Why Our Six-Week Builds Now Freeze Scope on Day Nine, Not Day One

The client almost never knows what they want on day one. They know on day eight, after two homepage drafts. We used to freeze scope at kickoff anyway, which produced the worst week-five rebuilds we've ever shipped. Here's what we changed, what it cost, and why we still aren't sure day nine is the right number.

18 October 2026 · 9 min read

The client almost never knows what they want on day one. They know what they want on day eight, after they've seen two homepage drafts side by side and realised the first one made the booking widget look like an afterthought and the second one made the room photography feel cramped. Freezing scope before that moment — which we used to do routinely, at the end of the kickoff call, in writing, counter-signed — is the single most expensive thing we did to ourselves for about three years running.

So we stopped. Mostly. The scope-freeze step in our six-week builds now sits at the end of day nine, not day one. This post is about what that actually changed on the ground, not a pitch for the build itself.

Why we used to freeze on day one

The reason was almost entirely about feeling disciplined. A signed scope at the end of kickoff looks like project management. It gives the delivery lead something to point at in week four when the client says "can we also add a gift-voucher flow". It makes the fixed price feel defensible. All of that is real, and none of it is wrong on paper.

The problem is that the artefact clients were signing on day one was a wireframe deck and a page list. Clients cannot read wireframes. Every designer who has sat across from an SMB founder with a wireframe deck knows this, and the industry still pretends otherwise because the alternative is uncomfortable. The specific failure mode: client approves the wireframes in the kickoff meeting, nods seriously at the page structure, agrees to the integration list, signs the document, and then three or four weeks later when the first staging link lands in their inbox they react as if they are seeing the design for the first time. Because they are. The wireframe was abstract geometry. The staging link is their actual hotel, with their actual photographs, and their brain finally has something to react to.

Late 2023, a boutique hospitality client in Jaipur — ten keys, direct bookings only. We did the day-one freeze properly. Scope locked, page list signed, the rooms-and-rates module specced as a single-page listing with per-room modal detail. Four weeks of build. We shipped the staging link on a Thursday. By Monday morning the client was back with a completely different mental model of how rooms should surface — they wanted seasonal rate cards, package combinations (room + breakfast + airport pickup as a single bookable unit), and a comparison view across room types. None of that was in the signed scope. All of it was obvious, in hindsight, for a property that sold 60% of its inventory as packages during wedding season. We rebuilt the module. Four days. Unbilled, because we'd signed the scope, and the honest reading was that we should have asked about packaging on day one and hadn't.

That rebuild is the pattern. Not that specific one — the shape of it. Every two or three builds we'd hit a week-five rebuild of somewhere between four and six days, almost always in the module the client cared most about commercially. The signed scope didn't protect us. It just gave us a theoretical right to refuse the rework, which we never actually exercised, because refusing the rework meant shipping a site the client would quietly resent for two years.

What day nine actually looks like now

Day nine is the end of the second design iteration. First iteration lands around day four or five (homepage plus one interior template, usually the one with the most commercial weight — rooms page for hospitality, product page for e-commerce, service detail for a studio). Client reacts, we take notes, Mira reworks. Second iteration goes out around day eight. We sit with the client on day nine — on a call, screen shared, no deck — and we talk through what actually changed between iteration one and iteration two, and what they're noticing now that they couldn't articulate on day one.

At the end of that call we produce a two-page document. That document is the new scope lock. It has:

  • A wireframe snapshot of the agreed homepage and the agreed primary interior template — not the full Figma, just a flattened PNG with annotations
  • The integration list, named explicitly: payment gateway (usually Razorpay or Cashfree, specified by name and plan), booking engine where relevant, GA4, WhatsApp Business API or the click-to-chat fallback if they're not ready for the API
  • A named list of every page in the build, with a word-count ceiling against each one (e.g. "About — 450 words", "Rooms index — 180 words plus per-room 220 words × 6 rooms")
  • What is explicitly not in scope, written as a short list, in plain English, with the understanding that any of these can come back as a change request at a stated day-rate

Two pages. Signed. That's the freeze. The rest of the build — three weeks and change — runs against that document, and in our experience the week-five rebuild largely stops happening. Not entirely. We still get the occasional "can we also…" but it lands as a change request against an artefact the client genuinely understood, not against a wireframe they'd nodded at four weeks earlier out of politeness.

