Simple Ethnography: Watching a Learner Take a Course Inside Someone Else's Store

Every usability session we’d run on the learner experience happened the same way: a test account, a demo course, our own staging environment, a learner recruited through a panel, task list in hand. It’s a reasonable setup for catching UI bugs, and it had caught plenty. What it structurally couldn’t catch was anything about the actual context a real learner experiences a course in — embedded inside someone else’s branded storefront, reached by clicking a random product link, surrounded by that merchant’s other products, promotions, and design choices, none of which our staging environment reproduced. Simple Ethnography is the Liberating Structure built for exactly that gap: instead of bringing the user to a test environment, you go watch them in the real one, doing what they’d do anyway.

What it is and how it runs

Simple Ethnography, in the Liberating Structures formulation, strips ethnographic fieldwork down to a lightweight, repeatable practice any team can run without formal training: pick a real setting where the behavior you care about actually happens, go there, watch quietly, and record what you observe using a simple structure — separating what you see and hear (objective observation) from what you interpret it to mean (your inference), and explicitly noting what surprises you. The discipline of keeping observation and interpretation in separate columns is the whole method; conflating them is the single most common way informal “let’s go watch some users” sessions produce weak, over-confident conclusions instead of real findings.

A session is typically short — twenty to sixty minutes of quiet observation, in the actual environment, followed immediately by a structured debrief where the observer(s) compare what they wrote in each column and specifically discuss the surprises, because a surprise is a strong signal that the team’s existing mental model of the user is wrong somewhere.

What we watched

We arranged, with a willing merchant’s permission, to sit in on one real learner enrolling in and starting a beginner cooking course embedded in that merchant’s storefront — a real customer, doing this because she’d actually decided to buy the course, not because we’d recruited her for a test.

Observed: She found the course through a “you might also like” carousel while browsing the merchant’s cookware pages, not through any course-specific navigation. She clicked in, read the course description, then — before enrolling — scrolled back out to look at the merchant’s product reviews page, cross-referencing whether the course creator’s cookware reviews looked legitimate. She enrolled, and the confirmation screen took her to a lesson player that used our platform’s default neutral styling — visually distinct enough from the merchant’s storefront branding that she paused for three or four seconds, said “oh,” and re-checked the browser tab to confirm she was still on the same site.

Interpreted: The “oh” pause and the tab-check were, in our read, a small but real trust hiccup — a jarring shift from a fully-branded storefront experience to a generic, platform-branded course player, at exactly the moment (immediately post-purchase) where reassurance matters most. Her earlier detour to the merchant’s product reviews, before enrolling, suggested that for a course embedded in a storefront she wasn’t already loyal to, she was evaluating the merchant’s general credibility, not just the course description — a form of trust transfer our course-discovery flow had never been designed around, because it assumed a learner arriving at a course page already trusted the source.

Surprise: The single biggest surprise was how little the “course” framing mattered to her navigation at all. She never used a course category page or search; she found it exactly the way she’d find any other product, through a recommendation carousel, and evaluated it the way she’d evaluate any other product — via the merchant’s own reputation signals, not course-specific ones like instructor bio or curriculum preview, both of which existed on the page but which she scrolled past without pausing.

What changed because we watched instead of tested

None of this would have surfaced in a staging-environment usability test, because a staging test can’t reproduce “a learner deciding whether to trust an unfamiliar merchant” or “a learner discovering a course via a product recommendation carousel embedded in that merchant’s actual catalog” — both depend on the real storefront context existing around the course, which by definition a demo environment strips away.

Two concrete changes came out of the debrief. First, the lesson player got a lightweight theming hook — inheriting the merchant’s primary brand color and logo in the header, rather than defaulting to fully neutral platform styling — specifically to close the “oh, wait, am I still on the same site” gap at the exact moment we’d watched it happen. Second, we deprioritized a planned instructor-bio redesign (more prominent credentials, more detailed curriculum preview) in favor of investigating whether merchant-level trust signals — store reviews, “verified merchant” badges — could be surfaced directly on the course card itself, since that’s the trust signal we’d actually watched a real learner go looking for.

Why watching beats testing here

A usability test answers “can a user complete this task in this interface.” Simple Ethnography answers a different, often more important question: “what does the real surrounding context do to how a user approaches this at all” — and that second question is invisible in an isolated test by construction, because isolating the test is removing the surrounding context. For an embedded product like a Shopify course app, where the course is never encountered in isolation from a specific merchant’s storefront and reputation, that context is not a minor variable — it may be the dominant one, and it took one real observation, not a battery of staging tests, to see it.

Failure modes and when to skip it

The most common failure is letting observation and interpretation blur together in the moment — writing “she was confused” instead of “she paused, said ‘oh,’ and checked the browser tab” throws away the raw material a debrief needs and quietly substitutes the observer’s assumption for what actually happened. Keep the columns separate, especially in the room, not just when writing up.

It also requires genuine, low-intrusiveness access to a real setting, which isn’t always available — a merchant has to be comfortable with an observer present (even remotely, watching a session recording with consent), and a single observed learner is one data point, not a validated pattern; treat what you see as a strong hypothesis to test further, not a conclusion to build a roadmap on unchecked. And it’s the wrong tool when you need to compare a specific interface change against an alternative under controlled conditions — that’s what a usability test or an A/B test is for; Simple Ethnography tells you what’s actually happening in context, not which of two options performs better.