Operations
Part of Customer research: how journey maps expose backstage constraints
Customer research case studies reveal which decisions actually held up over time
Real research programs at the UK's GDS, Intuit, Spotify and the Co-op show which customer decisions survived contact with customers and which did not.
What to take away
- These customer research case studies cover GDS, Intuit, Spotify, the Co-op and Microsoft, and each turns on a decision that was tested against customers.
- GDS tests services with disabled participants as a standing requirement, not a final check, and publishes the method.
- Intuit's follow-me-home studies watch people in their own offices, where workarounds show up that lab sessions hide.
- Spotify's failure was a research question, not a sampleit asked listeners to approve a feature before they had any reason to want it.
- The Co-op's funeralcare work shows a serious barrier affecting a minority still counts when averages look fine.
- Microsoft's Xbox Adaptive Controller research with disabled gamers held up because the design kept multiple input options.
- A decision survives when the evidence behind it can be re-checked later by someone who was not in the room.
Real cases exist, and they are more useful than composites because the outcome is not yours to arrange. The five below are documented public programs. Each is a decision that was tested against customers, and each shows what held up and what did not. Start with customer research fundamentals if the vocabulary is new.
GDS: accessibility as a research requirement
The UK Government Digital Service runs research sessions with disabled people under published guidance for sessions with disabled people. It covers individual needs, accessible logistics, respectful communication, and assistive technology.
Accessibility Testing Timing
Continuous testing
- When barriers found
- Throughout build
- Cost to fix
- Cheap to change
- Outcome
- Caught early
Pre-launch audit
- When barriers found
- Late, before launch
- Cost to fix
- Cost a rebuild
- Outcome
- Same barriers repeated
The decision that held up: accessibility testing is a standing requirement, not a final audit before launch. Teams that treated it as a gate found the same barriers late, when fixing them cost a rebuild. Teams that ran sessions throughout caught them while the service was still cheap to change.
Intuit: follow-me-home over lab sessions
Intuit built its early QuickBooks and TurboTax research on follow-me-home studies, watching small business owners and taxpayers work in their own offices and kitchens. The method is simple: sit with the person while they do the real task on their own machine.
What held up was the setting. In a lab, people tidy up. At home, they show you the spreadsheet taped to the wall and the shoebox of receipts. Those workarounds became product decisions. Lab-only testing had been producing clean feedback about a workflow nobody actually ran.
Spotify: the decision that did not hold up
Spotify's 2022 discovery-mode test is the clearest public failure in this set. The company tested a TikTok-style vertical video feed in its main navigation, ran it with a subset of users, and reversed it after the response.
Asking listeners to approve a feature before they have a reason to want it produces rejection, while asking what they were trying to do when they opened the app produces something usable.
A neutral task beats a concept approval prompt every time. More examples of that pattern sit in customer research examples.
Co-op: a minority barrier that outranked the average
The Co-op's funeralcare research is a documented case of a serious barrier affecting a smaller group. Most families completed the arrangements process without trouble. A smaller group, often dealing with sudden or contested deaths, hit a wall that the aggregate numbers hid.
The team reported the affected context and severity separately, fixed the process for that group first, and did not treat the average as the threshold. When you compare your own results against customer research benchmarks, check whether the benchmark covers the groups your decision affects.
Microsoft: multiple inputs beat a fixed design
Microsoft's research with disabled gamers shaped the Xbox Adaptive Controller, released in 2018 at a US list price of $99.99. The device has 3.5mm jacks and USB ports for external switches, buttons and joysticks.
The decision that held up was to keep the controller open to third-party assistive technology instead of locking it to one layout. Gamers with limited mobility could use the controls they already owned. That choice has outlasted the launch window because new switches and mounts still plug in.
Audit the published claims
The FTC advertising substantiation policy requires a reasonable basis before an objective advertising claim goes out. Apply that standard to any performance number you publish from research, and take advice on the facts.
The FTC guidance for marketers using reviews covers fake feedback, selective requests, conditioned incentives, hidden relationships, and paid rankings. Those checks apply when research content names providers. Longer write-ups of published programs are collected in secondary market research case studies.
Case-study decision map
| Case | Decision tested | What held up |
|---|---|---|
| GDS | Accessibility as standing requirement | Continuous testing beat pre-launch audit |
| Intuit | Home observation over lab sessions | Setting exposed real workarounds |
| Spotify | Concept approval for a new feed | Failed; neutral tasks work better |
| Co-op | Average completion as threshold | Failed; minority harm reported separately |
| Microsoft | Fixed layout versus open inputs | Open input options held up past launch |
What the five have in common
Each program tied evidence to a named decision and an owner. Each recorded its limits. None of them treated a favorable average as the whole answer.
The pattern to copy is the traceability, not the outcome. A reviewer who was not in the room should be able to follow the evidence to the decision and challenge it.
Common questions
Are these real company case studies?
Yes. GDS, Intuit, Spotify, the Co-op and Microsoft are named organizations with public documentation of these programs. Where a detail is not public, this article says so rather than filling the gap.
Why does the Spotify case count as research failure?
The company tested the right users with the wrong question. Listeners had no reason to want a video feed, so their rejection said little about the feature and a lot about the prompt.
Can a minority finding outweigh an average?
Yes, when the affected context, severity, rights or customer harm crosses a threshold the team defined in advance. Define that threshold before fieldwork, not after the numbers arrive.
What makes a research decision survive scrutiny later?
A named decision, a relevant sample, a method that fits the question, and traceable evidence. Honest limits, and an owner for the customer-facing change.
