1. Courses
  2. Product Management
  3. Product Discovery
  4. Stop 2

35 minBeginner

Watching the work

What people tell you about their work and what they actually do are different, and most product opportunities hide in that difference. This lesson shows you how to visit a store or a partner shop, watch real work without getting in the way, spot the workarounds, and turn what you saw into a journey map and a service blueprint.

By the end you can

  • Plan and run a field visit at a store, a warehouse dock or a partner shop
  • Use the apprentice stance to see the steps people forget to mention
  • Spot workarounds such as side spreadsheets, notebooks and chat groups, and name the need behind each one
  • Put a cost on a workaround in minutes and errors across all your sites
  • Draw a journey map and a service blueprint with the workarounds marked

The everyday picture

Ask a good home cook for her recipe and she says, "Some onions, a little salt, cook until it's done." Stand in her kitchen and you see three onions sliced very thin, two big pinches of salt, a squeeze of lime she forgot to mention, and the pan pulled off the heat the moment the smell changes. She didn't lie. She has made the dish so often that the important parts became invisible to her.

Work is the same. A receiving clerk will tell you the official steps. Watch her and you see the notebook in her apron pocket, the walk to the front of the store to find Wi-Fi, and the photo she posts to the store's chat group. Those are the parts you need.

Why you have to watch

Interviews (User interviews that don't lie) capture stories. Watching captures what people no longer notice. Three things mostly show up only when you watch:

  • Tacit steps. is the know-how people use without thinking: which carton to open first, which barcode never scans, which colleague to ask.
  • The environment. Noise, glare on the screen, gloves, a queue at the till, no signal at the back of the store, a customer interrupting every two minutes.
  • Workarounds people think are normal. The notebook, the spreadsheet, the chat group. Nobody mentions them because they've stopped seeing them as unusual.

The classic method is , developed by Hugh Beyer and Karen Holtzblatt and described in their book Contextual Design (1998). You go where the work happens and learn the job the way an apprentice learns from a master: they work, you watch, and you ask about what you see as it happens. Nielsen Norman Group sums up its four principles as context (go where the work is), partnership (they steer as much as you do), interpretation (check what you think you saw with them) and focus (know what you are there to learn).

Watching also beats predicting at a much larger scale. When Safaricom and Vodafone piloted M-Pesa in Kenya from October 2005, the service was designed for microfinance customers to receive and repay loans. Pilot customers adopted it for other uses, and the service was refocused on sending money home and making payments before its full launch in March 2007. The famous product came from watching what pilot users actually did with it.

Quick check

A clerk tells you receiving a delivery takes "about 20 minutes". You time a real delivery at 51 minutes. What is the most likely explanation?

Running a field visit on the shop floor

Before you go. Pick one workflow and one real occurrence of it: the morning DC delivery at a store, Ravi's Tuesday call with his sales rep, a rider's evening shift. Watch the real thing, never a demo of it. Get permission from the store manager, and tell the area manager. For a partner shop, the owner's yes is enough. Plan for safety and space: stand behind the line at the loading dock, never block the till, and wear what staff wear if the site requires it. Bring a field-note sheet, a watch for timing and a phone for photos, which you take only with permission and never of customers or card details. Go as a pair at most. More visitors turn the visit into an inspection.

While you are there.

  • Set the stance in one sentence: "Please work as you normally would. Treat me like a new hire who is shadowing you."
  • Watch first. Ask at natural pauses: "What did you just do there?" "Why that way?" "When did that last go wrong?"
  • Time every step, and note every screen, paper, device and person it touches.
  • Ask to see the artefacts: the notebook, the spreadsheet, the chat group. Look only at what they choose to show you.
  • Watch one person at a time. If a manager comes over to explain how it "should" work, thank her, offer to talk later, and keep watching the clerk.
  • If the visit turns into a complaint session, write the list down, thank them, promise to pass it on, and return to the work.
  • Accept a cup of tea, then let them get back to work. You are an apprentice, not a guest.
  • Keep it under about two hours, the limit Nielsen Norman Group recommends for field sessions. Book a second visit rather than staying all day.

Afterwards. Write up your notes the same day. Then play them back to the person you watched: "Here is what I think I saw. What did I get wrong?" That check, the interpretation principle, catches your misreadings before they reach a spec.

