Ethnographic Research SOP

An ethnographic research SOP puts you in the room where the problem happens, and a round of 6 to 10 site visits takes one to three weeks. You watch a nurse finish a shift, a baker plan tomorrow's orders or a warehouse clerk reconcile stock, and you write down what they do, what they skip and what they curse at. People describe their work one way in an interview and do it another way in real life. Observation catches the gap, and the gap is where products get built.

What is the ethnographic research SOP and when should you use it?

Ethnographic research means observing people in their own setting while they do the task you care about, with short questions asked in the moment. The startup version borrows from contextual inquiry, the method Hugh Beyer and Karen Holtzblatt describe in Contextual Design: you act as an apprentice and the customer acts as the master who shows you the work.

Use this SOP:

• When the job is physical or happens on site, such as retail, healthcare, logistics, hospitality or trades.
• When interviews give you vague answers like "it's fine, we manage", and you suspect a hidden workaround.
• When the customer does the task so frequently that it became invisible to them. People forget the sticky notes and spreadsheets they built years ago.

If you need people to record their own experience over days, the pain point diary SOP is lighter. If you only need to hear their story, start with a customer interview.

What do you need before you start?

• One observation goal in a sentence: "See how independent pharmacies handle prescription refills on a busy morning."
• 6 to 10 hosts who match your customer profile and agree to let you watch.
• A consent form that covers notes, photos and audio, and says how you store and delete them. In the EU, recordings of people fall under GDPR.
• A field note template (below) on your phone or a paper notebook.
• Permission from whoever owns the space, such as a manager or clinic owner, if your host is an employee.

How do you run it, step by step?

1. Pick the moment worth watching

Choose the time of day or week when the problem peaks: Monday morning, month-end close, the dinner rush. Ask hosts in a short call beforehand when the task is hardest. Two hours at peak teach you more than a full quiet day.

2. Recruit hosts across different settings

Vary size, location and experience. Six sites from one chain show you one company's habits. Offer something useful in return: a summary of what you saw across sites, a small fee or an hour of help with something unrelated.

3. Observe first, ask second

Spend the first 30 minutes watching without interrupting. Note what they touch, which tools they open and what they write by hand. Then switch to apprentice mode: "Can you show me why you did that?" and "What happens if you skip this step?"

4. Log with the AEIOU frame

Record Activities, Environments, Interactions, Objects and Users, the AEIOU framework developed at the Doblin design firm. Time-stamp every workaround: a printed list taped to a screen, a second phone, a WhatsApp group used as a database. Photograph objects, never faces, unless you have explicit consent.

5. Write the say-do gaps the same day

After each visit, compare what the host told you in the opening chat with what you saw. "We always check stock before ordering" next to "ordered from memory twice in two hours" is a say-do gap. These gaps point at problems the customer cannot name in a survey.

6. Cluster across sites

After four visits, put every observation on a sticky note in Miro and group them. Count how many sites show each workaround. One site is an anecdote; four sites is a pattern.

7. Measure the cost of the top pattern

For the strongest pattern, estimate minutes lost per occurrence and occurrences per week, using your time stamps. Multiply by an hourly wage. That number goes into your decision memo and later into a problem impact quantification.

Template you can copy

FIELD NOTES: Site [#] | Date | Host role | Time window

Opening chat (5 min, before observing)
- How would you describe [task] in your own words?
- What is the hardest part?

Observation log
Time | Activity | Tool / object used | Workaround? Y/N | Quote | Photo #

AEIOU summary
Activities:    what steps, in what order, how long
Environments:  noise, space, interruptions, devices present
Interactions:  who they talk to, hand-offs, waiting
Objects:       paper, apps, spreadsheets, sticky notes, phones
Users:         everyone involved, including those who only watch

Say-do gaps
Said: ______________________  Did: ______________________

Top 3 observations + estimated minutes lost per occurrence
1. ____  2. ____  3. ____

Follow-up OK? Y/N   Referral to another site? ______

Worked example: stock control in small florists

This is an illustrative case with invented numbers. A founder thinks florists need better point-of-sale software. She visits 8 independent florists for three hours each on Friday mornings, when wedding and weekend orders pile up.

The till turns out to be fine. At 6 of 8 shops she sees the same workaround: the owner photographs the cold room on her phone, then scrolls the photo while taking phone orders to guess what stock remains. Each check takes about 2 minutes, and she counts 15 to 25 checks per morning. At 5 of 8 shops, an order is promised from flowers already reserved for another customer at least once during her visit.

In the opening chat, every owner said stock was "under control". The say-do gap sends her to test a stock-reservation tool for florists instead of a new till.

What should you measure, and when do you stop?

Stop when two visits in a row add nothing new to your clusters. If you reach 10 sites and patterns still scatter, your customer profile is too broad; split it and pick one segment.

Which mistakes should you avoid?

• Helping instead of watching. The moment you fix their printer, you change the work you came to observe.
• Visiting on quiet days. Calm hours hide the problem. Book the peak.
• Recording without clear consent. Ask for written permission, store files securely and delete them on the date you promised.
• Writing interpretations as facts. Write "she checked the photo 4 times in 10 minutes", and keep "she seems stressed" in a separate column.
• Stopping at two sites. Two visits produce stories. Six produce evidence.

How can AI speed this up?

Dictate voice notes in the car after each visit and use an AI transcription tool to turn them into text. Then paste all field notes into an AI assistant and ask it to list every workaround with the site numbers where it appeared. AI is useful for counting and sorting. It cannot tell which workaround made the host sigh, so keep your own ranking of emotional weight next to its count. Remove names and identifying details before you paste anything into an AI tool.

Where does this SOP fit in your validation plan?

Ethnography sits early in Customer Discovery, next to interviews.

1. Use customer interviews to find hosts and form a hypothesis.
2. Run these site visits to see the real workflow.
3. Turn your notes into a customer journey map and an empathy map.
4. If your product is digital, compare what you saw with the online data from digital behavior tracking.

All 15 methods are in the Customer Discovery category. The Fe/male Switch startup game gives first-time founders field tasks like this with deadlines, and the F/MS startup course helps you build a startup in four weeks on top of the evidence.

FAQ

How many observations do I need for ethnographic research?

Plan 6 to 10 site visits of two to four hours each. Stop earlier if two visits in a row add no new patterns.

Can I do ethnographic research remotely?

Yes, for screen-based work. Ask the host to share their screen during a real work session and narrate what they do. For physical work, a remote session misses too much.

What is the difference between ethnography and a customer interview?

An interview captures what people remember and choose to tell you. Observation captures what they do, including the habits they no longer notice.

Do I need to pay participants?

A fee helps for long visits, between $50 and $150 for a half day in most consumer and small-business settings. For B2B, a summary of patterns across sites works as a thank-you.