Lesson 10 of 10
Asking well
Ask Claude answers a question worth one of the ten a month it allows by naming one figure this account's own tools can look up — and checks the answer the identical way the rest of this course is checked.
Lesson nine explained why a figure in Claude's Conclusions can be trusted: it names a fact rather than writing a number, and a sentence that cannot name one is thrown away before Marleigh & Rowe ever sees it. Ask Claude is the same machinery, held to the same rule, pointed at a question this account actually typed rather than a reading run on a schedule. Everything below is about the two things that machinery alone does not decide for you: what makes a question worth asking, and how to read what comes back.
The same tools, the same figures
Ask Claude answers by calling the exact functions every page in this course has already walked through — the same revenue, the same contribution, the same break-even and target ACoS lesson six built. An answer about this window’s revenue can never disagree with the Profit page, because both are reading the same arithmetic rather than two separate ones that happen to agree today.
A question worth one of ten
The allowance is ten questions a month, and it is spent by an answer that actually checks out, not by an attempt — see the panel on the Ask Claude page itself for exactly how that count is kept. A question earns its place by being plain and single: one thing about this account’s own figures, not a request for an opinion nothing here can check, and not several questions folded into one sentence, since a follow-up is a fresh question against the same allowance rather than a continuation of the last one.
Take a real row from this account’s own table in lesson five: Marleigh & Rowe 8.5" x 11" Self-Healing Cutting Mat - 5 Pack - Double Sided - Rotary Quilting Mats for Workshops and Classrooms, on Amazon.ca, is the thinnest margin among the 12 trading rows shown there, at 33.8%. “Which product’s margin is thinnest this window, and why?” is a question worth asking — it names one figure, over one window, that this account’s own tools can actually look up and break down by product. “How can I make more money?” is not, however reasonable it sounds out loud: nothing behind this feature can price an opinion, only a figure it already has a definition for.
Reading an answer that cites its sources
Every figure in an answer arrives as a plain number, already checked — the citation itself is never shown, the same way the placeholder behind a figure in this course’s own tables is never shown either. Ask for a chart and the same rule holds one step further: the chart is drawn from the exact rows the question’s own tool call returned, never from a number written free-hand to illustrate the point. A nomination for a chart that points at nothing a tool call actually returned is not drawn at all, and the answer says so rather than quietly leaving a gap.
What “could not be checked” and “stopped early” mean
Neither is a failure. An item marked as unable to be checked was a sentence or a chart that named a figure the verifier could not confirm, left out rather than shown with a caveat attached to it. An answer marked as having stopped early ran into the limit this feature places on how much it will look up for one question before it had checked everything it set out to — and says so plainly, rather than presenting a partial look as though it were the whole picture. Both are the same honesty this whole course has practised from lesson one onward, applied to a feature that talks back.
Why an attempt costs nothing
A question that gets declined, one that fails outright, or one where every figure it tried to cite failed the check, spends none of the ten — only an answer that actually held up counts against the allowance. That is worth knowing before asking a first question: there is no penalty for a wording that comes back empty, so a question worth asking is worth asking plainly, on the first try, rather than saved for a moment that feels safer to spend it on.