Supplement · 3 · Technical influence
Saying the True Thing, Usefully
A decision nobody understands is not a decision, it is a delay with documentation. This supplement is about the sentences: how to present an architecture choice to executives, give a forecast that is honest and still credible, and escalate in a way that produces an answer.
Mission tie-in: lesson 0050 insists the truth stays cheap to see. That only works if someone can say it in a room where the vocabulary is money, dates and risk — without softening it into meaninglessness.
Why this is here
The trigger is a specific kind of bad meeting. You have the analysis — the drivers, the option table, the ledger — and thirty minutes with people who will fund or stop the work. Twenty-five minutes in, someone says “so are we on track?” and everything you presented is compressed into a yes or a no that you did not intend to give.
Three failure modes produce that moment, and all three are avoidable. Burying the decision: the ask arrives on slide fourteen, after the context nobody needed. Hiding the uncertainty: a single date is quoted because a range “sounds unprepared,” and the range then arrives anyway, later, as a surprise. Hiding the trade: the scope reduction is described as a “phasing decision” and the cost of that phasing never gets said out loud, so nobody consents to it.
The remedy is not charisma. It is a small set of structures — a memo shape, a range with a named variable, an options table, an escalation format — and the discipline to keep the cost visible in every one of them.
Knowledge: five conversations, five shapes
1. Presenting a decision to executives
Executives are not a less technical audience; they are a differently accountable one. They are optimising a portfolio, so what they need from you is the shape of the commitment: what it costs, what it forecloses, when it becomes irreversible, and what you need from them. Depth of technical reasoning is available on request — it is not the opening move.
Lead with the decision and the ask. Everything else is support.
Buries the decision
“I want to walk through the returns architecture. We looked at three approaches for carrier integration. The first was a direct integration with each carrier, which has some advantages around control… the second…”
Twelve minutes in, no one knows whether they are being informed, consulted, or asked. Attention is spent on orientation instead of on the question.
Leads with the decision
“We’re buying carrier integration rather than building it. That’s a €24,000-a-year commitment and it hands the team back about four engineer-weeks of capacity a year. I’m not asking you to approve it — it’s mine to decide. I am asking you to tell me if a vendor dependency on the label path is a problem for you, because that’s the one objection that would change it.”
Decision, price, decision rights, and the specific input wanted — in four sentences.
Note the third sentence in the good version. Naming who owns the decision prevents the most common executive-meeting failure: a decision that was yours gets accidentally re-opened, or one that was theirs gets accidentally made by you. State it explicitly, every time, using the decision-rights language from lesson 0048.
2. Explaining uncertainty without sounding unprepared
The fear is that a range reads as ignorance. It reads as ignorance only when it is bare. A range with a stated confidence, a named driving variable, and a date on which it narrows reads as control — because it demonstrates that you know precisely what you do not yet know.
Sounds unprepared
“It’s hard to say. Somewhere between four and ten weeks? It depends on a lot of things. We’ll know more once we get into it.”
No confidence level, no driver, no narrowing date. This is the version that makes people ask for a single number instead.
Sounds in control
“Eighty per cent confident the pilot is live between 23 September and 14 October. The spread is almost entirely guest-order support: in, it’s the late date; out, it’s the early one. We decide that on 8 September, and the range narrows to about four days that afternoon.”
Same uncertainty, but bounded, attributed to one variable, and with a date when it collapses.
A point estimate does not remove uncertainty; it transfers it silently to the listener, who then plans as if it were a commitment. When it moves, they experience a broken promise rather than an updated forecast — and the trust they withdraw is not about the date, it is about whether your numbers mean anything. A range does the opposite: it makes the uncertainty a shared object that both of you can act on. The listener can now decide whether to hedge, and you can be right about the range even when you are wrong about the day.
The named variable is what converts the range from a hedge into a plan. “It depends on many things” gives no one a lever. “It depends on guest orders, decided 8 September” gives your sponsor something they can pull — and gives you a date on which you have promised to be more precise, which is the part that reads as competence.
