What Market Research Answers for a Small Business

What to take away

Small business market research questions are usually about method: which evidence answers this decision, and what that evidence cannot show. The answers below cover the mechanics, the ethics and the report itself.

What Census Business Builder actually does

The Census Bureau's Census Business Builder overview describes the tool as a way to pick a business type or NAICS code and a geography, then view resident and business data for that area.

The geography can be a metro, county, city, ZIP Code or census tract. Choosing a tract instead of a metro changes the count more than most owners expect, because the tract is where the customers actually walk in.

Write down five things before you estimate anything: the NAICS code, the geography, the data release, the unit of measure, and the exclusions. A count of establishments in a ZIP Code is not a count of your competitors, and it is not demand.

The tool is a U.S. starting point. It gives you a denominator, not a total addressable market. Reachable demand needs eligibility, access, frequency and your own capacity applied on top.

Here is that chain with numbers attached. The count is 1,200 households in one tract. An eligibility rule, such as owner-occupied and inside one income band, cuts it to 480. Access, meaning inside your service radius and not already tied to a competitor, leaves 300.

At two visits a year, those 300 households produce about 600 visits a year. Multiply by your own average ticket to get an annual revenue figure. Substitute your own count and rules; the shape of the arithmetic is the point.

Once demand is bounded, the budget question moves to positioning and channels, which small business marketing strategy covers.

Competitor research without crossing a line

Study direct providers, substitutes, the workaround customers use instead, and the choice to delay. Doing nothing is a competitor.

Record what public evidence shows: audience, offer, price basis, and availability. Also record proof, geography, and constraints. Published menus, posted rates, permit filings and job ads are all fair game.

Do not impersonate a customer to get a quote, bypass access controls, or solicit information someone is not free to give. That is not research, it is a liability, and it corrupts the data besides.

Which question goes to which method

The table is the point of the exercise. Most wasted research spend is a "why" question sent to a survey, or a "how many" question sent to six interviews.

Interviews, surveys and search interest

There is no universal interview count. One workable pattern is 12 interviews spread across four participant types, three each. Keep going while each new interview adds a pattern, a variation or a counterexample. Stop when three in a row add nothing new. Report who was missing.

Survey size follows from several factors. These include population, design, the estimate you want, and expected variation. Other factors are precision, confidence, weighting, and subgroups. Nonresponse also matters. None of those have a default.

For a 5,000-customer population at 95 percent confidence and a margin of 5 points, the textbook calculation gives about 370 completed responses. The arithmetic is 5,000 / (1 + 5,000 x 0.05 x 0.05). It assumes the widest spread of answers.

A tighter expected spread lowers the count. If one invitation in five comes back, invite about 1,850 people to land 370.

Google Trends shows sampled, normalized interest scaled from 0 to 100, not raw search volume. It separates topics from exact terms and asks you to compare like with like.

Treat a pattern as relative interest for one query, place and period, then check it against customer and commercial evidence. Small business content marketing trends covers which evidence is worth building pages around.

What the report contains

The question, the method, the sample, the sources with dates:

A one-page report still carries the decision, the count or pattern behind it, the source dates, and the limitation.

Every important number traces to an original source and a documented transformation. If you cannot reproduce a figure from the report alone, it does not belong in the report.

Cost and time depend on how many participant types you recruit and how hard they are to reach. Set a stop rule before you start.

The GAO evaluation design guide ties evaluation questions to evidence needs and design choices. Apply that discipline when you work through how to research small business market research questions. Federal guidance does not make a local marketing result causal or transferable.

When you draft the consent step for interviews or surveys, use the W3C Privacy Principles statement for shared vocabulary. Record what you collect, why you hold it, how long you keep it, and who can see it. Review your own data collection against the governing law and your configuration.

Common questions

How many customer interviews are enough?

There is no universal count. Recruit the relevant range of participant types and stop when the decision has enough well-documented coverage of patterns, variation and counterexamples. Report who was not represented.

A 12-interview pattern across four types, three each, is one workable starting shape. Keep adding interviews until three in a row raise nothing new. If each one raises a new counterexample, coverage is not there yet.

How large should a market survey be?

Calculate it from the population, the sampling design, and the precision you need. Also include the confidence level and expected response distribution. Consider subgroup needs, nonresponse, and decision risk:

Worked from a 5,000-customer population at 95 percent confidence and a 5 point margin, the same formula gives about 370 completed responses. At one response in five, invite about 1,850 people.

What belongs in a market research report?

A decision record should cover these elements:

• The decision, scope, assumptions, sources and dates, participants, methods, instruments, field dates, analysis, contradictions, limitations, economics, conclusion and next action. Every number should be reproducible from what the report itself states.