Separate attention, buying intent and purchases
Search volume describes estimated search activity under a tool's settings. It is not a count of unique customers, delivered orders or units. One person may research several times, compare brands and never purchase. Another may buy without searching because a friend recommended the product.
Classify queries before interpreting their numbers. "How to clean a travel mug" describes an existing-owner problem. "Leakproof commuter mug" may describe product evaluation. "Travel mug replacement seal" can indicate demand for a component rather than a new mug. Mixing them produces a bigger number and a less useful business question.
Choose a product hypothesis, geography and purchasing situation. Record which queries plausibly support that hypothesis and which belong in a separate cluster. The goal is not to maximize a spreadsheet total. It is to understand the specific demand you would need to reach through your actual channels.
Use Trends and Keyword Planner for different jobs
Google Trends helps inspect relative interest and timing. Values are normalized within the selected geography and time range and scaled from 0 to 100. A value of 100 represents peak relative interest in that comparison, not 100 searches, 100 sales or universal popularity. Equal indexes in different regions do not imply equal absolute search volumes.
Keyword Planner offers historical metrics and advertising forecasts. Keep those separate: forecasts depend on bids, budgets, seasonality and historical ad quality. An estimated advertising click is still not a purchase. Save tool settings and the date alongside an export so the evidence can be reviewed later.
Use Trends to ask whether interest is sustained, seasonal or recently unusual. Use keyword evidence to examine relevant query language where available. A missing or low-volume trend series does not prove nobody wants a niche product. It may mean the tool cannot provide enough signal at that level of specificity.
| Source | Useful question | Invalid conclusion |
|---|---|---|
| Google Trends | When is relative interest higher under these settings? | This index predicts orders. |
| Keyword Planner | Which query clusters merit investigation? | Every search is a distinct buyer. |
| Interviews | What problem and workaround exist? | A small sample proves prevalence. |
| Offer test | Will this audience act at this price? | One test establishes long-term demand. |
Do not add every keyword into one market
A product can be described by material, size, use case, benefit and brand. "Insulated commuter mug," "leakproof travel cup" and "coffee mug for a train commute" may contain overlapping intent. Tools may group close variants or report related terms differently. Adding every row without understanding the methodology creates false precision.
Build intent clusters: product category, problem, use case, comparison and replacement. Read actual results to determine what the language refers to. Record uncertainty where the same phrase describes different products. Keep brand queries separate when you cannot plausibly capture brand-specific demand.
Separate countries, languages and time windows too. A worldwide yearly average can hide a seasonal peak in a country you cannot profitably ship to. For a gifting product, compare complete cycles rather than using the most exciting week as a launch forecast. The supplier's lead time belongs in the interpretation.
Demand Evidence Worksheet: an illustrative travel mug
This worksheet contains hypothetical observations, not current keyword metrics or a real market study. It uses qualitative signals so you can practice interpretation without mistaking invented numbers for a dataset. Replace each row with dated evidence from your geography and product proposition.
Suppose purchase-oriented queries exist, multi-year relative interest has recurring peaks, and interviews describe cleaning frustration. That supports testing a cleanable commuter proposition. It does not establish willingness to pay your price, acquisition cost or units you can sell. A viral video would add context but not repair those missing assumptions.
The useful output has scope and confidence: "There is enough aligned evidence to test this commuter use case in our region; willingness to pay and channel economics remain uncertain." That is more defensible than an unsupported sales forecast.
| Signal | Illustrative observation | Interpretation | Next verification |
|---|---|---|---|
| Search intent | Phrases focus on leaks and commuting. | A focused problem may be discoverable. | Inspect results and cluster intent. |
| Trend | Relative interest repeats seasonally. | Timing matters; a rise may not be structural growth. | Compare complete years in the region. |
| Interviews | Some participants describe awkward cleaning. | Usability may offer a differentiating angle. | Observe a sample test; avoid prevalence claims. |
| Video/community | A review discusses leakage and cleaning. | Useful vocabulary, not purchase proof. | Check independence, recency and relevance. |
| Missing proof | No priced offer test yet. | Willingness to pay is unknown. | Test the intended delivered price. |
Use social and community evidence as context
Videos, discussions and reviews reveal the words people use, comparisons they make and frustrations worth investigating. They help form better interview questions. They are much weaker for estimating a buying audience's size.
A high-view video may entertain people who never intend to buy. A discussion can be dominated by a few vocal contributors. Affiliate incentives, sponsorship, outdated products and regional availability change what an example means. Record these limitations rather than assuming every interaction is organic commercial demand.
Look for independent convergence, but do not double-count the same underlying story. Five posts reacting to one viral launch are not five independent demand signals. Prefer a small set of relevant observations to a pile of unrelated mentions. Customer pain is valuable evidence, not permission to manufacture a testimonial.
Express confidence without disguising uncertainty
Separate confidence in the problem, the proposed solution and your ability to reach buyers. You might have strong evidence for commuter frustration, mixed evidence for a lid design and none for acquisition economics. An overall adjective hides this unevenness.
Use an evidence log: observation, source, geography, date, inference, alternative explanation and next test. Label inferences Supported, Tentative or Unknown. These are practical judgments, not statistical confidence intervals. Explain why a label changes; do not improve it just because you spent longer researching.
An established store can add conversion, return and repeat-order behavior for related products. A new store cannot borrow those values from competitors. Neither should import an unrelated conversion rate and call the multiplication a demand forecast. The missing connection between searches and purchases is what a real offer test must investigate.
- Known: what the source measures and under which settings.
- Inferred: the product hypothesis suggested by the measurement.
- Unknown: conversion, capture share, delivered units and profit.
- Next: an experiment addressing the most consequential unknown.
Make a test decision, not a pretend sales estimate
Stop researching when you have a focused hypothesis, the main uncertainty and an affordable test. More screenshots of the same trend do not create independent evidence. A useful process narrows the next action instead of prolonging indecision.
BluePocket's Product Opportunity Reports organize available search, trend, keyword and customer/problem context into readable decision support. Source availability and quality vary. Use the report alongside supplier costs, samples and first-party experiments; it does not expose competitors' private sales or guarantee inventory returns.
Take the evidence into the pre-inventory scorecard. Decide whether to TEST, NARROW or SKIP, and document why before committing stock. The win is avoiding an expensive decision that depends on evidence you do not have, not claiming a product will sell an exact number of units.