The Forward Deployed Iteration Loop

The Forward Deployed Engineer (FDE) model is designed to solve high-stakes, "messy" real-world problems by embedding builders directly with users. Instead of traditional "customer interviews," the goal is to live the customer's problems and execute 4-5 rapid iteration cycles per week to build software that achieves specific business outcomes (e.g., increasing factory production or finding fraud).

The Weekly FDE Rhythm

Follow this specific schedule to maximize trust and iteration speed:

1. Monday (On-Site Integration): Arrive at the customer's office. Get a desk inside their building. Attend their internal meetings. Identify the "burning problem" that the CEO or COO cares about most.
2. Monday Night (The Build): Stay late to build a prototype or feature that addresses a specific pain point discovered during the day.
3. Tuesday (The Feedback Loop): Show the build to the actual users (e.g., factory floor managers or analysts). Observe them using it.
4. Tuesday Night (The Iteration): Refine the code based on the day's feedback.
5. Wednesday - Thursday: Repeat the Cycle. By Thursday afternoon, you should have executed 4 full iterations before leaving the site.
6. Friday (Core Synthesis): Return to the home office. Synthesize what you built into a generalizable product feature.

The Murder Board Ritual

Before starting or pivoting a major project, conduct a "Murder Board" to stress-test the strategy:

1. Prepare the Doc: Write a 2-page plan including the vision, goals, and specific tactics for the next three months.
2. Define Principles: Include 2-3 principles that are disagreeable. (Avoid "Move Fast"; use something like "Prioritize data ingestion over UI polish").
3. Invite the Killers: Bring in 3-4 smart peers who have no context on the project.
4. The Session: Their sole job is to "tear apart" the plan, questioning every assumption, frame, and tactic.
5. Outcome: Refine the plan based on the holes they poked, or abandon the project if the core logic fails.

Mapping the "Ontology" (Alien to Human)

Large organizations often store data in "alien language" (e.g., SAP table S3_F1_Z). Your job is to translate these into a human-legible "Ontology":

1. Identify the Core Concepts: What are the 3-5 things the business actually talks about? (e.g., "Aircraft," "Part," "Work Order," "Oil Well").
2. Extract and Join: Pull the raw, messy tables from the client's disparate systems.
3. Map to Humans: Create a layer where the user can query "Where is Aircraft 79?" instead of joining 15 tables with cryptic names.
4. Build UI on Top: Only build UI once the data mapping (the "Ontology") reflects the real-world operation.

Examples

Example 1: Manufacturing Production

• Context: A plane manufacturer needs to ramp up production from 4 to 16 planes per month.
• Application: The FDE embeds in the factory, identifies that "work order carryover" is the bottleneck. They map the SAP data for "parts" and "stations" into a unified dashboard.
• Output: A "Project Management" tool (similar to Asana for planes) that allows station managers to see exactly what work is remaining and where the parts are physically located.

Example 2: Public Health Response

• Context: A government agency needs to coordinate vaccine distribution during a pandemic.
• Application: The FDE sits with civil servants and clinicians. They build a data adapter that pulls real-time inventory from thousands of pharmacies into a single "Source of Truth."
• Output: A supply chain visibility platform that allows the agency to reallocate doses based on daily demand spikes.

Common Pitfalls

• Acting as a "Solutions Architect": FDEs must be empowered to write code and ship new products. If the person is only "listening" and "architecting" without building, the loop breaks.
• Solving the Wrong Problem: Customers often ask for "better data infrastructure." Usually, they actually need a specific business outcome (e.g., "lower wait times"). Always anchor to the outcome, not the tech stack.
• Remote Disconnection: Trust is built over dinners and shared desks. Attempting this model via Zoom loses the "vibe check" and the ability to see the non-verbal frustrations of the users.
• Generic Principles: Using principles like "Move Fast" during a Murder Board. If no one can reasonably disagree with your principle, it isn't a principle; it's a platitude.