Pattern synthesis
Pattern synthesis without a fake sample
A pattern, in this guidebook, is a mechanism that shows up in more than one account, in the buyer’s language, with the condition still attached. It is the step between claude notes and positioning. It is not a word cloud, and it is not a license to say “most buyers.” trykux would rather keep a pattern small and true than wide and decorative.
The rule
Three accounts, one mechanism, the condition intact
“Everyone wants easier onboarding” is not a pattern. It has no mechanism, no role, and no limit. “Security reviewers stall when tenant isolation is explained only in the vendor’s nouns, and the champion cannot translate” can be a pattern if you heard that mechanism in separate accounts and can say which roles spoke. You still name the count. You do not promote the count into a share of the market.
Four things a pattern can be
01
A repeated stall
The same purchase friction at the same kind of touchpoint. Example shape: proof gaps that surface in security review, not in the first demo.
02
A repeated mismatch of roles
Champions describe a problem the economic buyer does not recognize, or the reverse. This pattern is about the buying group, which is why decision-maker research has to come first.
03
A repeated artifact
The object people trust: a questionnaire, a sandbox, a clause, a coworker’s warning. Market observation is usually how you notice the artifact. Interviews alone tend to forget it.
04
A repeated phrase
Only if the phrase still has a job. “We need a partner, not a vendor” is usually a slogan. A phrase earns its place in claude notes when it changes what the room can approve.
A procedure you can do on paper
Spread the claude notes so each account is visible as its own pile. Do not pool quotes first. Pooling is how one eloquent champion colonizes the study. Read across piles for a mechanism, not for a mood word. “Frustrated” will match everything and teach nothing. “Cannot forward the sentence” might match three piles and teach something.
Write the pattern as a sentence with a condition: when this role reaches this touchpoint, this mechanism stalls the purchase, unless that condition is already met. Then try to break it. Look for an account in the set where the condition was present and the stall did not happen. A broken pattern is a better result than a protected one. It tells you the condition was incomplete.
Record the disconfirming note in the brief. Hiding it is how qualitative marketing research acquires a reputation for saying whatever the sponsor hoped. trykux’s standard is the opposite. The disconfirming claude note stays on the page.
Language you should retire
“Customers consistently said” is retired unless you can replace “customers” with roles and “consistently” with a count of accounts. “The data shows” is retired, because these are notes, not a dataset with a frame. “ROI of insight” is retired entirely. Insight is not a security you can price from nine conversations.
What you may say is plainer and stronger. “In four of the six evaluations we sat with, the security reviewer stopped when the answer was a product noun. In one, a solutions engineer had already rewritten the answer in the bank’s own control language, and the review moved.” That is a pattern with a limit. It is ready to be considered for positioning. It is not ready to be a market statistic.
Group by account first. Themes that appear only after the piles are mixed are the ones to distrust.
Hand-off
A pattern that survives the break test can move to positioning synthesis. The writer’s job there is to keep the condition in the sentence. A pattern that does not survive goes back to customer interviews as a better question, not forward as a softer claim. The glossary keeps “pattern” from sliding into “trend,” which this guidebook does not use.