The same method works beyond stores. At a partner shop, sit in during the rep's call and count the interruptions. At a distribution centre, stand at the dock through one truck. With riders, ride along for an evening shift and watch what they do on the app between deliveries.

Quick check

You are shadowing a receiving clerk when the store manager walks over and starts explaining how receiving is supposed to work. What do you do?

Spotting workarounds

A is something people do to get the job done that the official tool or process doesn't support. Each one is a need your product misses, plus the user's own solution to it. That makes workarounds the most valuable thing you can find on a visit. Look for these signals:

  • Paper next to a screen: notebooks, printed lists, sticky notes on the scanner, a list on the back of a calendar.
  • A side spreadsheet that nobody officially owns and everyone depends on: "the expiry sheet", "Rina's file".
  • Chat groups doing system work: orders sent as photos, price changes posted as images, approvals given with a thumbs-up.
  • Retyping: the same data entered into two systems, or read off one screen and typed into another.
  • Phone calls to check something the system already shows, because nobody trusts the screen.
  • Personal tools at work: a personal phone's camera or calculator, a personal mobile wallet.
  • An unofficial expert: "Ask Ama, she knows how to fix that."
  • Steps in odd places: walking to the front of the store to get a signal so the app will sync.

Read a workaround like a spec. The columns of a side spreadsheet are the fields people need. The chat group shows who needs to know what, and how fast. The notebook shows what the clerk checks on Monday morning. Don't rush to ban workarounds either. They often do jobs you can't see yet. A store's chat group carries price changes, and it may also be where the area manager thanks staff for a good week. Remove it without replacing those jobs and people will rebuild it the next day.

Then put a cost on each one: minutes per occurrence, times occurrences per day, times the number of sites, plus the errors it causes. Twenty-one minutes of workarounds per delivery, at about one delivery a day across 320 stores, is roughly 112 staff hours every day. That number turns "staff find it annoying" into a case a director can act on (Business cases).

Quick check

Mangrove's order desk retypes about 1,500 phone orders a week from photos in the sales reps' chat group. Each takes about 3 minutes, and about 1 order in 12 later needs a correction. Roughly what does this cost each week?

Journey maps and service blueprints

A shows one person's steps over time in one scenario, with what they do, use, think and feel, and where it hurts. Draw it from what you observed and heard, never from imagination in a meeting room. Keep it to one kind of person and one scenario, such as "a clerk receiving the morning DC delivery". Mark how long each step took and which steps are workarounds.

Step Minutes (one delivery) Check seal, papers 4 Count 40 cartons 8 Scan items into GRN 18 Expiry dates in book no field in the app 12 Walk to front to sync no signal at the dock 6 Damage photo to chat manager checks there 3 51 minutes, 21 of them workarounds x 320 stores = about 112 staff hours a day
A journey map reduced to its timeline: one clerk receiving one 40-carton delivery. The highlighted bars, with a reason under the step name, are workarounds. They take 21 of the 51 minutes, which across 320 stores is about 112 staff hours a day.

Seeing behind the counter

A adds the people and systems behind a journey. G. Lynn Shostack introduced the idea in Harvard Business Review in 1984, and Nielsen Norman Group calls it "a part two to customer journey maps". It has lanes for the customer's actions, the frontstage (the staff and screens the customer deals with), the backstage (work the customer never sees) and support processes (systems and other teams). The lanes are separated by the .

For internal tools and B2B products the blueprint is often more useful than the journey map, because the pain sits backstage. Ravi sees a friendly phone call. Behind the line of visibility, the rep's notebook becomes a photo in a chat group, an order desk retypes it, and about one order in twelve picks up an error on the way. Every between lanes is a place where data is copied, delayed or lost. Circle every one.

Quick check

In a service blueprint of phone ordering, where does "the order desk retypes the rep's photo into the back office" belong?

Shop owner Sales rep line of visibility Order desk Systems Writes list on calendar Reads list aloud 2 lines arrive wrong Writes it in notebook Photo to chat group Retypes into back office Back office order DC picks and ships 6 pm call copy 2 copy 3 goods workaround (copy 1 is the notebook)
A service blueprint of one phone order. Ravi sees only the top lane and the rep's call. Below the line of visibility the order is copied three times, by hand, before the back office knows about it, and errors creep in at each copy.

