A first paid agency pilot should prove that a client will pay for a bounded deliverable you can produce at a sustainable cost. Choose one recurring problem, agree what the client will receive and track all the work needed to deliver it. The first useful result is an accepted piece of work with an honest cost record.
This guide produces a pilot proposal, delivery checklist and evidence log. It suits a solo operator or small team with a real service skill and limited proof. Budget a few focused sessions for qualification and scoping, then a delivery window that fits the work and client dependencies. A pilot is not permission to promise revenue, rankings or results you do not control.
1. Select a problem you can solve with a concrete output
Start with recurring requests you have actually heard: a small software company needs its onboarding emails checked, or an established publisher needs a content inventory reconciled. Choose a job where the output and its usefulness can be examined without waiting for months of downstream performance.
For a hypothetical pilot, offer an onboarding-email audit covering one signup journey, up to five messages and a prioritised repair list. Exclude implementation, legal review, new design assets and unlimited stakeholder interviews. A first pilot that includes everything makes it difficult to learn what the client valued or what the work cost.
Use a problem statement: “New customers receive inconsistent next-step instructions across the welcome sequence.” Ask the prospect to show examples and explain the operational consequence. If the prospect wants a guaranteed increase in paid conversion, clarify that the pilot can identify and document defects; it cannot guarantee how customers will respond.
Early user contact and iteration are useful general principles, but they do not establish your service's demand or economics. Treat them as a method for learning, then collect your own paid-delivery evidence. Y Combinator's startup advice.
2. Qualify prospects before writing proposals
Create a short list of organisations with the problem, the required inputs and a person who can accept the work. Prefer relevant, permission-respecting introductions and channels where your message is welcome. Personalise the observation; do not manufacture a case study or imply a prior relationship.
A truthful opening might say: “I reviewed the public signup path and noticed that the welcome message and help page describe different first steps. I can audit one journey and deliver a repair list. I have a demonstration audit available.” Label the demonstration as a sample using public or fictional material, not paid client success.
On the qualification call, ask who owns the journey, what has already been tried, what access is available and what decision the report should support. Confirm budget authority and a named reviewer. A friendly contact without access or authority can create substantial unpaid coordination.
Record disqualifiers: the client cannot provide the necessary materials, demands an outcome guarantee, wants work outside your competence or expects uncontrolled access to sensitive data. A polite no-fit decision is part of a functioning pipeline. Do not turn every objection into a promise to absorb more scope.
3. Write a proposal that can be accepted objectively
Use plain operational terms: problem, deliverable, inputs, exclusions, timetable, price, acceptance method and change process. For material agreements, have qualified advisers review the actual commercial terms. The outline here is a scoping exercise, not a contract template or legal opinion.
For the onboarding example, the deliverable is a spreadsheet mapping five messages to the signup journey, an annotated report of defects and a handover call. Required inputs are current message copies, the intended user journey and one stakeholder interview. The client appoints one reviewer to consolidate comments.
Acceptance could mean every in-scope message is inventoried, each recommendation points to an observed issue and the repair list separates factual defects from hypotheses. It should not mean “the client feels happier” or “conversion rises by a specified percentage.” Those conditions either lack a clear test or depend on later work outside your control.
Name the dependency clock. For example, the planned delivery window starts when the agreed materials are complete; missing access pauses the schedule and triggers a revised date. Decide what happens if the supplied materials differ substantially from the scoped journey before starting the analysis.
4. Price the pilot with visible labor
Estimate research, execution, checking, meetings, revisions and administration. Assign a cost to owner time even if you do not immediately pay yourself that amount. Otherwise a supposedly profitable service can depend on unpaid evenings.
Here is a hypothetical calculation in one currency:
| Cost item | Assumption | Cost |
|---|---|---|
| Owner delivery and analysis | 10 hours at $40 | $400 |
| Independent checking | 2 hours at $35 | $70 |
| Meetings and administration | 3 hours at $40 | $120 |
| Project-specific tools | Actual planned expense | $30 |
| Total planned delivery resources | Stated items only | $620 |
At a pilot price of $900, the planned contribution after those delivery resources is $280, or about 31% of revenue. That is not total business profit: acquisition effort, general overhead, tax and other costs remain. If delivery overruns by five owner hours, contribution falls by $200 to $80.
The arithmetic gives you a decision. Reduce the work, improve the process or charge enough to cover the real resource envelope. A discount may be a deliberate learning expense, but record it as such. Do not call a subsidised experiment a proven profitable service.
