Market evidence
Market signals versus demand proof: an evidence ladder
A product can appear on many pages, attract search interest and collect reviews without proving current unit demand. The useful question is what each evidence class permits the team to say and which next test reduces uncertainty.
Begin with the claim, then choose the evidence
My default is to write the strongest claim the team wants to make before collecting more rows. A listing can support presence at a source; a search series can support relative attention within its defined scope; a transaction record can support a bounded commercial outcome. None should be promoted into a stronger class by repetition alone.
Evidence is useful when it changes the next decision—not when it merely increases the row count.
Treat public observations as the lowest rung
Matched listings, prices, specifications, promotions and review counts can reveal patterns worth testing. Record source diversity, coverage, timing and match confidence. For the narrower rules around product-page availability, use the separate public-availability note rather than treating a visible label as stock data.
Search and attention signals
Google explains that Trends uses sampled, anonymised, aggregated and normalised search data. Values express relative interest for a selected geography and period; low-volume series can contain noise. Google says Trends is not a scientific poll and should be one data point among others.
Transaction proxies
Basket tests, lead forms, quote requests, dealer enquiries and reservations move closer to commercial intent, but each has selection, availability and conversion limits. Define the denominator and remove duplicates before comparing them.
Authorised first-party evidence
Orders, invoices, shipments, cancellations, returns, stockouts and lost-sale records support stronger conclusions only within their entity, channel, geography and period. Channel fill is not sell-through, and unavailable stock can suppress sales.
Build the evidence row
- Decision question and strongest permissible statement.
- Source class, coverage and timestamp.
- Product-match and collection confidence.
- Observed value and comparison basis.
- Alternative explanations and missing evidence.
- Next reversible test, owner and deadline.
Claims the evidence does not support
- Listing count is presented as assortment sales.
- Search interest is translated into units or market share.
- Review volume is treated as current demand without cohort context.
- Public availability is called inventory.
- Orders are called sell-through without cancellations, returns or channel context.
- Coverage limitations and unknowns disappear from the summary.