The Two-Week Paid Pilot: How We Replaced the Free Proposal

We killed the free proposal in 2024. Not because we got too busy — because we noticed every deal we lost was preceded by a forty-page document nobody read. The replacement is a paid two-week pilot, and it has changed the shape of our pipeline more than any sales hire.

This is the playbook: what's in the pilot, what we charge, how we contract it, and the honest numbers on what it does to your funnel.

Why free proposals quietly bleed agencies

The free proposal is a holdover from an era where agencies competed on production quality of the PDF. You'd get an RFP, spend 30–60 hours of senior time on architecture diagrams, estimates, team bios, and a SoW, then email it into a void. In our experience, win rates on cold proposals sit somewhere between 10% and 20%. So you're paying for four losing proposals for every one that closes.

Worse, the proposal is the wrong artefact. It's a sales document pretending to be an engineering document. The estimates are guesses dressed up in Gantt charts. The risks section is boilerplate. And once the client signs, you discover the actual scope in week three of the build — which is exactly when fixed-bid projects start losing money.

The diagnosis we kept arriving at: we were doing the discovery work for free, badly, and giving it away. A paid pilot fixes all three problems.

What a two-week pilot actually contains

The pilot is not a discount on the full project. It is a standalone engagement with its own deliverables, its own acceptance criteria, and its own invoice. If the client never hires us for the build, they still walk away with something they can hand to another vendor.

Here's the deliverable bundle we settled on after about a dozen iterations:

• A technical architecture doc — chosen stack, key trade-offs, integration map, data model sketch
• A prioritised backlog broken into milestones, with rough estimates as ranges (not point values)
• A risk register with named risks, likelihoods, and mitigation options
• A working spike — usually the riskiest integration or the gnarliest UX flow, built in code
• A fixed-bid proposal for phase one, only valid because the pilot derisked it

The spike is the thing that separates this from a glorified discovery phase. By the end of week two, the client has something running on a staging URL. It might be ugly. It might be one screen. But it proves we understood the problem.

How we scope the spike

We pick the spike before the contract is signed. In the qualification call, we ask: "If this whole project failed, what's the single thing that would have caused it?" The answer is the spike. Sometimes it's a Stripe Connect flow with split payouts. Sometimes it's an LLM pipeline with a retrieval step that has to return in under two seconds. Sometimes it's whether a legacy ERP will let us write to it at all.

The spike is intentionally narrow. One screen, one endpoint, one integration. If you let it sprawl, you've just sold a small project at pilot pricing, which defeats the point.

Pricing the pilot

We charge between $12k and $25k for a two-week pilot, depending on team composition. The lower end is one senior engineer plus a part-time PM. The upper end adds a designer and a second engineer for parallel tracks.

A few pricing rules we hold to:

1. The pilot is profitable on its own. It's not a loss leader. If we never close the follow-on, we still made margin.
2. The pilot fee credits against the build, partially. We credit 50% of the pilot fee against phase one if the client signs within 30 days. This rewards momentum without giving the work away.
3. No free revisions on pilot deliverables. If the client wants the architecture redone three ways, that's a new engagement.

The credit mechanic is the one most founders push back on. The instinct is either to credit 100% ("it should feel free if they hire us") or 0% ("we did the work, we keep the money"). 50% is the number that signals goodwill without training clients to treat the pilot as refundable.

The contract structure

The pilot needs its own short-form agreement. Don't bolt it onto an MSA you haven't negotiated yet — that's how pilots get stuck in legal for three weeks and lose all their urgency.

We use a two-page letter agreement covering:

1. Scope: bullet list of the five deliverables
2. Timeline: two weeks, with a defined start date
3. Fee: fixed, 50% on signature, 50% on delivery
4. IP: client owns deliverables on full payment
5. Exclusivity: none — we can both walk away
6. Build option: 50% credit if phase-one SoW signed within 30 days

The exclusivity clause matters. Some clients want a no-shop during the pilot. We refuse. The whole point is that both sides are evaluating fit, and locking either party in changes the dynamic.

The kickoff ritual

Day one is not a kickoff deck. Day one is a working session where we sketch the architecture on a shared canvas with the client's most technical person in the room. If they don't have a technical person, we ask them to bring whoever will be the day-to-day product owner.

By end of day one, we have a one-page architecture sketch the client has signed off on verbally. Everything we build in the next nine days flows from that page. This is the part that scope-creeps without discipline, so we treat the day-one artefact as the source of truth and reference it in every standup.

What it did to our funnel

The headline change: we talk to fewer prospects, and a much larger share of them become clients.

In our experience, roughly one in three serious inbound conversations now converts to a paid pilot. Of those pilots, somewhere between 70% and 85% convert to a phase-one build. So the end-to-end conversion from qualified lead to paying build client is meaningfully higher than the old free-proposal funnel, even though the top-of-funnel number is smaller.

The other changes are harder to measure but more important:

• The clients who say yes to a pilot are better clients. They have budget authority, they make decisions, and they don't disappear for three weeks mid-conversation.
• Our phase-one estimates are more accurate, because they're based on two weeks of real work instead of two hours of guessing.
• Senior engineers stopped resenting sales. Writing a spike is engineering. Writing a proposal is not.

There is a cost. You will lose deals you would have won under the old model — usually to agencies still doing free proposals. We've made peace with that. The deals we lose to free-proposal shops are the deals we didn't want.

When the pilot model doesn't work

A few situations where we still write a traditional proposal:

• Government and large enterprise procurement. The process requires a fixed-format response. The pilot can sometimes happen after award, framed as a discovery sprint.
• Existing clients on a new project. We already know each other; the pilot would be theatre.
• Projects under about $40k total. The pilot fee becomes a disproportionate fraction of the engagement.
• Pure staff augmentation. There's no scope to derisk; the client just wants engineers on a project they're running.

For everything else — and especially for first-time clients on fixed-bid greenfield builds — the paid pilot has become the default.

Where we'd start

If you're an agency founder reading this and your last five proposals took 40 hours each and closed two of them, run the experiment. Pick the next qualified inbound that isn't a procurement-driven RFP. Send a one-page pilot offer instead of a proposal. Price it at whatever feels uncomfortable but defensible — probably $15k for two weeks. See what happens.

The first one will feel awkward. You'll want to discount. Don't. The second one will feel normal. By the fifth, you'll wonder why you ever wrote a free proposal.

If you want to talk through how to structure a pilot for a specific kind of engagement, that's the sort of conversation we have with founders before they hire us — see our services page for what we typically run pilots on.