5. Deliver against an observable checklist
At kickoff, confirm the input bundle, scope and reviewer. Save a versioned copy of the materials you received so late client changes do not silently alter the baseline. Use the least access needed, with client-approved handling and deletion arrangements for sensitive material.
Follow a repeatable sequence: inventory the messages, trace the user journey, identify inconsistencies, prioritise repairs and check the report. Distinguish an observed defect from an interpretation. “The button leads to an unavailable page” is different from “a shorter subject line may improve engagement.”
Track time by activity while the work happens. Capture interruptions and rework as well as focused delivery. If the client supplies a new sequence halfway through, log it as a scope event and agree the next action instead of quietly doubling the project.
Before delivery, have someone follow the recommendations using the source materials. They should be able to reproduce the factual issues and understand why each repair matters. Remove unsupported estimates and claims you cannot substantiate.
6. Close the pilot with acceptance and evidence
Deliver the report with a short explanation of what was checked, what remains uncertain and what the client can do next. Ask the named reviewer to confirm acceptance against the agreed criteria and to consolidate any corrections. Do not use silence as evidence of satisfaction unless the actual agreement clearly addresses acceptance.
Complete an evidence log containing scope version, received inputs, delivery date, actual time, expenses, invoice and payment status, accepted output and exception notes. Keep payment timing distinct from revenue recognition and obtain accounting advice where needed. An invoice sent is not cash collected.
In the hypothetical pilot, suppose actual owner delivery takes 12 hours instead of 10. The extra $80 lowers the stated contribution from $280 to $200. If the client also requests implementation, quote or decline that work separately. Absorbing it makes the audit's economics impossible to interpret.
Use a completed record rather than a folder of unconnected emails. For the same hypothetical pilot:
| Evidence field | Illustrative closing entry |
|---|---|
| Scope | One signup journey, five emails, report and handover; implementation excluded |
| Acceptance | Named client reviewer confirms the inventory and evidence-linked repair list; correction log retained |
| Delivery cost | 12 owner delivery hours plus three meeting/admin hours at $40, two checking hours at $35 and $30 tools: $700 |
| Revenue and collection | $900 fee; keep invoice date, due date, amount collected and collection date as separate fields |
| Acquisition effort | If qualification and proposals took another four owner hours at $40, include $160 separately |
| Decision | $200 after stated delivery resources becomes $40 after that acquisition effort, before shared overhead and tax; reprice or shorten the next pilot if that buffer is insufficient |
Acceptance and payment are distinct checks: an accepted report with an unpaid invoice has not demonstrated cash collection. Agree the payment schedule and what happens when inputs, approval or payment are delayed before beginning; obtain advice for the actual terms rather than treating this worksheet as a contract.
Ask for specific feedback: which recommendation was useful, what was missing and whether the client could act without another explanation. Obtain explicit permission before using their name, materials or words publicly. A useful private comment is not automatic permission to publish a testimonial.
7. Decide whether the service deserves repetition
Repeat when the problem recurs, the deliverable is accepted and actual cost leaves a credible contribution after the work required to win clients. Revise when the value is clear but inputs, meetings or checking consume more time than expected. Stop when clients will not pay enough or the result depends on skills you cannot responsibly supply.
Use a small pilot review:
| Question | Evidence | Next action |
|---|---|---|
| Did the client receive the promised output? | Acceptance record and correction log | Repair incomplete scope before upselling |
| Did the work fit the resource envelope? | Actual activity-level hours and expenses | Adjust scope, price or process |
| Is demand repeatable? | Similar qualified requests, not compliments alone | Test another bounded engagement |
| Can another person deliver it? | Checklist and independent checking result | Improve instructions before delegating |
Do not turn one accepted pilot into a claim that you have a scalable agency. It is one operating observation. A second project with different client inputs can expose exceptions the first did not reveal.
The productized-agency guide shows how to turn those exceptions into a repeatable scope and evidence packet. A marketplace decision comes later, after you can explain the client relationships, obligations and delivery economics.
Should the first pilot be free?
A free demonstration can show your method, but it cannot prove willingness to pay. If you choose one, cap the effort and label the result accurately. A paid pilot provides different evidence, especially when payment and acceptance are documented.
What if the client asks for guaranteed returns?
Explain the output you can control and the downstream variables you cannot. If the purchase depends on a guarantee you cannot support, decline or redesign the engagement. A clear refusal is less costly than an unfulfillable promise.

