Small Business Market Research: Trends, Cohorts, Competitors

What to take away

What Google Trends actually measures

Trends draws on sampled, anonymized, categorized and aggregated search data. Google normalizes the result to a 0 to 100 scale within the query you run. That scale is relative interest, not a count of searches.

Change the geography, the time window or the category and the index moves even when demand has not. A term that returns zero may simply sit below the sampling floor. Low-volume local terms hit that floor often.

Treat Trends as one signal. It tells you when interest rises and falls, and which of two terms your market searches more. It cannot tell you how many people searched, or whether they bought.

Reading Trends without fooling yourself

Compare like with like. Same region, same period, same category, same device mix. A term compared against itself across two years is a cleaner read than two different terms in one window.

Seasonality explains most spikes in trades with weather, school terms or tax dates. Plot two full years before you call anything a trend.

Then validate. Pull your own numbers for the same weeks: quote requests, calls, bookings, web form fills. If Trends rose and your leads did not, the interest is real but it is not reaching you. That gap is a channel problem, not a demand problem.

For a broader frame on how search, advertising and retention fit together, see small business marketing strategy.

Cohort views for retention

A retention rate for the month tells you what happened to everyone at once. It blends new buyers with ten-year regulars and hides the leak.

Split customers into monthly cohorts by first purchase. Track what share of each cohort buys again in month two, month three, month six. Cohorts that decay faster than the ones before them are your warning.

Use a spreadsheet such as Google Sheets and define the metric before comparing groups: M1, M3, and M6 mean the first, third, and sixth month after the first purchase. At each checkpoint, divide the cohort members who made a repeat purchase during that month by the cohort's original first-purchase count. Use the same interval and denominator for every cohort, and do not compare a cohort at M6 until it has six months of follow-up. For subscriptions, report active customers at each checkpoint and cancellations separately; define “active” consistently.

The same split works on cancellations. Group by signup month and count how many are still active at each interval. A flat line means you are replacing churn with new sales and standing still.

Cohort work supports customer retention for small businesses because it separates the customers you keep from the ones you keep buying.

Map competitors and substitutes

List direct competitors, indirect solutions, internal workarounds and doing nothing. Delay is a real substitute, and often the cheapest one for your customer.

Record the target customer, the offer, the price basis, and proof.
Record availability, service area, channel, and constraints. Use public evidence only: their site, their posted prices, their reviews, their job ads.

For local competitor discovery, search Google Maps and Yelp by service and neighborhood, then verify each result against the business's website and Google Business Profile. Record where and when you found each detail. For local population and industry context, use U.S. Census Bureau data.census.gov to examine American Community Survey data and County Business Patterns. These sources describe population and establishment patterns; they do not identify every competitor or prove demand.

Reviews show what customers notice and what goes wrong. Job ads show where they are investing. Both are public and both are checkable.

Do not misrepresent yourself, bypass access controls, scrape against terms, or ask a competitor's employee to disclose protected material.

Interview recent buyers, not friends

Recruit people who can answer the question: recent buyers, lapsed customers, qualified nonbuyers, decision-makers, staff. Friends and current customers are convenient and systematically exclude the people whose behavior decides the outcome.

For structured screening or survey questions, Google Forms can provide a consistent way to collect answers. Keep interview prompts tied to a specific recent event, and record how participants were recruited so you can distinguish interview themes from representative estimates.

Ask about one specific recent event. What triggered it, what they did first, and what alternatives they considered. What they paid, what nearly stopped them, and what happened next.

Never ask whether someone would buy an undefined product. Test a choice instead: book a slot, request a quote, join a waitlist. Explain the costs and conditions first, and never take payment for something you cannot deliver.

Protect the people who answer

Say who is running the research and why. Explain what participation involves. State whether you record. State what you pay. Explain how you use and share the data. State how long you keep it. Explain how someone withdraws.

Collect only what the decision needs. Restrict access, secure storage and transfer, separate identifiers where you can, and delete on schedule.

Health, finance, employment, children and other sensitive topics need qualified review before you start. The W3C Privacy Principles statement gives designers a shared vocabulary and warns against shifting privacy work onto the individual. It does not replace the law, your contracts or a review of your actual setup.

Decide, record, revisit

Write down the question, recommendation, alternatives, evidence, method, sample, sources, dates, calculations, limitations, and dissent. Link every number to its origin.

State what would reverse the recommendation and when the evidence goes stale. A decision record with no reversal condition is a report.

For evaluation questions and evidence needs, the GAO evaluation design guide explains how design choices follow from the question. It is written for federal program evaluation. Use its discipline as a check on your method, not as proof that a marketing result is causal.

When you buy a tool, the GOV.UK technology selection guidance recommends choices that can change over time, preserve data control, address security risk and include ownership cost. Those are public-service rules, not private-sector mandates, but they make good buying questions.

The U.S. Small Business Administration's market research and competitive analysis guide separates existing sources from direct consumer research and frames demand, market size, location, saturation and pricing as research questions. Verify the exact market and data for your own decision.

Market research decision record

Common questions

Does Google Trends show search volume?

No. Trends reports relative interest on a 0 to 100 scale, normalized within the query you run. It shows when interest rises and which term your market favors, not how many people searched.

How many responses do I need?

It depends on the estimate, the population, the sampling method, the expected variation and the precision you want. A large convenience sample can estimate the wrong population very precisely. Report how you recruited, who answered and what you cannot claim.

For a probability sample estimating a proportion, a common starting calculation is n = z² × p × (1 − p) ÷ e², where z is the selected confidence multiplier, p is the assumed share, and e is the desired margin of error. This calculation assumes probability sampling; it does not make a convenience sample representative. For local buyer interviews or opt-in forms, report who you recruited and use responses to identify themes, not to claim a population-wide percentage.

What is a cohort view in plain terms?

Group customers by the month they first bought, then track what share of each group buys again at set intervals. It shows whether newer customers behave worse than older ones, which a single monthly average hides.

Can research prove a launch will succeed?

No. It reduces uncertainty and exposes assumptions. Execution, price, competition, timing, capacity and outside conditions still decide the result.