0% completed
The Customer Engagement Lifecycle
On This Page
The Seven Stages
Discovery
Scoping
Prototype
Pilot
Production
Handoff
Feedback
Outputs and Exit Conditions
Exposure at Each Stage
When a Stage Fails
Using This in an Interview
Key Takeaways
Practice Questions
Most FDE engagements follow the same order of work, called the engagement lifecycle. Its seven stages are discovery, scoping, prototype, pilot, production, handoff, and feedback. Each stage removes one kind of uncertainty. The insurer engagement from What a Forward Deployed Engineer Does went through all seven in six weeks.
The Seven Stages
Every stage produces one output, and the next stage starts only when that output is good enough. The condition that ends a stage is its exit condition. Writing it down in advance keeps the customer and the FDE in agreement.
The stages do not take equal time. For the insurer, discovery and scoping took week 1. The prototype and the pilot took two weeks each, and production and handoff shared week 6.
Discovery
Discovery is the stage in which the FDE learns how the work is done today, who makes each decision, and what data exists. The FDE read 300 claims with the adjusters and found three facts the demo had missed. Photos decide most glass and minor damage claims. Scanned police reports decide most theft claims. Some fraud conditions had never been written down.
The output is a short description of the current work. The exit condition is that the FDE can explain how an adjuster chooses each of the six queues.
Scoping
Scoping is the stage in which the FDE and the customer agree on the smallest version that delivers value, and on the number that will show it works. For the insurer, confident claims would be routed automatically. The pilot would cover glass and minor damage first. Success would be measured as agreement with the adjusters on past claims.
The output is a one-page scope that also lists what is excluded, like payment decisions. The exit condition is that the claims manager accepts it.
Prototype
A prototype is a first working version, tested on past data with known answers before any real work depends on it. On 2,000 past claims, the prototype routed 1,400 automatically and matched the adjusters on 97 percent of those, so about 42 were wrong. The error rate is now known before a real claim is routed. The exit condition is that the customer accepts it for the pilot.
Pilot
A pilot runs the system on real work, limited to one part of it, while people check the results. The insurer's pilot covered glass and minor damage, about 600 claims a day, with adjusters checking a sample. The pilot checks whether agreement on real claims stays close to the prototype's 97 percent.
Production
Production means the system handles all of the work it was built for, and the customer's operations depend on it. In week 6, the system began routing all six queues, about 1,200 claims a day. Production also needs counts and alerts, so the customer's team notices when automatic routing falls.
Handoff
Handoff is the stage in which the customer's own team takes over operating and changing the system. It succeeds when that team, here the insurer's 4 engineers, can make a change without the FDE. A good handoff has three parts. The team reads the code, a short guide covers each alert, and the team makes one change while the FDE is still there.
Feedback
Feedback is the stage in which the FDE reports to the product team what the product could not do. The insurer engagement produced three reports: scanned reports read poorly, there was no decision log, and only one threshold was allowed. Feedback is a stage of its own because, without it, only one customer benefits from what the engagement learned.
Outputs and Exit Conditions
| Stage | Output | Exit condition for the insurer |
|---|---|---|
| Discovery | A description of the current work and its data | The FDE can explain how each of the six queues is chosen |
| Scoping | A one-page scope with a success measure | The claims manager accepts it |
| Prototype | A system tested on 2,000 past claims | The insurer accepts 70 percent automatic routing at 97 percent agreement |
| Pilot | The system running on two queues | Agreement on real claims stays close to the prototype's |
| Production | The system running on all six queues | Counts and alerts show normal operation |
| Handoff | A team that can change the system | The 4 engineers make a change without the FDE |
| Feedback | Reports to the product team | Each gap is written up with the evidence that supports it |
Exposure at Each Stage
Each stage lets the system affect more real work than the stage before it. The amount of real work the system can affect is its exposure. For the insurer, it was zero real claims in the prototype, about 600 a day in the pilot, and 1,200 in production.
The arithmetic shows why the order matters. In production, 840 automatic decisions a day at 97 percent agreement means about 25 wrong ones. The pilot's 420 give about 13. The team measures real error at half the exposure.
When a Stage Fails
When a stage fails its exit condition, the engagement returns to the stage before it. If the pilot had shown lower agreement, the team would have returned to the prototype and added the new real claims to the test data. The insurer's own demo skipped discovery and misrouted 1 claim in 4.
Using This in an Interview
In a project deep dive, describe your project stage by stage, with the exit condition you used at each step. This shows that you controlled risk.
"We spent the first week on discovery, reading 300 claims with the adjusters. Then we tested a prototype on 2,000 past claims, where 70 percent were routed automatically with 97 percent agreement. The claims manager accepted that rate, so we piloted on the two largest queues, about 600 claims a day, with adjusters checking a sample."
Expect two follow-up questions:
- "How did you decide the prototype was good enough to pilot?" Name the exit condition and who accepted it, here the claims manager accepting 97 percent agreement.
- "What would you have done if the pilot results had been worse?" Return to the prototype, add the failing real claims to the test data, and pilot again.
Key Takeaways
- The engagement lifecycle has seven stages: discovery, scoping, prototype, pilot, production, handoff, and feedback.
- Each stage produces one output and ends when its written exit condition is met.
- Exposure increases one stage at a time, from no real claims to about 600 a day to all 1,200.
- A failed stage sends the engagement back one stage.
The next lesson, How FDE Interview Loops Test the Job, shows which interview rounds test which parts of this work.
Practice Questions
1. Why was the prototype tested on past claims instead of new claims?
<details><summary>Show answer</summary>Because past claims already had a known correct queue, so every decision could be scored. New claims have no known answer until an adjuster routes them.
</details>2. During the pilot, the system routes 420 of the 600 daily claims automatically, and agreement on real claims falls to 94 percent. How many claims a day reach the wrong queue, and what should the team do?
<details><summary>Show answer</summary>About 25 claims a day, and the team should return to the prototype. Six percent of 420 is 25.2, about twice the 13 a day expected at 97 percent. The exit condition has failed, so the team adds the misrouted claims to the test data and improves the prototype before piloting again.
</details>3. The claims manager asks to skip the pilot and enable all six queues at once to save two weeks. What do you say?
<details><summary>Show answer</summary>Explain the exposure and offer a measured path. Without a pilot, about 25 claims a day could reach the wrong queue before anyone measured real error. The pilot covers the simplest half of the claims and still delivers about half of the value during those two weeks.
</details>Reading Progress
0%
On This Page
The Seven Stages
Discovery
Scoping
Prototype
Pilot
Production
Handoff
Feedback
Outputs and Exit Conditions
Exposure at Each Stage
When a Stage Fails
Using This in an Interview
Key Takeaways
Practice Questions