3. Communicating scope options
“We can’t make the date” is a problem. Three dated options with named costs is a decision. Build the table before the meeting, and make sure at least one option is one you would not choose — a table where every road leads to your preference is a presentation, not a choice.
| Option | Scope | Pilot live (80% range) | What it costs |
|---|---|---|---|
| A | Full: registered plus guest orders, plus exchanges | 4 Nov – 2 Dec | Nine weeks past the 30 September commitment; the quarter’s support-cost target is missed entirely. |
| B (recommended) | Registered-customer domestic returns, automated labels | 23 Sep – 14 Oct | Guest orders — 29% of return demand — keep the current 72-hour path until the next increment. |
| C | B without automated carrier labels; ops desk creates labels manually | 9 Sep – 18 Sep | Manual review queue rises from ~34/day to roughly 120/day at 10% exposure. This breaks the stated guardrail of 50/day and needs two additional ops staff or a guardrail change. |
Option C is the important row. It is the fastest option, it is genuinely available, and it breaks a guardrail the initiative committed to. Presenting it with that cost stated in bold is how you stay honest about a trade you do not want, rather than quietly omitting the option and hoping nobody asks. If the sponsor chooses C anyway, they have chosen it with the guardrail in view — and that is a legitimate decision that you can execute without resentment.
4. Negotiating quality, scope, time and risk without hiding the trade
Pressure arrives as a request to move one variable while holding the others still. The dishonest response is to accept and absorb the difference in unstated quality — skipped tests, an unrehearsed rollback, a security review deferred. That is not a compromise; it is a loan taken out in someone else’s name.
Hides the trade
“We’ll do our best to hit the 30th. We might have to cut some corners on testing but we should be OK.”
“Some corners” is not a decision anyone can consent to. The risk has been accepted on the sponsor’s behalf without telling them, and the first they will hear of it is the incident.
Names the trade
“I can hold the 30th if we drop guest orders — that’s option B. What I can’t do is hold the 30th and full scope: the only remaining variable would be the rollback rehearsal and the accessibility pass, and I’m not willing to trade those without you deciding it explicitly. If you want to, say so and I’ll write down what we accepted.”
The refusal is specific, the alternative is concrete, and the escape hatch is real but requires an explicit, recorded decision.
The last clause matters more than the refusal: “say so and I’ll write it down.” It converts an implicit risk transfer into a recorded one. Most quality trades survive only in the dark; almost nobody will sign their name to “we shipped the returns flow without an accessibility pass” once it has to be written into an ADR.
5. Dissent, and disagree-and-commit
A team that never disagrees is not aligned, it is quiet. The lead’s job is to make dissent cheap to express and expensive to leave unresolved. Two moves do most of the work: solicit the objection by name, and record it when it is overruled.
Suppresses dissent
“Any concerns? … No? Great, we’re agreed.”
An open floor with a five-second pause reliably produces silence, which is then reported upward as consensus. The objection still exists; it now surfaces as slow delivery instead of as words.
Surfaces dissent
“Priya, you argued for building the integration. I went the other way. Can you state the strongest version of your case, so it’s in the ADR — and tell me what evidence would show I got this wrong within a quarter?”
Names the dissenter, asks for the case at full strength, records it, and defines the signal that would reverse the decision.
Then commit properly. Disagree-and-commit is a real obligation on both sides: the decider owes the dissenter a recorded rationale and a revisit trigger; the dissenter owes the decision genuine execution, not compliance. The phrasing that carries both: “I still think we should have built it. We’re buying it, that’s decided, and I’ll make the adapter work. If the vendor misses the latency driver twice in the pilot, I’m reopening it — and that trigger is in the ADR.”
† The revisit trigger is what makes commitment sincere rather than performed. Without one, a dissenter’s only route back to the argument is for the project to visibly fail — which gives them an interest in it failing. Name the trigger and you have removed that incentive entirely.The five-minute decision memo
Most of the above collapses into one artifact. Write it before the meeting; if you cannot fill in every field, you are not ready to ask for the decision. Read aloud it takes about two minutes, which leaves three for the actual question.
- Decision
- One sentence, active voice, in the present tense. “We buy carrier integration rather than building it.”
- Decision rights
- Who decides, who advises, who is merely informed. State whether this memo is a notification or a request.
- The ask
- Exactly what you need from the reader, and by when. “Confirm by Thursday that a vendor dependency on the label path is acceptable.”
- Why now
- What makes this the moment — and when the option expires or gets more expensive.
