Feedback Loupe · Embedded engagements
One hard product problem. Six to twelve weeks. A principal designer embedded with your team.
You end with a validated direction, a working prototype, and a decision you can defend.
The engagement
Evidence over artifacts, inside your team, on a clock.
One problem, one designer embedded with your engineers and your users. Long enough to put working behavior in front of the people who will depend on it; short enough that the answer arrives while the decision is still open.
What you get is a decision you can defend: a validated direction and a functional prototype, or a recommendation not to build at all. I have made that second recommendation more than once. It is a much cheaper answer in week nine than in month nine.
The method is documented on Approach and visible across the work. This page is its commercial shape.
The problems
Products get stuck three ways. I’ve spent a career inside all three.
01 · AI you can trust
Accuracy is not the problem; confidence is. The work is making uncertainty visible, sources legible, and the workflow real, until the people using the output trust it enough to act on it. On a clinical voice AI, field research in week one killed the co-pilot concept and moved the product upstream. That judgment is the job.
Healthcare voice agent → review, escalation, and safe handoff model
02 · Operational workflows
Complex systems drift toward interfaces that explain the database instead of the work: internal logic masquerading as user experience. The job is reducing complexity without hiding what matters. The discipline is the same whether the surface is a booking flow serving 31 million passengers a year or a security analyst’s investigation floor.
CLI security tooling → an investigation product analysts navigate without the engineer
03 · Zero to one
When the capability is ahead of the product, opinions multiply and the loudest one wins. Evidence settles it: direct time with users, prototypes that test behavior instead of taste, and a stage gate that is allowed to say stop. Most concepts don’t survive that. The ones that do are worth building.
Care-coordination research → onboarding beta, and a defensible stop at the market ceiling
How it runs
Four moves. The same discipline every time.
Kickoff
Name the problem, the users, and what done means, including what will not get built.
Access
Time with the people who will use the thing, inside the system they work in. Not a stakeholder tour.
Evidence
Working behavior in front of real users. Assumptions break here, early, while breaking is cheap.
The call
You leave knowing whether to proceed, pivot, stop, or hand off. The evidence and the working prototype come with it, close enough to build from.
The engagement needs three things from your side: one owner who can change the decision, users I can reach, and a decision that is genuinely open. Direct access to users is non-negotiable. It is where the evidence comes from.
Fit
The shape protects the outcome.
A timebox that ends in a decision only works if it stays what it is. So the engagement is deliberately not:
If what you need is one of those, I will say so in the first conversation and point you somewhere better suited.
You don’t need a brief. A sentence will do.
Say which of the three problems is yours, what’s stuck, and whether the people who use the thing are reachable. If the shape fits, the first conversation is thirty minutes.
Feedback Loupe LLC · Virginia / remote · led by Christopher Ryan Mullins. Prior engagements were performed as an employee or subcontractor of other firms, shown here as evidence of the principal’s capability.