Foundations & core concepts of customer research

In 2004 LEGO was weeks away from running out of cash. The prevailing internal view, backed by trend research and confident forecasts, was that children raised on video games would never again sit still for a 500 piece build. So LEGO made the bricks bigger, the sets simpler, the builds faster. Sales kept falling. What turned the company around was a different kind of looking: researchers sent to spend time with families in Germany and the United States, watching children play over days rather than asking them what they liked. What they saw contradicted the forecasts. Children were chasing mastery, and mastery needs difficulty and time.

Both inputs were called research. Only one was evidence about behaviour. This lesson sets the vocabulary the rest of the module runs on: what qualitative and quantitative each buy you, where the line sits between evidence and opinion, and who actually counts as a customer.

What customer research actually means

Customer research is the systematic collection and analysis of information about the people who buy, use, or reject your product, so you can decide how to position, price, message and build. The word "systematic" is load-bearing. Reading five G2 reviews is not research. Systematic means a defined question, a defined group of people, a method chosen on purpose, and an output written down somewhere the whole go-to-marketgo-to-marketThe strategy defining how you'll launch a product: target segments, channels, value proposition and coordinated action plan.View full definition → team can act on it.

Three things you are trying to establish with precision: the problem the customer is trying to solve, the words they use for that problem, and what made them pick you or somebody else. Everything else in this module is machinery for getting those three answers reliably.

Sub-concept 1: qualitative vs. quantitative research

Qualitative research gives you the why and the wording. Quantitative research gives you the how many and the how often. They answer different questions, and the common failure is treating one as a cheaper version of the other.

Qualitative work is interviews, observation, session recordings, open-text analysis: small numbers of people, studied closely. Ten interviews with churned customers will tell you why they left and in what emotional register they felt the pain. It cannot tell you how widespread that reason is. Quantitative work is surveys, product telemetry, funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → data, pricing tests: large numbers of people, measured shallowly. A survey of 500 users tells you what share of the base shares that reason. It cannot tell you what the reason means.

LEGO's turnaround shows the asymmetry. The quantitative trend data on shrinking attention spans was not fake, but it was a measurement of something adjacent to the question. Only ethnographic observation, watching a specific child in a specific bedroom, produced the finding about mastery that gave LEGO permission to make sets harder again. Numbers told them something was wrong. Watching told them what.

Order matters more than proportion. Qualitative first to find out what to measure, quantitative to size it, qualitative again to explain the anomalies. April Dunford, who has spent more than two decades on positioningpositioningThe mental space you want your brand to occupy in your target customer's mind relative to alternatives.View full definition → work, argues that a handful of well-structured customer interviews produces more positioning clarity than survey data read without them.

Sub-concept 2: evidence versus opinion

An opinion is a claim about customers held by someone inside the company. Evidence is an observation of what customers said or did, with a record of where it came from. "Our customers want a self-serve tier" is a hypothesis wearing the clothes of a finding. It becomes evidence when you can name the customers, the date, the method and the raw quote.

The UK Government Digital Service built its whole operating model on that distinction. When GDS was set up in 2011 and GOV.UK replaced Directgov, Business Link and a long tail of departmental sites in 2012, the first of its design principles was to start with user needs, not government needs. In practice that meant a written user need behind each piece of content, and content removed when no need could be shown. The service manual goes further and tells teams that everyone, including developers and senior stakeholders, should spend at least two hours every six weeks watching someone use the service. The point of that rule is political rather than methodological: an executive who has watched a real person fail to complete a form stops arguing from taste.

Three tests for whether you are holding evidence. Can you say who said it? Can you say what question they were asked? Would a colleague reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → the same reading of the transcript? If any answer is no, label it a hypothesis and keep it in a separate list.

Sub-concept 3: who counts as a customer

"Customer" is looser than it looks, and most weak research is research aimed at the wrong person.

At minimum, separate four groups. The user operates the product day to day. The buyer signs and holds the budget. The blocker (security, legal, procurement, IT) can stop the purchase without ever using the thing. And the non-customer, the person who evaluated you and went elsewhere or did nothing, usually holds the most useful information and is the hardest to reach. In B2B these are often four different people with four different definitions of success.

