Standard discovery advice assumes you can reach your users. In institutional financial products you frequently cannot. The people who use the system are a buy-side operations team at a client firm, access is controlled by a relationship manager who does not want you near their account, and the users themselves bill their time in a way that makes an hour of research genuinely expensive.
This does not mean discovery is impossible. It means the cheap methods are unavailable and you have to work the ones that remain much harder.
Start with what already exists
Before asking anyone for access, mine what your firm has already collected. In institutional businesses this is usually far richer than in consumer products, and almost nobody reads it.
| Source | What it tells you | Its bias |
|---|---|---|
| Support and service tickets | Where the product actually breaks | Skews to complainers and to the loudest accounts |
| Implementation notes | What every client asks to be changed at onboarding | Reflects sales promises as much as user need |
| RFP responses | What buyers say they require | Buyer is often not the user |
| Usage logs | What is really used, and what was built and abandoned | Shows behaviour, never motive |
| Churn and non-renewal notes | The highest-signal material in the business | Written by the person who lost the account |
A week spent reading implementation notes across ten clients will tell you more than three badly-recruited interviews, and it costs nobody an hour of their time. It also earns you the access you were refused: showing up to a relationship manager with a specific, evidence-backed hypothesis is a different conversation to asking for "some user research".
Go through the relationship manager, not around them
The RM's incentive is protecting the account. An unstructured product conversation is pure downside for them: it can surface complaints they were managing, or raise expectations they cannot meet.
Make it safe and specific:
- Ask for one named person and 30 minutes, not "some users".
- Send the questions in advance, so the RM can see nothing embarrassing is coming.
- Invite them to join. Their presence changes the answers, but access beats purity.
- Report back what you learned and what you will do. Do this reliably and the second request is far easier.
Your largest, unhappiest client is simultaneously the richest source of insight and the one you will be kept furthest from. Expect this. Build the internal trust to get there eventually rather than forcing it early and getting locked out permanently.
Use your internal proxies properly
Operations, client services and sales engineers all use your product against real client problems every day. They are not your users, and treating them as a substitute produces products optimised for internal convenience. But as a source of hypotheses they are excellent and available immediately.
The distinction that keeps this honest: internal colleagues tell you where to look; clients tell you whether you are right. Skip the second step and you will confidently build the wrong thing with a great deal of internal support.
Watch the workaround
In institutional finance the workaround is almost always a spreadsheet, and it is the most informative artefact available to you. If a client operations team exports your data every morning and rebuilds something in Excel, that spreadsheet is a specification for the product you failed to ship.
Ask for a copy. Then ask:
- Which columns did they add that you do not provide?
- What did they filter out, and why is it noise to them?
- What do they do with the output — who receives it, and when?
- What happens on the day it is wrong?
That last question usually reveals the real job. The spreadsheet is rarely the point; it is the thing standing between an operations lead and an uncomfortable conversation with a portfolio manager at 8am.
Industry events put your users in a room without an RM gatekeeping the conversation, and in a mood to complain freely to a peer. One afternoon at a trade association working group can be worth a quarter of formal research requests.
Adjust your expectations on sample size
Consumer discovery guidance says five to eight interviews per segment. In institutional products your entire market may be forty firms, and the segments are genuinely different — a tier-one bank and a boutique asset manager do not share a workflow.
With small populations the useful shift is from statistical thinking to case thinking. Three deep, longitudinal relationships where you understand a firm's workflow properly will serve you better than fifteen shallow conversations. Depth substitutes for breadth when the population is small.
The technique fundamentals still hold — see customer interviews that produce decisions, and note that the past-behaviour rule matters even more here, because institutional users are unusually fluent at describing an idealised process that nobody actually follows.