An engineering advisory engagement letter contains six things, and all six are written down before the first invoice is sent. There is the question — the specific decision, read, or artefact the work will answer. There is the corpus — the documents, code, and conversations the studio will read in order to answer it. There is the deliverable — the memo, the decision record, or the architectural review. There is the named senior reader who will sign the deliverable. There is the window — the calendar weeks inside which the work happens. And there is the closing date — the day on which the file is closed, after which neither side has a standing claim on the other. None of the six is novel. The discipline lies in writing all six down in plain English on a single page.
We have come to think of the letter itself as part of the work, rather than its administrative wrapper. A senior engagement that cannot be described on one page tends to be an engagement no one has thought through. The act of writing the letter forces the unspoken assumptions onto the page. Sometimes the engagement is reshaped at that point. More usefully, sometimes it is declined.
What sits on the page
The letter opens with the question. We try to write it as a sentence the operator could read aloud to a board without translation. Should the team build the orchestration layer in-house, or should it adopt a commercial vendor for the same function? Was the eleventh-of-March incident caused by the change made on the ninth? Is the architecture, as currently drawn, defensible to a future technical due diligence? The question is not the brief. The brief is what the studio will read, where the studio will go, and to whom the studio will speak. The question is the shape of the answer the operator is paying for.
After the question comes the corpus. It is a short, named list — the architecture diagrams as they stand, the decision record from January if one exists, read access to the relevant repositories, two hours with the head of platform, one hour with the engineer who owns the affected service. The corpus is short by design. A scoped read whose corpus runs to forty documents is, almost always, a scoped read whose question has not been narrowed enough. The studio's job in the letter is to draw the boundary tightly; in the engagement, to read everything inside it.
The deliverable is then named in a single phrase. A written memo of three to five pages, signed by the named senior reader, with a recommendation, the reasoning, and the conditions under which the recommendation should be revisited. The memo is a document the operator can keep, re-read in nine months, and hand to a future board member or to the next senior engineer joining the team. We prefer the memo over the slide deck for a simple reason: the slide deck is rarely re-read, and the memo, when written well, is.
The named senior reader is then committed to in the same sentence as the deliverable. The studio does not staff scoped engagements out to whoever happens to be free that week; the engagement is a piece of one person's senior judgment, with the studio's name behind it. If the named reader cannot deliver inside the window, the engagement is paused and the operator is told; another reader is not silently substituted. This is the part of the letter the operator tends to quote back to us a year later.
The window comes next, and is short. Two to four weeks for most reads, six for a scoped diligence with code in scope. The window is calendar-bound rather than effort-bound; the deliverable is due on the date written, not after a vague number of hours have been worked. If the window has to slip, the studio writes to the operator before it slips, and the cause is named.
The closing date is the last line on the first page. It says, in one sentence, that the engagement ends on that date and that the studio will not invoice for follow-on work without a fresh letter. The line is small and load-bearing. It distinguishes the scoped engagement from the ambient retainer, and it is the sentence the studio has most often had to defend with operators trained, by previous suppliers, to expect the relationship to drift. The engagement closes. If the next question wants a senior read, a fresh letter starts.
The letter is the engagement. If the engagement cannot be written on one page, neither side has yet decided what the engagement is.
A redacted example
The shape is easier to see on a page than in description. A real letter from earlier this year, with the client and the system redacted, opened with the question — does the proposed orchestration layer warrant being built in-house, or should the team adopt a commercial vendor for the same function? The corpus that followed was short: the architecture diagram dated the fourteenth of February, the build-versus-buy memo prepared by the head of platform on the twenty-first, read access to the prototype repository, a two-hour conversation with the head of platform, a one-hour conversation with the engineer prototyping the in-house option, and a one-hour conversation with the operator. The deliverable was named in a single phrase: a written memo of three to five pages, signed by the named senior reader, with a recommendation, the reasoning, and the conditions under which the recommendation should be revisited. The reader was named. The window was three calendar weeks from the date of countersignature. The closing date was the Friday at the end of the third week, on which the final version of the memo would be delivered and the file would be closed. No follow-on work without a fresh letter.
The letter was one page. The fee sat on the second, the terms on the third. The whole document fitted inside a single email attachment and was countersigned inside two business days. The engagement ran as written. The memo was delivered on the named Friday. The file was closed.
Three observations are worth making about the example. First, the question is small; the studio has written letters whose questions are larger, and those tend to be letters whose engagements run longer. Second, the corpus is named in full at the start, and nothing was added to it during the engagement without a written change to the letter. Third, the closing date is not an option — it is the end. A follow-on read on a related question would have begun under a fresh letter, on its own merits, rather than appended to this one.
The studio's stance
We will not start work on a scoped engagement that has not been written down on one page. The reason is not legal caution. It is that the work which is not described on a single page is the work that drifts, that quietly expands, that ends without a clean record, and that neither side is later able to point to as a piece of senior judgment well delivered. The page is the discipline. When the operator and the studio can both look at the same six lines and recognise them as the engagement, the engagement is one we are willing to take.
When the page cannot be written — because the question is not yet a question, because the deliverable is not yet a deliverable, or because the operator wants the studio to be on hand without naming what the studio would do — we say so, in writing, and decline. The firm whose engagements cannot be described on one page is the firm whose engagements, three years on, no one will remember the value of.