Do it

Plan one field visit, take timed notes, cost the workarounds, and draw the blueprint.

  1. Write a one-page visit plan

    One workflow, one real occurrence, clear permission, a safety note and the questions you'll keep in mind.

    Markdown
    ## Field visit plan: morning delivery, Ama's store
Workflow: receiving the DC delivery in the Store app (GRN)
Why now: staff say receiving "takes 20 minutes"; GRN sync errors rose this month
When: Thursday, truck due 7:30 am; stay until shelves are filled (2 hours max)
Permission: Ama (store manager) agreed; area manager informed
Who we shadow: the clerk on that day's roster, not the store's best clerk
Safety: stand behind the yellow line at the dock; no photos of customers
Bring: field-note sheet, watch, phone (photos only with permission), water
Roles: Imran observes and asks; Tomás times steps and writes
Keep in mind:
- Where does the clerk leave the Store app, and why?
- Which paper, chat or spreadsheet touches this delivery?
- What happens to damaged or short items?
After: write up the same day; play it back to the clerk on Friday

  2. Take timed field notes

    One line per step, with the tool or artefact, whether it's a workaround, and the person's own words.

    Markdown
    | Time  | Step                  | Tool or artefact   | Workaround? | Quote or note                    |
|-------|-----------------------|--------------------|-------------|----------------------------------|
| 07:34 | Checks seal, papers   | Delivery note      |             |                                  |
| 07:38 | Counts 40 cartons     | Delivery note, pen |             | "The note is nearly always right" |
| 07:46 | Scans items into GRN  | Store app, scanner |             | Promo packs scan twice           |
| 08:04 | Writes expiry dates   | Notebook (photo 2) | Yes         | "The app has no place for this"  |
| 08:16 | Walks to front, syncs | Store Wi-Fi        | Yes         | No signal at the dock            |
| 08:22 | Photos damaged items  | Store chat group   | Yes         | "Ama checks the group first"     |
| 08:25 | Moves stock to shelf  | Trolley            |             | Older yogurt pushed to the back  |

  3. Put a cost on the workarounds

    Use your own wage figure; the point is the order of magnitude, stated with its assumptions.

    Output
    Workaround minutes per delivery: 12 (expiry notebook) + 6 (walk to sync) + 3 (chat photo) = 21
Deliveries: about 1 per store per day, 320 stores
21 min x 320 = 6,720 min a day = about 112 staff hours a day
Assume a staff hour costs about $3 (use your own number):
112 x $3 = about $336 a day, or about $120,000 a year
Not yet counted: expired stock written off because the notebook was missed

  4. Draw the service blueprint

    Start with the customer lane from your notes, add the frontstage and backstage lanes, then the systems. Mark workarounds and circle the handoffs.

    Markdown
    ## Service blueprint: partner shop phone order (one cycle)
Customer (Ravi):    writes low items on calendar -> reads list on 6 pm call
                    -> gets stock Thursday, 2 lines wrong -> phones rep
Frontstage (rep):   calls -> writes order in notebook -> promises a fix
--- line of visibility ---
Backstage (desk):   photo arrives in reps' chat group -> retypes into back office
                    -> logs the correction after delivery
Support (systems):  back office order -> DC picking -> delivery -> invoice
Workarounds:  calendar list, rep's notebook, chat photo, retyping
Handoffs:     rep -> chat group, chat group -> desk, desk -> back office
Cost:         about 75 desk hours a week; about 1 order in 12 corrected

  5. Play it back

    Show the blueprint or journey map to the person you watched and to one other person who does the same job. Ask what you got wrong and what happens on a bad day. Change the drawing, not their words.

Bring it to your team

Ask these before anyone draws a solution for frontline work.

  • When did someone on this team last watch this work happen, in person, at a real site?
  • Which workarounds did we see, and what job does each one do that our product doesn't?
  • What does each workaround cost in minutes and errors across all our sites?
  • Which backstage handoffs does this feature remove, add or move?
  • If we switch off the chat group or the spreadsheet, what will people lose?

