Turning this into an operating cadence

The cycle, repeated

You have now run the loop once. The practice is that same loop on a monthly cadence: re-measure, diagnose the largest remaining gap, ship one change matched to its failure mode, verify, record.

The prompt set should stay stable for several cycles so movement stays comparable, then be revised deliberately rather than continuously — when your category shifts or when a prompt has been settled in your favour long enough to stop being informative. Do not delete prompts that show zero presence. A prompt you have never won is where the opportunity is; deleting it only improves the average.

What to report upward

Report the experiment, not the dashboard. What did we believe, what did we change, what happened, what do we now know that we did not know last month.

That framing survives contact with a sceptical executive far better than a visibility score, because it is falsifiable and it accumulates. It also protects you when a month produces an inconclusive result — you still learned something, and you have the record to prove the process is working even when the number did not move.

Connect it to the commercial question when you can. Presence on a recommendation prompt in your core category is worth more than presence on a category-definition prompt, and saying so keeps the conversation about pipeline rather than about a number.

What to stop doing

Stop reading single answers as evidence. Stop shipping bundles of changes and attributing the result to whichever one you liked best. Stop reporting the aggregate score as the outcome. Stop deleting the prompts you lose.

And stop treating advice — including everything in this course — as true without testing it in your own category. The point of building a measurement loop is that you no longer have to take anyone's word for it, including ours.

Where this goes next in SolCrys

• Paid tiers add scheduled daily tracking and more engines, which shrinks the variance problem by giving you many more samples.
• The MCP server exposes your workspace to agents, so the monthly read can be drafted programmatically rather than by hand.
• More Deep Analyses and Content Audits per month means the diagnosis step stops being rationed to one.

Lab: Schedule the next cycle

No quota consumed.

1. Put the next cycle in the calendar, one month out.
2. Write down the gap you will target next and why it is the highest-value one remaining.
3. Start a running log of experiments — hypothesis, change, outcome, effect size. Three entries in, it will be more useful than any external best-practice list.

Check yourself

• What is the highest-value gap remaining, and what failure mode is it?
• Can you state last cycle's result in the what-we-believed, what-we-changed, what-we-learned form?
• Which of the four stop-doings is your team most guilty of?

Go deeper

This lesson is the spine. These guides are the depth.

Continue

• Previous: Signal or noise: the third run
• Last lesson — back to the course overview