The reason day nine works — insofar as it does — is that by then the client has seen their own content inside two different design directions. They've had the small internal argument with their co-founder about whether the hero should lead with the view or the room. They've looked at the rooms module twice and noticed, on the second viewing, that they sell packages and the current layout doesn't accommodate that. The reaction that used to happen in week five, when the staging link landed, now happens on day eight when the second iteration lands. We absorb it inside the design phase, where absorbing things is cheap.

The commercial cost we had to swallow

One extra design iteration costs us somewhere between ₹18,000 and ₹25,000 in internal time that we don't bill — Mira's hours, the review call, the production of the lock document. Call it a full designer-day plus change. That's real money for a studio our size and it comes straight out of margin on a fixed-price build.

What we used to spend on week-five rebuilds, before this change, was ₹60,000 to ₹1,20,000 of unbilled dev time per affected project — four to six developer-days at our internal cost, and that's before counting the morale tax of asking the build team to rip out and redo work they'd shipped clean. Not every project had a week-five rebuild. Maybe one in three. But the ones that did were disproportionately painful, and the ones that didn't still carried the risk of one, which meant we were pricing a buffer into the fixed number that we couldn't honestly justify to the client.

So the maths, roughly: we eat ₹20,000 of extra design time on every build, in exchange for not eating ₹80,000-ish on roughly one in three builds. The expected value works out, give or take. More importantly, the fixed-price number we quote now actually reflects the work we're going to do, which is a thing we were quietly lying about before — not deliberately, but the buffer was there.

Riya pushed back on this hard when we first proposed it, and she wasn't wrong to. The arithmetic that matters to a project lead is calendar, not rupees. Day-one freeze gave us a five-week build window after kickoff. Day-nine freeze compresses that to four weeks and three days. On a project that's already tight — and a six-week build with real content production is always tight — losing a week and a bit of execution window is not nothing. She was overruled, partly, by the fact that the week-five rebuilds were killing her weekends more than the compressed build window was. But she still thinks day eleven would be safer than day nine and we haven't fully resolved that argument. (It's possible she's right. We'll see.)

One honest caveat. This doesn't work for every client type. Enterprise procurement teams — the kind who need a signed SOW with scope annexures before they can raise a PO — still get a day-one freeze, because the alternative is that the project doesn't start. For those builds we price in the rebuild risk explicitly, usually as a 15-20% contingency on the dev line, and we tell the procurement contact what it's for. Sometimes they accept it, sometimes they push back and we end up absorbing it anyway. That whole category of client is a separate conversation — procurement-driven buying has its own pathologies and this post isn't the place for them.

The thing we actually don't know. Day nine is the end of iteration two in our studio, with one design pod, with Mira drawing. If we hire a second designer next year — which we've been talking about on and off — iteration two might land on day seven, or day eleven, depending on who's drawing and how fast they move through revisions. In which case the scope-freeze point drifts with them. Which means day nine isn't really a principle, it's a side-effect of our current staffing, dressed up as a process decision. The honest answer is that we don't know yet whether the right rule is "freeze at the end of iteration two, whenever that lands" or "freeze on calendar day nine regardless", and those two rules diverge the moment the design team gets bigger. We'll find out the hard way, probably, the first time a build runs with a different designer and the scope-freeze call lands on a day that doesn't match the number we've been quoting internally for the last year —

Ready to build something that works?

Get a free website mockup for your business. No commitment, no fluff — just what it could look like.

WhatsApp replies in seconds.

WhatsApp Call Free mockup
WhatsApp us · replies in seconds