GDS pushed the definition wider still. A government service has no opt-out audience: the people who most need it include those with low digital confidence, poor connectivity or a disability, and the Service Standard requires teams to research with them and to provide assisted digital support rather than designing for the confident majority. Commercial teams rarely need to go that far, but the discipline transfers: if your research pool is drawn only from customers who answer emails and enjoy talking to marketing, you have sampled your fan club.

An ideal customer profileideal customer profileIdeal Customer Profile: a precise description of the company or customer type that gets the most value from your product and is most likely to buy and retain.View full definition → is the operational answer to who you should be researching and selling to: company size, sector, technical setup, buying trigger, and how badly the problem hurts. It is not a personapersonaA semi-fictional, research-based representation of your ideal customer: their goals, frustrations, behaviours and decision criteria.View full definition → with a stock photo and a first name. Segment, which sells customer data infrastructure and therefore has a stake in this argument, learned the lesson the hard way. Its founders' first product, a classroom feedback tool, failed. The open-source analytics library they had built as scaffolding was the thing developers actually used, and the company that resulted was built around those developers rather than the customers the team had imagined.

Sub-concept 4: jobs-to-be-done and voice of customer

Two terms you will meet throughout the rest of this module.

Jobs-to-be-done, associated with Clayton Christensen, holds that customers hire a product to make progress on a job. The canonical example is the McDonald's morning milkshake, bought less as a drink than as something that occupies a boring commute and lasts until lunch. The research move is behavioural: watch what people do around the purchase, and ask what they fired to hire you.

Voice of customer means the exact words customers use about their problem, captured verbatim and kept unedited. Not your paraphrase. Sources include win/loss calls, onboarding recordings, support tickets, sales call transcripts and review sites. The value is that customer phrasing removes translation work for the next reader, which is why VoC belongs in a repository rather than in one person's notebook.

Real-world cases

LEGO, 2004 onwards: The ethnographic programme did not just save the product line, it changed who got to make claims. Sets became more complex again, instructions longer, and the adult builder was recognised as a real segment rather than an oddity. The mechanism worth copying is cheap and unglamorous: send people to observe in the setting where the product is actually used, and give the observation the same standing in the room as the sales forecast.

UK Government Digital Service, 2011 onwards: GDS made user research a named profession with its own career path, embedded researchers in delivery teams rather than in a central insights function, and published its methods openly in the service manual. That last choice matters for anyone building a research practice: written, public standards mean a new team does not renegotiate what counts as evidence every time it starts work.

Segment: The pivot from a failed classroom tool to analytics infrastructure came from reading behaviour rather than intent. Nobody had asked for the library; developers simply kept using it. Usage data is qualitative-adjacent evidence in that sense, a record of what people chose when no researcher was watching, and it is the cheapest correction available to a team whose stated customer and actual customer have drifted apart.

CMO action items

• Write your current beliefs about your customers as a numbered list of hypotheses, then mark each one E (backed by a dated, sourced observation) or O (opinion). The ratio is your starting diagnostic.
• Book four to six win/loss conversations this quarter, split between buyers who chose you in the last 90 days and buyers who chose someone else. Record, transcribe, and keep the customer's phrasing intact.
• Start a shared VoC repository that sales, support and marketing all add to weekly, and read it before any campaign brief is written.
• Adopt the GDS two hour rule in some form: put your executives in front of real users on a fixed cadence rather than in front of a summary slide.

Common mistakes that kill results

Calling internal consensus a finding. LEGO's simpler-sets strategy was not lazy work. It was a plausible story that everyone senior believed, and it cost the company years. Belief held by many people inside the building is still opinion.

Researching the wrong person. If your interviews only reach end users, your messaging will delight the person who logs in and lose the person who signs. MapMapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → the buying committee first, then design the sample to include the buyer and the blocker. Government teams face the same trap in a different shape: the official who commissions a service is never the citizen who has to complete it at 11pm on a phone.

Sampling only the people who like you. Happy, responsive customers are easy to recruit and tell you least. The people who left, and the people who never arrived, hold the information you are missing.

Treating research as a project with an end date. Markets move, competitors enter, priorities shift. Rolling interviews at a quarterly cadence keep the vocabulary current; a research deck from two years ago is a historical document.