Skip the RFP: a better way to buy creative work

Every few months a 40-page RFP lands in our inbox with a five-day turnaround, an NDA that forbids us from mentioning rivals exist, and a request for "three creative concepts with projected ROI." We decline all of them — politely, and with this article attached.

This isn't diva behaviour. It's pattern recognition: in more than a decade, we have never seen a great creative engagement begin with a great RFP. Not once. The correlation isn't coincidence; it's selection. The RFP process reliably identifies the studios that are best at responding to RFPs, and that is a different skill — almost a different industry — from building brilliant software. If you want the best builder, you need a process that selects for building.

Why RFPs select the wrong winner

Four structural flaws, none fixable with a better template:

The concept tax. Asking for speculative creative work selects for agencies with a bench of people idle enough to produce free concepts — which usually means the work is done by whoever's lightest on bookings, presented by whoever's best in a boardroom, then handed to an entirely different team if you win. You're buying the pitch, then receiving the bench. Every agency knows this. None of us will tell you during the RFP, because we're all standing in the same line hoping you can't tell.

Requirement theatre. RFPs demand that agencies answer 80 detailed questions about technical approach — for a project whose actual shape hasn't been discovered yet. So we write confident answers to unanswerable questions, and procurement scores the confidence. A truthful proposal would say "it depends, here are the three shapes it might take and how we'd find out" — and that proposal loses to the fiction that says "Phase 1: Discovery, $12,000." The process punishes honesty at the exact stage where honesty is most valuable.

Price-anchoring too early. Asking for fixed costs against invented scope means the price you compare isn't a price — it's a guess about a guess. We wrote about what honest estimation actually looks like in how studios price and in more detail in estimating software honestly. A spreadsheet of guessed prices ranked lowest-to-highest is a machine for selecting the most optimistic liar.

The good ones self-select out. Studios with full books and waitlists — the ones you're trying to find — decline RFPs, because the win rate doesn't justify the week. The response pool skews toward whoever needs the work. Needs the work is not a character flaw, but it's not a quality signal either.

Add it up and the RFP is a finely tuned instrument for hiring the best available proposal writer. If your project were a proposal, this would be perfect.

The better way: request for conversations

Here's the process we'd run if we were the client. It takes three to four weeks, costs one paid engagement, and selects for the thing you're actually buying: judgement.

Step one — write a two-page brief. Not 40 pages. Two. The problem, the context, the constraints, the budget band (yes, the real one — the budget conversation is free value, and hiding the number mostly hides it from the people who could help you spend it well), and what done looks like. Our guide to writing a brief that gets better proposals is the same shape: a brief that fits in two pages forces you to know what you're asking for.

Step two — shortlist three agencies from evidence, not applications. You don't need 30 RFIs. You need three names assembled from: work you admire (ask the client of that work how the engagement actually ran), people you trust who've bought this work before, and one wild card whose thinking in public — writing, talks, open source — shows a mind you'd want in the room. Our signals over showreels piece is a checklist for exactly this filtering: shipped durability, how they talk about failure, whether the people in the creds deck are the people who'll do the work.

Step three — conversations, not presentations. One hour each, working session format. Give them the brief a week ahead and forbid slides if you're brave — the deck is a proxy for the wrong skill. Ask them to interrogate the problem in front of you: what they'd need to know, what looks risky, what they'd push back on. You're watching for how they think about your problem, not how well they present their process. Red flags: they ask no hard questions, they agree with everything in the brief, they quote a fixed price in the room. Green flags: they tell you something in the brief is wrong and show their working.

Step four — paid discovery with two. This is the move that replaces the entire RFP circus. Take your two front-runners and pay each for a bounded, one-to-two-week discovery: interviews, technical audit, a shaped plan with a real estimate. Our discovery sprint playbook describes what good looks like on our side of the table; from your side, you're buying two things at once — a genuinely useful artefact (the plan, the audit, the estimate, usable whoever you hire) and two weeks of behavioural data no pitch can fake. Did they meet the people they said they'd meet? Was the estimate honest about uncertainty? Did the week end with clarity or with an invoice-shaped fog?

