A BEMER distributor in Hungary spent years paying developer agencies to build a rental management system. He never liked any of the results. A few months ago he picked up a Claude subscription and got further in a week than every paid team had managed combined.
That gap between what founders build themselves and what agencies deliver is not a headline. It is our family network's lived experience.
We know both sides. CZ Dev is a two-person agency. We also built our first product for a non-technical founder, our mother: ten years of operational data trapped in spreadsheets she understood and nobody else could touch. Six weeks to a working Streamlit prototype. Five months to a full React production build. Month-end reconciliation that took a full day now runs in minutes.
The founders who go through bad agency engagements are not unlucky. They walk into a specific, predictable failure mode that starts in week one.
The Real Failure Mode
The six-month silence then wrong-delivery pattern is not a talent problem. It is a discovery process that never defined done.
Most agencies treat discovery as a formality. They run one kickoff call, collect requirements in broad strokes, and issue a quote before they understand the actual data structure, the existing workflows, or the edge cases the client has been managing for years without documentation. The client assumes the agency will figure it out. The agency assumes the client knows what they want. Neither assumption is stated out loud. Six months later, both are wrong.
The BEMER distributor's outcome was not exceptional. He spent millions of forint paying developers over the years to build the same product and never liked any of the results. The pattern held every time because the pattern was structural, not personal.
Three Red Flags in the First Week
These three signals, visible before any code is written, predict a bad engagement with high reliability.
No discovery session that defines done in measurable terms. If an agency cannot tell you, in specific language, what you will be able to do at the end of phase one that you cannot do today, they have not done discovery. They have done intake.
A quote issued before the agency understands your data structure or existing workflows. Pricing a build without understanding what the data looks like is guessing. Guesses compound. The quote either inflates to protect the agency or shrinks to win the deal, and both outcomes punish the client.
No named person responsible for decisions on both sides. Agencies fail in ambiguity. If there is no single human on the client side authorized to say "yes, that is what we want" and no single human on the agency side accountable for delivery, every disagreement becomes a negotiation with no anchor.
Questions to Ask in Week One
These are not gotcha questions. They are the questions that reveal whether the agency has done this before in a way that produces working software.
"Show me a technical decision you made on a recent project and explain the tradeoff you rejected."
A competent agency answers this in two minutes. They chose a technology or architecture, passed on an alternative, and know why. The alternative matters as much as the choice. If the answer is a feature list or a client reference, not a tradeoff, they are not thinking in decisions.
"How do you handle scope change after kickoff?"
Scope changes. The honest answer involves a process: who flags it, who approves it, what happens to the timeline and cost. The dishonest answer involves flexibility. Flexibility without process is how engagements become open-ended and unpriceable.
"What does done look like for phase one, and how will we both know?"
Done should be observable. A specific screen the client can use. A workflow that runs end to end. A number that moves. If the agency cannot answer this in concrete terms before the engagement starts, they will not be able to answer it six months in either.
Realistic Timelines
The VitalRegistry CRM gave us the numbers founders ask about most.
Six weeks to a Streamlit prototype. That prototype was functional enough that our mother stopped using her spreadsheets and started using the app without being asked twice. It was also fast-and-ugly on purpose. A prototype is not a product. It is a question with a working answer attached.
Five months to the full React production build. The rebuild included a real design system, proper data models, contract generation, one-way Google Calendar sync, and in-app reminders surfacing follow-ups before they fall off the calendar. The second build was the one that looked and felt like a product. The first build was the one that told us whether a product was worth building.
Founders routinely underestimate both numbers. They expect a prototype in two weeks and a production build in two months. Agencies that promise those timelines are pricing to win the deal.
Four to eight weeks for a prototype with tight scope and a working discovery process. Four to six months for a focused vertical production build. Anything less is borrowing from your future.
The Discovery Process That Actually Works
The discovery process we use starts the same place a good agency kickoff should: with the existing workflow, not the desired feature list.
Before writing a line of code for VitalRegistry, we mapped the spreadsheet. Not the data model we wanted to build, the actual spreadsheet our mother used. The columns she had named, the color coding she had invented, the filters she had built to find overdue devices. That map told us more about what mattered to her than any requirements session would have produced.
The questions it generated were operational. Not "do you want a dashboard" but "what is the first thing you check every morning." Not "do you need reporting" but "how do you know at month-end whether you have collected everything you are owed."
Those questions define done. Done is: she opens the app and the first screen answers the question she used to open the spreadsheet to answer. Done is: month-end reconciliation runs in minutes instead of a day.
A discovery process that does not end with those definitions is a conversation, not a foundation.
The Principle
The founders who get burned by agency engagements are not naive. They are hiring into a process problem they have no way to diagnose from the outside.
The six-month silence followed by a wrong delivery is a failure baked in during week one, when nobody stopped to ask what done looked like, who was authorized to say so, or what the agency had decided to build before understanding what already existed.
Those questions do not guarantee a good engagement. They guarantee you will find out in week one, not month six, whether the agency knows how to run one.
The real question for founders considering their first agency engagement: do you know, in specific and observable terms, what you will be able to do at the end of phase one that you cannot do today? If you do not, that is the first conversation to have, before a quote is issued and before a contract is signed.