Red flags

  • A journey map drawn in a meeting room without a single visit.
  • Only the best store or the friendliest shop visited.
  • A plan to "ban the spreadsheet" with nothing that replaces what it does.
  • A visit timed for a quiet demo instead of a real delivery or a real order.
  • Field notes full of complaints and no timed steps.

Practice

Find and cost the workarounds

Tomás shadowed a nightly stock count in one store. Each store counts one aisle every night after closing. From his notes, list each workaround, the need behind it, and its cost per night across all 320 stores.

Output
21:30  Store closes. Two staff print the count sheet from the back office (2 min)
21:32  Count aisle 4 on paper, about 180 items (35 min)
22:07  Recount 12 items where shelf and backstore differ (10 min)
22:17  One staffer types the paper counts into a spreadsheet on the office PC (25 min)
       "The Store app can't do counts."
22:42  Emails the spreadsheet to the area manager, who forwards it to stock control (2 min)
22:44  Posts "done" and a photo of the sheet in the store chat group (1 min)
       "So the area manager sees we finished tonight."
  • You found at least four workarounds: printing, typing into the spreadsheet, emailing, and the chat post
  • Each need is written as a need ("counts reach stock control the same night"), not as a feature
  • You treated the recount as part of the real job, or explained why you counted it
  • Your cost uses minutes × stores and states its assumptions
  • You noticed at least one risk the minutes don't capture, such as typing errors
Show a model answer

Workarounds and needs. Printing the sheet (2 min): "I need the count list in my hand while I walk the aisle." Typing paper counts into a spreadsheet (25 min): "I need to record counts once, where I count." Emailing it on (2 min): "Counts should reach stock control the same night without me." Chat photo (1 min): "My area manager needs to see that we finished." The recount is real work; the walk to the backstore might shrink if the scanner showed backstore stock, so note it as a question.

Cost. 2 + 25 + 2 + 1 = 30 minutes per store per night. 30 × 320 = 9,600 minutes, about 160 staff hours a night, mostly late at night. Not counted: typing errors in the spreadsheet, and counts lost when the email is missed.

Blueprint a wrong-delivery fix

When Ravi's delivery has wrong lines, this happens: he phones his rep; the rep writes it in his notebook and posts a message in the reps' chat group; the order desk finds the original order in the back office and raises a correction; the DC adds the missing items to Ravi's next delivery; finance adjusts the invoice a week later. Draw this as a text blueprint with four lanes, mark the line of visibility, the workarounds and the handoffs, and name the step where you would expect most delays.

  • Four lanes: customer, frontstage, backstage, support systems
  • The line of visibility sits between the rep and the order desk
  • The notebook and the chat message are marked as workarounds
  • At least three handoffs are named
  • You name a likely delay and say why, such as the invoice fix a week later
Show a model answer
Markdown
Customer (Ravi):   finds wrong lines -> phones rep -> waits -> gets items next delivery
                   -> checks invoice a week later
Frontstage (rep):  takes call -> writes in notebook (workaround) -> posts in chat (workaround)
--- line of visibility ---
Backstage (desk):  finds original order -> raises correction in back office
Support:           DC adds items to next pick -> finance adjusts invoice (weekly batch)
Handoffs: Ravi -> rep, rep -> chat group -> desk, desk -> DC, desk -> finance

The likely delay is the finance adjustment, which runs weekly, so Ravi pays a wrong invoice or argues about it. The chat message is the likeliest place for a correction to get lost, because it scrolls away among 40 reps' messages.

Check yourself

On visits you notice that every receiving clerk keeps a notebook of expiry dates. Head office wants to ban the notebooks to push adoption of the Store app. What do you recommend first?

Which finding most needs a service blueprint rather than a journey map?

You can visit only one store in person for a receiving study. Which plan teaches you the most?

Flashcards

Open my review

Key takeaways

  • People describe the official process. Watching shows the tacit steps, the environment and the workarounds.
  • Go to a real occurrence of the work, take the apprentice stance, time the steps and ask about what you see.
  • Each workaround is a missed need plus the user's own solution. Read it like a spec and cost it across all sites.
  • A journey map shows one person's steps over time; a service blueprint shows the backstage handoffs where internal tools usually hurt.
  • Never retire a workaround until the product does every job it did.