Step five — score on a rubric that rewards judgement. Total cost is at most a quarter of the weight. Suggested rubric, adapted from what we've seen procurement teams actually defend well internally:

Criterion	Weight	What evidence looks like
Problem understanding	25%	They reframed something correctly in discovery
Team & continuity	20%	Named people, allocation, who leaves when
Plan & estimation honesty	20%	Ranges with reasons, risks stated unprompted
Working evidence	20%	How the paid fortnight actually felt
Price	15%	Value for the plan proposed, not lowest number

Then pick one, and tell the other why they lost. That feedback costs you ten minutes and it's the only part of your old RFP that was ever worth keeping.

Keeping procurement happy

The objection arrives reliably, usually from someone legally obliged to raise it: "We have procurement rules. Three comparable quotes. Fair process." Fair enough — those rules exist to prevent corruption and waste, and the conversation process satisfies both better than the RFP did:

• Three comparable quotes: you have them — but they're quotes for shaped plans instead of guesses for a vague brief, which is what "comparable" was always supposed to mean.
• Fair process: define the steps and rubric before the conversations, apply them identically, document the scoring. Fairness is a procedure property, not a page count.
• Conflict of interest: the paid discovery is disclosed, equal in fee, and the artefacts are yours either way. That's cleaner than most RFPs, where the winning bid's "free" concepts were prepaid by every client who funded the agency's loss-leading pitch team.

If the rules are truly immovable — public sector frameworks exist, we've met them — then run the RFP differently inside the rules: prequalify on evidence and references, pay a market rate for any creative work requested, and score the paid-discovery stage as its own line item. Even a constrained process can stop paying the concept tax.

What it costs, and what it saves

The honest arithmetic: two paid discoveries at, say, $15–25k each, versus a "free" RFP. But the RFP isn't free. Your team writes 40 pages, reads fifteen 60-page responses, sits through nine decks, and picks a partner based on documents written by people other than the ones who'll do the work. Weeks of senior time, spent auditioning the wrong skill. And the risk cost is the big one: a wrong agency choice discovered in month three is the most expensive line item in this entire article, and it's the one RFPs maximise.

Two weeks working with a studio tells you more than any document ever written about working with that studio. You are buying years of collaboration. Audition the collaboration.

Key takeaways

• RFPs select for proposal-writing skill, idle capacity, and optimistic fiction — not building skill.
• Speculative concepts audition the pitch team and deliver the bench; you'd never hire a person this way.
• The replacement: two-page brief, three-agency shortlist from evidence, hour-long working conversations, paid discovery with two.
• Score on a rubric weighted toward judgement; cap price at ~15% and score it as value-against-plan.
• Procurement's real requirements — comparable quotes, fair documented process — are better served by shaped plans than by invented scope.
• The paid discovery artefact has standalone value whoever you hire; the true cost comparison is against choosing wrong in month three.

FAQ

Isn't paid discovery just agencies monetising the pitch? It's the opposite. The pitch is already priced in — every agency's rates carry the cost of its unpaid pitch team. Paid discovery makes the cost explicit, caps it, and gives you an artefact you keep: the plan, the audit, the estimate. You stop paying for pitches in every invoice forever and start paying once, visibly, for thinking you own.

What if both paid discoveries are excellent? Congratulations: you have two good plans, a tested relationship with your second choice, and no wrong answer. Pick on team chemistry and engagement model fit — and consider hiring the runner-up for a smaller piece of real work later. Two good options isn't a procurement failure; it's the best possible outcome.

Won't agencies just send their best people for the discovery and then swap? Some will try. Counter it contractually and socially: ask for the discovery team's names and allocation in the eventual proposal — the discoverers should become the project core in a senior studio — and check it against reference calls. Continuity questions asked at reference stage ("who actually did the work, and were they in the creds deck?") are harder to charm your way past than any clause.

We're a small team — can we run a lighter version? Yes, and it gets better at small scale, not worse. Skip to two conversations and one paid discovery week. The whole thing can run in a fortnight; it's one of the few procurement processes where the budget version is the luxury version, because speed and directness are the quality signals.