Frontline Operator Feedback Loop
Use the product in the least privileged role, log friction, then fix it
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 99%
Shu continues making Deliveroo deliveries to experience the system as a rider rather than as a recognized executive. A field session exercises several layers at once: the rider app, travel and waiting, restaurant handoffs, food condition, and treatment by partner staff. He avoids disclosing his identity because special treatment would corrupt the observation. When friction occurs, he records it in notes, compares it with what other riders say, and takes the issue back to the company or restaurant for process improvement. The loop is use, observe, log, verify, and act—then repeat. Its value comes from exposing operational details and preserving empathy for the people doing the work. It should complement, not replace, larger datasets and structured feedback from a representative range of riders, restaurants, and consumers.
Origin
Shu was one of Deliveroo's first riders and says an early experience of being treated dismissively helped shape his concern for rider respect. Years later, as chief executive, he still makes deliveries to test the app and observe restaurant operations without revealing who he is.
Core principles
- 01Direct use reveals friction that reports can hide
- 02An unannounced operator sees the ordinary experience rather than executive treatment
- 03Frontline participants are customers of the operating model too
- 04Observation should become recorded evidence
- 05Field learning matters only when it changes product or process
How to run it
- 1
Select the frontline journey
Choose a real operating role and transaction that together expose the product, partner process, and human experience.
Pro tip Pick a journey with enough complexity to reveal handoff problems, not merely a demonstration path.
Watch out Follow the same safety and operating requirements as anyone else in the role.
- 2
Remove executive privilege
Use the normal tools and channels without announcing seniority or asking for exceptional handling.
Pro tip Preserve anonymity where practical so partners respond as they normally would.
Watch out Do not deceive people beyond what is necessary to experience the standard service.
- 3
Run the full experience
Complete real transactions and observe software behavior, waiting, handoffs, communication, and the condition of the final output.
Pro tip Notice both technical defects and small interpersonal behaviors that affect the day.
Watch out One unusual interaction should not automatically be treated as a system-wide pattern.
- 4
Log specific evidence
Record where the friction occurred, what happened, and which part of the process may have contributed.
Pro tip Write notes immediately instead of reconstructing the encounter later from memory.
Watch out Describe observed conduct without inflating motive or certainty.
- 5
Check the pattern
Compare the observation with input from frontline participants and available operational data.
Pro tip Ask whether others recognize the issue before choosing the size of the response.
Watch out The executive's personal experience must not overrule broader worker evidence.
- 6
Close the loop
Send validated findings to the relevant internal team or partner, improve the process, and revisit the journey to see whether conditions change.
Pro tip Frame partner feedback around waiting, food readiness, and respectful communication rather than personal status.
Watch out Fieldwork without follow-through becomes executive theatre.
In the wild
During deliveries the night before the interview, Shu says one restaurant kept him waiting, handed over food he considered cold, and responded dismissively when he raised it. He did not reveal his identity, recorded the incident, and heard another rider describe the restaurant as a recurring problem.
→ Shu said he would speak with the restaurant about its process and how staff treated riders.
Common mistakes
Requesting the executive experience
Revealing seniority can produce special treatment and hide the ordinary journey the fieldwork is meant to expose.
Generalizing from one encounter
A single session supplies a lead; other participants and operational evidence are needed to judge its prevalence.
Taking notes without action
The loop is incomplete until a validated observation reaches the team or partner able to improve the system.
Is it for you?
Best for
It is best for leaders of operational products where software quality and human service interactions jointly determine the outcome.
Not ideal for
It is not ideal as a substitute for representative research, worker voice, safety controls, or systematic performance data.
From the transcript
“i always test our rider app right so it's a good way to do it”
“i want the true experience right”
“i make sure to log log into my notes”
From the episode
Deliveroo Founder - From £0 to £5 Billion: Will Shu