Strategy
Part of Primary market research: the version that survives contact with reality
Primary market research examples with the reasoning laid bare
Primary market research examples for 2027 show how interviews, observation, surveys, and pilots can answer different decisions without overstating evidence.
What to take away
- Zappos, Intuit, Xerox PARC, Coca-Cola, Shimano with IDEO, and Airbnb all ran research that public accounts describe in enough detail to copy the reasoning.
- Sip tests measure preference, pilots measure purchase behavior, and interviews explain why something happens rather than how often.
- Sample counts follow the claimsmall interview rounds for discovery, hundreds of responses before you quote a percentage.
- New Coke is from 1985 and the Shimano cycling work is from the mid-2000s. Read both as method illustrations, not current market evidence.
Why these cases still work as examples
New Coke's taste tests ran in 1985. Shimano's cycling project with IDEO ran in the mid-2000s. Neither tells you what buyers want now.
Each case pairs one decision with one method. That pairing is the transferable part. The published outcome belongs to that company, that market, and that year.
Primary market research means data you collect yourself for your own decision. That definition sorts the cases below into directly relevant and background reading.
Discovery: Intuit's home visits and Zappos's demand pilot
Zappos founder Nick Swinmurn photographed shoes in local shops and bought them only after an order arrived. The pilot recorded purchase behavior in one sales channel.
Intuit's Follow Me Home program sent staff into customers' homes to watch them use Quicken and QuickBooks. Researchers saw workarounds that surveys had missed. That answered a discovery question, not a sizing question.
Both teams chose the method that could change the decision. Neither treated a small sample as a market forecast.
Pricing: New Coke and stated preference
Coca-Cola's 1985 sip tests favored the sweeter formula. The company misread attachment to the original brand. A sip test measures preference at first taste, not repeat purchase over months.
Willingness to pay needs a different instrument. Conjoint analysis presents product bundles at set prices and estimates trade-offs from the choices people make. Pair it with a limited behavioral pilot before a price change.
Replace leading questions with neutral choice tasks and comprehension checks. Set stop conditions for complaints, exclusions, and unexpected cancellations.
Example: service design at Xerox PARC and GOV.UK
Lucy Suchman's copier studies at Xerox PARC found that printed instructions assumed a plan users did not follow. Design attention moved to the real task. The evidence came from watching people work.
The GOV.UK Service Manual's pop-up research guidance describes short in-context interviews and usability tests, and names the limits: brief sessions, participant selection, and recording constraints.
Teams weighing interviews against observation can read our qualitative research guide first, because the method has to match the decision.
Retention: Shimano, IDEO and Airbnb
Shimano asked IDEO why many American adults had stopped riding bikes. Researchers interviewed lapsed cyclists and watched their routines. The work led to the Coasting line for casual riders.
Airbnb's New York photography program sent staff door to door to photograph listings and talk with hosts. Booking behavior shifted in that market. The idea came from field observation, not a survey.
Leaver interviews explain why people left. They do not say how many would have stayed. Keep the two questions apart in the report.
Example-to-method map
| Decision | Primary method | What the result can carry |
|---|---|---|
| Zappos market entry | Demand pilot | Purchase behavior in one channel |
| Intuit product discovery | Home observation | Behavior in the homes observed |
| Coca-Cola formula change | Sip taste test | Preference at first taste |
| Xerox PARC service design | Copier observation | Task behavior in the settings observed |
| GOV.UK service redesign | Pop-up research | Short in-context sessions |
| Shimano and IDEO | Lapsed-rider interviews | Why people stopped, not how many |
| Airbnb host program | Field photography | Booking behavior in one market |
A market research strategy checklist turns that map into a sequence you can run before collection starts.
Sample size and participant counts
The count follows the claim you intend to write. Six interviews can establish that a problem exists. They cannot tell you what share of customers meets it.
Typical ranges quoted in practice, not rules: 5 to 8 participants per distinct user group for discovery interviews, and 5 to 8 per round for usability tests.
Roughly 300 to 1,000 completed surveys support a proportion reported with a margin of error near plus or minus 3 to 6 percentage points.
Sample size determination explains how confidence level and margin of error set that number.
Turning an example into your own test
Five steps, in order, before any recruitment starts.
- Write the decision and the threshold that settles it.
- Name the evidence gap and the group that holds the answer.
- Pick the method that matches the claim, and report each method separately.
- Set the participant count, the consent wording, and the stop conditions.
- Record your prediction, then compare it with later business results.
Keep the original notes and the recruitment record, so a later result can be traced back to its source.
Common questions
Are these cases real and verifiable?
Yes. They come from published accounts: Intuit's Follow Me Home program, Zappos's early inventory test, and Lucy Suchman's copier work at Xerox PARC. Coca-Cola's 1985 tests, Shimano with IDEO, and Airbnb's New York photography program are documented too. We did not run or commission any of them.
Why pair interviews with a survey?
Interviews tell you which reasons exist. A survey tells you how common those reasons are within a defined frame. Our guide to common primary market research questions covers the ones buyers raise first.
Can one case prove a best practice?
No. A case shows a process that worked once. You still have to test it against your market, your users, and your budget.
How many participants do I need?
Enough to support the sentence you intend to write: small rounds for discovery, hundreds of completions before you quote a percentage.