Further reading

Episode transcriptFifty-one minutes, not twenty · about 10 minutes · tap any line to play from there
  1. NadiaImran, how long does it take a clerk at Ama's store to receive the morning delivery?
  2. ImranShe told me about twenty minutes.
  3. NadiaAnd when Tomás timed it?
  4. ImranFifty-one. I didn't believe him until I saw his notes.
  5. NadiaNobody lied. Today's episode is about the gap between what people tell you and what they actually do, and why so many product ideas live in that gap.
  6. NadiaHere's the picture from the lesson. Ask a good home cook for her recipe. Some onions, a little salt, cook until it's done.
  7. ImranMy grandmother, word for word.
  8. NadiaNow stand in her kitchen. Three onions sliced very thin, two big pinches of salt, a squeeze of lime she never mentioned, and the pan off the heat the moment the smell changes.
  9. ImranShe didn't hide the lime. She just doesn't think about it anymore.
  10. NadiaThat's tacit knowledge: know-how people use without thinking, so they never mention it. A clerk's version is the notebook in her apron pocket.
  11. ImranSo interviews get stories, and watching gets the things people no longer notice.
  12. NadiaThree kinds of things. Tacit steps. The environment, like noise, glare, gloves and no signal at the back of the store. And workarounds people think are normal.
  13. ImranIs there a name for the method?
  14. NadiaContextual inquiry. Hugh Beyer and Karen Holtzblatt described it in their book Contextual Design, in nineteen ninety-eight. You go where the work happens and learn like an apprentice. They work, you watch, and you ask about what you see as it happens.
  15. ImranAn apprentice. I like that. It changes how you stand in the room.
  16. NadiaIt does. Here's a bigger example. When Safaricom and Vodafone piloted M-Pesa in Kenya from twenty oh five, it was designed for microfinance customers to receive and repay loans.
  17. ImranAnd people used it for something else.
  18. NadiaPilot customers adopted it for other uses, and the service was refocused on sending money home and making payments before its full launch in March twenty oh seven. Watching what people did shaped the product the world knows.
  19. ImranOkay, the practical part. I'm visiting a store. What do I do?
  20. NadiaPick one workflow and one real occurrence of it. The morning DC delivery, never a demo of it. Get permission from the store manager and tell the area manager. Stand behind the line at the dock. Photos only with permission, and never of customers.
  21. ImranAnd once I'm there?
  22. NadiaSet the stance in one sentence: please work as you normally would, and treat me like a new hire who's shadowing you. Watch first. Ask at natural pauses. What did you just do there? Why that way? When did that last go wrong?
  23. ImranAnd time everything.
  24. NadiaTime every step, and note every screen, paper, device and person it touches. Ask to see the artefacts: the notebook, the spreadsheet, the chat group. Only what they choose to show you.
  25. ImranWhat usually goes wrong on visits?
  26. NadiaThe visit turns into a complaint session, so write the list down, thank them, and go back to the work. Or it turns into a group explaining the system, so watch one person at a time. Or you become a guest. Accept the tea, then let them work.
  27. ImranAnd the manager who walks over to explain how receiving should work?
  28. NadiaThank her, offer to talk later, and keep watching the clerk. Her version is the official one, and you came for the real one. Keep the whole visit under about two hours.
  29. ImranNow workarounds. What am I looking for?
  30. NadiaPaper next to a screen. A side spreadsheet nobody officially owns. Chat groups doing system work, like orders sent as photos. Retyping. Phone calls to check something the system already shows.
  31. NadiaAlso personal phones used for work, an unofficial expert everyone asks, and steps in odd places, like walking to the front of the store to find Wi-Fi.
  32. ImranThat last one is exactly what Tomás saw.
  33. NadiaIf you can see the screen, the first diagram is that delivery as a timeline. Six steps, with a bar for the minutes each one took. Three bars are workarounds: expiry dates in a notebook, twelve minutes. Walking to the front to sync, six. A damage photo to the chat group, three.
  34. ImranTwenty-one of the fifty-one minutes.
  35. NadiaAnd here's why that matters. At about one delivery a day across three hundred and twenty stores, that's roughly a hundred and twelve staff hours every day.
  36. ImranThat's a number a director listens to.
  37. NadiaMinutes per occurrence, times occurrences per day, times sites, plus the errors. That turns staff find it annoying into a business case.
  38. ImranSo should we just ban the notebook?
