unusually supportive or have an unusual operation. The exception is a deliberately bespoke high-value service, where repeatability matters less, but pipeline, concentration and pricing must still support the business.
What if the customer keeps adding requests during the pilot?
Refer to the intervention boundary and price changes separately. New requests are useful evidence of adjacent needs, but completing them for free destroys the pilot's time and contribution test. Ask whether the request is required for the agreed measure.
If yes, decide whether the original scope was defective and document the change. If no, record it for the closing decision. The exception is a safety, legal or data issue that must be corrected immediately. Stop affected work and obtain appropriate advice rather than using scope as a reason to ignore material risk.
Should the pilot automatically become a full contract?
No. End with a deliberate decision using the agreed pass rules and updated commercial terms. Automatic continuation can surprise the buyer and conceal whether they actively value the service. State what happens to data, access, unfinished work and customer materials at the end.
If both parties want to continue, use a new agreement or clearly activated next stage. The exception is a pre-agreed conversion mechanism requested by an experienced buyer, provided cancellation, price and notice are explicit and lawful. Requirements differ by jurisdiction, so obtain suitable advice for ongoing contracts and renewals.
Comments (0)