0%
Forward Deployed Engineering
The Role
Customer Data and Systems
Building in the Field
Running Deployments
Discovery and Problem Framing
Every deployment begins with a request, and requests are rarely the real problem. A customer may ask for a dashboard when they need faster decisions, or for a model when they need cleaner data. Discovery is the work of finding out what the customer actually needs before building anything. This lesson covers why discovery matters, how to run it, how to watch users work, how to frame a problem and how to test that framing.
The request is not the problem
Customers describe problems in terms of solutions they can imagine. A logistics manager might say we need a real time map of every truck, when the underlying problem is that late deliveries are discovered too late to warn customers. A map might help, but an alert when a delivery is likely to be late might help far more, at a fraction of the effort.
The cost of skipping discovery
When FDEs skip discovery, they often build exactly what was requested and then watch nobody use it. The dashboard is accurate but answers the wrong question, or the model is clever but its output does not fit any decision. Weeks of work are lost, trust drops and the customer starts to doubt the product. A few days of discovery would usually have prevented this.
Discovery as a habit, not a phase
Discovery is often described as the first phase of a deployment, but good FDEs treat it as a habit that continues throughout. Each demo, each user conversation and each data surprise is a chance to learn more. Plans change as understanding improves. The first discovery effort sets direction, and later discovery keeps it correct.
Who discovery involves
Discovery involves many people: the sponsor who funds the project, managers who own the process, users who do the daily work, data owners who control the systems and IT and security teams who set the rules. Each sees a different part of the problem. An FDE who talks only to the sponsor gets a strategic picture with little detail about how work actually happens.
What discovery produces
Discovery should produce a clear written problem statement, a description of current workflows, a list of data sources and their quality, the decisions the system must support, success criteria, constraints and risks. This does not need to be a long document. A few pages that everyone agrees on are far more useful than a long report nobody reads.
How long discovery takes
For a typical deployment, initial discovery takes from a few days to two or three weeks. Small, well defined problems need less. Large, vague goals such as improve operations across all plants need more. Discovery that drags on for months is usually a sign of unclear sponsorship or a customer unable to decide, which is itself an important finding.
Discovery under time pressure
Customers sometimes push to skip discovery and start building immediately, especially after a long sales process. FDEs can respect this pressure while still doing discovery quickly: a few focused interviews, an afternoon watching users and a quick look at the data. Building something small early can even be part of discovery, as long as everyone treats it as a learning tool.
Discovery and trust
Discovery builds trust. When an FDE asks thoughtful questions, listens carefully and reflects back what they heard, customers feel understood. That trust pays off later, when the FDE needs patience during a difficult integration or a sponsor to defend the project internally. Customers forgive many technical setbacks when they believe the FDE truly understands their problem.
Reading before the first meeting
Before discovery meetings, read everything available: the sales notes, the contract, the customer website, annual reports and any public information about their industry. Arriving informed lets you skip basic questions and ask better ones. It also shows respect for the time of the people you interview, who may have explained their business to vendors many times before.
Separating facts from opinions
During discovery, people mix facts and opinions. A manager might say the data is terrible, which is an opinion, while a data owner might say twelve percent of records lack a customer id, which is a fact. Both are useful, but they need different handling. Record opinions as hypotheses to check, and verify important claims against the data yourself.
Discovering politics
Discovery also reveals politics: which teams compete for budget, who opposed the purchase, whose role might change because of the system. These dynamics affect the deployment as much as technical issues. FDEs do not need to play politics, but they need to understand them, so they can involve the right people and avoid surprises.
When the customer already knows
Sometimes the customer has done careful work already and knows exactly what they need. In that case, discovery is shorter: confirm the understanding, check the data and agree success criteria. Respect their work. Insisting on a long discovery process when the customer is well prepared can feel patronising and waste time on both sides.
Discovery documents as shared memory
Write up discovery findings and share them with the customer. A written record prevents later disagreements about what was agreed and helps new people join the project. It also serves your own team, since the next FDE on the account can read it rather than starting again. Keep it updated as understanding changes.
Hypotheses before the first meeting
After reading the background material, write down a few hypotheses about the problem: what you think is happening, why and what might help. Then use discovery to test them. Hypotheses focus your questions and make it easy to notice when you are wrong. Hold them loosely, though, and drop any that the evidence does not support.
The beginner mindset
Even experienced FDEs benefit from a beginner mindset in discovery. Assume you do not yet understand how this customer works, even if you have deployed in the same industry many times. Every organisation has its own history, systems and habits. Patterns from past deployments help you ask better questions, but they should not replace the questions themselves.
Discovery with your own team
Discovery also involves your own company. Talk to the sales team about the deal, to support about any past tickets, to the product team about relevant features and to other FDEs who have deployed in the same industry. Internal discovery takes a few hours and often saves days, because someone has usually seen a similar problem before.
A short example
An FDE joins a utility company that asked for a predictive maintenance model. Discovery reveals that the real issue is crews arriving at sites without the right parts. The first deployment becomes a simple parts checklist driven by work order data, delivered in three weeks. Crews love it, and the maintenance model follows later, with better data and a sponsor who trusts the team.
Discovery as a skill you can improve
Discovery is a skill, and like coding it improves with deliberate practice. After each deployment, look back at what discovery missed and why. Perhaps you never spoke to night shift users or never checked a key data source early. Turning each miss into a new question on your checklist makes the next discovery round sharper and faster.
Customers describe solutions they can imagine, so discovery uncovers the problem behind the request before anyone builds. Treat it as a continuing habit, involve sponsors, managers, users, data owners and security, separate facts from opinions and write the findings down so everyone shares one understanding.