The short version: from this month, every custom-development contract we sign carries a Delivery Handbook 2.0 as an annex. There is no "deep industry expertise" or "full-stack enablement" inside — only three things: three points that require a signature, one two-week rhythm, and a written answer for every "not passed". We publish it because in more than twenty projects last year, the arguments were almost never about technology. They were about what counts as done.
The first version was two pages: a milestone table and a defect-severity note, kept as a project manager's private memo. Over two years every project changed it once — each "you never told us that" from a client became a line in the handbook. Version 2.0 gathers those practices into three chapters: how to start, how to run, how to close.
1 · Three sign-offs: no signature, no next phase
| Gate | Entry criteria | Who signs | If it fails |
|---|---|---|---|
| 1 · Requirement baseline | Spec, prototype and quote agree; every screen is clickable | Client project owner + our PM | Back to discovery; no extra fee, schedule shifts |
| 2 · Test entry | Self-test pass rate met; no open P1; deploy script runs on an empty environment | Our QA + dev lead | Not accepted into integration testing; dev resolves internally first |
| 3 · Go-live | UAT signed; historical data reconciled both ways; rollback rehearsed | Client business owner + our consultant | Go-live postponed; new cutover window scheduled |
The point of the gates is not to block people; it is to move the argument to a moment when change is still cheap. The typical bad project looks like this: nobody reads the prototype during requirements, the first demo happens two months later, and someone says "this is not what we meant". Changing at that point costs three times more; not changing means acceptance drags on forever. Gate 1 exists so that sentence gets said in week two.
2 · The rhythm: three touchpoints for the whole project
- D1 planning (45 min). Your product owner sets priorities; we confirm what these two weeks will and will not include.
- D5 mid-sprint demo (30 min). Half-finished but clickable and operable. Issues are logged on the spot, not saved up for acceptance.
- D10 acceptance (60 min). We walk the acceptance sheet line by line: sign, or write down precisely what failed.
Everything else is written: a weekly status note covering what was done, what is next, and which risks need your decision. Anything agreed verbally in chat is overridden by the defect log and the change order — not out of stubbornness, but because verbal promises leave no trace and both sides remember them differently a month later.
3 · The change budget: 10-15% reserved in the contract
No custom project ships without change. The handbook's answer: reserve 10-15% of the contract value as a change budget for the small adjustments that surface before go-live; beyond that, changes go through a written estimate — who asked, which modules, how many person-days, does the date move — one page, your decision, we only build what you sign.
The rule has saved both sides. A warehouse project raised 23 changes during UAT: 19 were absorbed inside the budget, 4 went through estimates and pushed the date by two weeks. Had all 23 been treated as "just a quick tweak", that project would have slipped two months and nobody could have explained where the time went.
Writing down what counts as done is the cheaper option for everyone. A vague contract reliably ends with the vendor doing more work and the client waiting longer.
4 · You can use the handbook to audit any quote
Every role and every deliverable in the handbook should map to a line in the quote. If a quote carries "project management, 15%" but cannot say what project management does, which documents it produces and who owns each stage, take chapter 2 of the handbook and ask.