- Options considered
- Two or three, each with its cost and the reason it was not chosen. Include the one you would not pick.
- Cost accepted
- What gets worse if this is right. Every real decision has one; a memo without this line is advocacy.
- Reversibility
- How hard is undoing it, and until when? Name the date the door closes.
- Evidence
- The numbers, with their assumptions. Link the ledger; do not paste it.
- Dissent
- Who disagrees and their strongest argument, in their words rather than yours.
- Revisit trigger
- The observable signal that reopens this decision, and who watches for it.
Escalating with a concrete decision request
An escalation that asks for “visibility” or “support” produces sympathy and no change. An escalation is a request for a specific decision or resource, from a named person, by a named date, with the consequence of no answer stated plainly.
Produces sympathy
“Flagging that we’re blocked on the Payments team for the settlement contract. It’s been a couple of weeks and it’s starting to affect our timeline. Any help appreciated.”
No decision requested, no date, no named owner, no cost. The most likely outcome is “thanks for the heads-up” and another two weeks.
Produces a decision
“Decision needed from you by Thursday 11th. The RefundSettled contract has been waiting 14 days against Payments’ published 3-day review commitment. That has cost us 9 working days on slice A, which carries €1,600/week of delay. Two options: (a) Payments reprioritises the review this week, or (b) we ship the pilot reading settlement from the monolith directly, which we’d need to unpick in Q1 — roughly 40 hours of rework. I recommend (a). If there’s no decision by Thursday, we take (b) by default and I’ll record the rework as debt with a repayment trigger.”
Named decider, date, quantified cost, two real options, a recommendation, and a stated default. It is answerable in one line.
The default is the sharpest instrument in that paragraph. A stated default converts silence from a way of avoiding a decision into a way of making one — which is exactly the pressure needed on a queue that is not yours.
| You hear yourself saying | What it actually does | Say instead |
|---|---|---|
| “We’ll try to hit the date.” | Accepts the date while privately planning to miss it. | “I can hit it with scope B; I cannot hit it with scope A.” |
| “It should be about six weeks.” | Transfers your uncertainty to the listener silently. | “80% confident between five and nine weeks; the driver is X.” |
| “There’s some technical debt here.” | Names a category, requests nothing, changes nothing. | “This costs us six hours a month; repaying it takes three days.” |
| “Just flagging that we’re blocked.” | Transfers the problem without transferring a decision. | “Decision needed by Thursday; options A or B; default is B.” |
| “Everyone’s aligned on this.” | Reports silence as agreement. | “Priya disagrees, here is her case, here is the revisit trigger.” |
| “We might have to cut some corners.” | Accepts risk on someone else’s behalf, unrecorded. | “The remaining variable is the rollback rehearsal. Do you accept that?” |
- “Executives want confidence, so give them a single date.”
- They want to be able to plan, which is not the same thing. A single date they cannot rely on is worse than a range they can hedge against — and the second time a confident date moves, every subsequent number you give is discounted regardless of how good it is.
- “Simplifying for a non-technical audience means leaving out the trade-off.”
- Simplify the mechanism, never the consequence. “We hold the order system as the source of truth, so returns can be briefly out of date after a refund” is understandable by anyone and contains the whole trade. Removing the second clause is not simplification, it is omission.
- “If I present the options well, the right one will be obvious.”
- Options with different risk profiles are chosen by whoever carries the risk, and that may not be you. Your job is to make the trade visible and state a recommendation; if the sponsor picks differently, they have exercised a decision right you do not have.
- “Disagree-and-commit means the argument is over.”
- It means the argument is paused against a named trigger. A commitment with no route back is not commitment, it is suppression, and it converts a colleague’s expertise into a grievance with a long memory.
- “Escalating makes me look like I can’t handle it.”
- Escalating vaguely does. Escalating with a decision request, a date, a cost and a default reads as someone managing a dependency they do not own — which is precisely the lead-level behaviour being looked for.
When this is the wrong move
- The decision is yours and reversible. Do not build a memo and book a meeting for something you can decide in ten minutes and undo in an afternoon. Ceremony scaled to reversibility is lesson 0046’s rule; it applies to communication too.
- The relationship, not the argument, is the problem. If a stakeholder does not trust the team, a better options table will not fix it. Deliver one small thing you promised, on the day you promised, and try again with the table afterwards.