  39. NadiaNot yet. Read a workaround like a spec. The notebook tells you which dates matter and when someone checks them. And workarounds often do jobs you can't see. A store chat group carries price changes, but it may also be where the area manager thanks people for a good week.
  40. ImranSo if we kill it without replacing those jobs, people rebuild it the next day.
  41. NadiaExactly. Replace every job a workaround does, then retire it.
  42. ImranOkay. Journey maps.
  43. NadiaA journey map shows one person's steps over time in one scenario, with what they do, use, think and feel. Draw it from what you saw, never from imagination in a meeting room. The timeline we just described is a stripped-down journey map.
  44. ImranAnd the service blueprint?
  45. NadiaThe blueprint adds what's behind the counter. G. Lynn Shostack introduced the idea in the Harvard Business Review in nineteen eighty-four. It has lanes for the customer's actions, the frontstage people and screens they deal with, the backstage work they never see, and the support systems.
  46. ImranSeparated by the line of visibility.
  47. NadiaRight. The second diagram is a blueprint of Ravi's phone order. Picture four horizontal lanes. At the top, Ravi writes his list on the calendar and reads it out on the six pm call. In the next lane, the rep writes it in a notebook and posts a photo to the reps' chat group.
  48. NadiaBelow the dashed line of visibility, the order desk retypes that photo into the back office. Then the DC picks and ships, and two lines arrive wrong.
  49. ImranThree copies by hand before the system even knows the order exists.
  50. NadiaAnd every handoff is a place where data gets copied, delayed or lost. For internal tools, that backstage lane is usually where the pain is.
  51. ImranThat's what I found at the order desk. Two staff retype the photos that about forty reps post, roughly fifteen hundred orders a week, most of them between seven and nine in the evening. Their correction log showed about one order in twelve fixed after delivery.
  52. NadiaAnd nobody on the Partner app team had ever seen that desk.
  53. ImranNobody. And it changed the project. Moving phone shops into the app doesn't only help shop owners. It frees about seventy-five hours of desk time a week and removes the biggest source of wrong deliveries.
  54. NadiaOne more tool, for when you can't visit every store. A diary study.
  55. ImranHow does that work with clerks?
  56. NadiaAsk eight clerks in four stores to send a photo and a thirty-second voice note every time they receive a delivery, for a week, through the messaging app they already use. Cheap, more sites, more shifts. Pair it with at least one visit, because people don't report what they don't notice.
  57. ImranThe Do it, in my words. Write a one-page visit plan with one workflow, permission and a safety note. Take timed field notes, one line per step, marking the workarounds. Put a cost on the workarounds with your own wage number.
  58. ImranThen draw the blueprint with four lanes and circle the handoffs. And play it back to the clerk, asking what you got wrong.
  59. NadiaThat last step matters most. It catches your misreadings before they reach a spec.
  60. ImranQuestions for my next review. When did someone on this team last watch this work, in person, at a real site? What job does each workaround do that our product doesn't? And if we switch off the chat group, what will people lose?
  61. NadiaTake a second on this one. You can visit only one store. Do you pick the best store, the flagship near head office, or an ordinary store on a normal morning?
  62. NadiaAn ordinary store on a normal morning. The best store hides the workarounds the others need, and the flagship shows you the official version. Add a diary study to reach more stores.
  63. ImranMy takeaways. People describe the official process, and watching shows the tacit steps, the environment and the workarounds. Each workaround is a missed need plus the user's own solution, so cost it across every site. And the blueprint shows the backstage handoffs where internal tools usually hurt.
  64. NadiaOne nuance. A single visit can be an unusual day. Check your timings against a second visit or the diary study before you build a case on them.
  65. ImranNoted.
  66. NadiaNext time, what to do with the pile. Eight interviews, two visits, two hundred tickets, and an AI tool that's very eager to help.
Finished reading?Mark it complete to fill in its stop on your course line.

Use it at work

Process map: as-is and to-be

A swimlane map with lanes per team, decisions with labelled outcomes and a validator; the today and the proposed process side by side.

Open in the Product & BA Lab
Next stopFrom research to opportunities
Was this lesson helpful?
Questions and discussion

Loading the discussion…