- You do not yet know the answer. Ranges are for genuine uncertainty, not for avoiding work. If the spread is wide because nobody has looked, the honest message is “I’ll have a range on Friday after a two-day spike,” not a range invented on the spot.
- The audience needs the depth. An architecture review with peers is not an executive briefing. Leading with the conclusion and withholding the reasoning insults the room and wastes the best review you will get.
- During an incident. Structured options and revisit triggers are peacetime instruments. In an incident, say what is broken, what you are doing, and when you will next speak. Everything else waits.
Skill: choose the sentence
A sponsor asks for a date. The most credible answer names a range, plus:
Genuine disagree-and-commit requires the decider to supply:
What most reliably turns an escalation into an actual decision?
Practice: rehearse on your own initiative
- Write one five-minute decision memo for a decision you are actually facing, filling every field. The field you cannot complete — usually “cost accepted” or “revisit trigger” — is the part of the decision you have not made yet.
- Convert your current forecast into a range. State a confidence level, name the single variable that drives the spread, and name the date on which it narrows. Then check: is the variable actually the biggest one, or the one you find easiest to talk about?
- Build a three-option scope table with real dates. Include one option you would not choose, with its cost written out in full. If every option flatters your recommendation, rebuild it.
- Find the trade you are currently absorbing. Name one quality that has quietly become the flexible variable on your team — tests, accessibility, rollback rehearsal, security review. Write the sentence that hands that decision back to whoever should be making it.
- Rewrite your last escalation in the concrete format: named decider, date, quantified cost, two options, a recommendation, a default on silence. Compare the two and note which one you would have answered.
- Solicit one dissent by name. Pick a decision you made recently and ask the person most likely to have disagreed for the strongest version of their case. Record it, with a revisit trigger, wherever the decision lives.
- When did you last give a single date you knew was a range? What did it buy you, and what did it cost when it moved?
- Which trade-off are you currently absorbing rather than presenting — and who would be surprised to learn it is being made?
- Whose dissent on your team is currently expressed as slow delivery rather than as words? What made speaking it more expensive than not?
- Look at your last escalation. Could the recipient have answered it in one line? If not, it was a status update wearing an escalation’s clothes.
Reveal: the consent test
After any decision conversation, ask yourself one question: if this goes badly in six months, will the people in that room say they knew what they were agreeing to?
If the answer is no — if the cost was mentioned once, quietly, in a subordinate clause, or if the option that broke a guardrail was never shown — then you obtained agreement rather than consent, and the difference will surface exactly when you most need the relationship. Everything on this page is machinery for making that answer yes: the cost line in the memo, the bold guardrail in option C, the recorded dissent, the stated default. None of it is about being persuasive. It is about being confirmable later.
Your win
You can open with the decision and the ask, express uncertainty as a bounded range with a named driver and a narrowing date, put a scope option you dislike on the table with its cost in full, refuse a hidden quality trade while offering a real alternative, record dissent with a revisit trigger, and escalate in a form that can be answered in one line.
Read deeper
- Crucial Conversations, Patterson, Grenny, McMillan & Switzler — making it safe to say the difficult thing, and the difference between agreement and consent.
- Thinking, Fast and Slow, Kahneman — part III on overconfidence and the planning fallacy: why point estimates feel more credible than ranges and are usually worse.
- How to Measure Anything, Douglas Hubbard — ch. 5 on calibrated estimation: how to give a 90% interval you would actually bet on.
- The Manager’s Path, Camille Fournier — the tech-lead and staff-engineer chapters on communicating upward and escalating well.
- Software Architecture in Practice, Bass, Clements & Kazman — ch. 21 on architecture evaluation with stakeholders: the option table as a shared artifact rather than a pitch.
- “Is High Quality Software Worth the Cost?”, Martin Fowler — the argument to reach for when quality is being treated as the flexible variable.
- Course: lesson 0046 (ADRs and scaling ceremony to reversibility), lesson 0048 (decision rights, escalation triggers), lesson 0050 (forecast honesty). Portfolio: the executive summary format. Reference: decision rights, ADR.
Have your agent-teacher play a sceptical CFO against your decision memo — then check whether your answers added information or only added confidence.