Why Agencies Fail Serious Operators

Most agencies are competent at building software. That is not where they fail. They fail at the decision that precedes the build — whether the thing deserves to exist at all — because nothing in their business model rewards them for getting that decision right.

This memo is not an attack on any firm. It is an incentive analysis. If you are an operator considering a serious build, the incentives of your vendor are as material as their portfolio, and considerably easier to read.

Follow the billing

An agency bills for output: hours, deliverables, releases, retainers. Revenue scales with scope and duration. Whether the product finds a paying market does not appear anywhere on the invoice.

Understand what this means in practice. A vendor who questions your idea shrinks their own contract. A vendor who tells you "not yet" loses the engagement to someone who will say yes. Nothing corrupt is happening — everyone behaves exactly as their incentives instruct. The result is a market of firms structurally unable to tell you the one thing you most need to hear.

The vendor's yes costs nothing to give. It is only expensive to receive.

The wrong first question

"What features do you want?" sounds like service. It is risk transfer. The moment your guess becomes the specification, every commercial risk in the project belongs to you — and every hour of building it is billable to them.

The right first question is about the buyer, not the product. Who is this for? How often does the pain occur? What do they pay today to tolerate it? Why would they switch? A firm that cannot hold that conversation is not equipped to protect your capital, however well it engineers.

Watch, too, what happens to scope once the feature list becomes the contract. Every ambiguity resolves in one direction. Every new idea is a change order. The economics of the engagement improve as the build grows — for one party. Operators who have managed construction projects will recognize the pattern immediately. The difference is that in construction, the owner usually commissioned a survey first.

The pitch is senior. The build is not.

There is a second structural problem: leverage. Agency economics improve with the ratio of junior hours billed to senior hours worked. The principals who impressed you in the pitch are the firm's business development function. The moment the contract is signed, their attention moves to the next pitch, and your product — the asset your capital is buying — is thought through by the least experienced people in the building.

For routine work this is tolerable. For a first product, where every early decision compounds, it is quietly fatal. The expensive mistakes in a build are not typing mistakes. They are judgment mistakes: what to include, what to refuse, which corner of the market to enter first. Judgment does not delegate down the org chart.

Anatomy of a failed serious build

The pattern below is real. We have reviewed versions of it many times — details changed, sequence identical: a strong operator, a weak thesis, and a vendor who said yes.

Note what never happened. No one tested willingness to pay. No one priced the switching cost. No gate stood between conviction and capital. The operator's insight was genuine — two decades of it. The failure was not the idea, and it was not the code. It was the absence of any structure that could have caught a weak thesis before month nine.

What operators should demand instead

Five things. None of them are technical.

1. Evidence before scope. Research on buyers, workflows, and substitutes — before any feature is specified or priced.
2. Decision gates before capital. Defined points where the work stops and you decide, with evidence in hand, whether it continues.
3. Senior judgment in the room. The people who won the engagement should be the people doing the thinking. If the principals disappear after signature, the judgment did too.
4. Working proof before commitment. Software you can put in front of real buyers — before the serious money moves.
5. The right to hear "not yet." Ask a candidate firm when they last told a client to stop. If the answer is never, they are not filtering. They are billing.

Cheap builds are the most expensive

The discounted bid is priced on hours, and hours are the one input that says nothing about outcomes. What the cheap build omits is exactly the work that protects capital: research, validation, senior review, and the discipline to stop. Those line items are the first to disappear under price pressure because they are the hardest to demonstrate in a proposal — and the only ones that determine whether the asset earns anything.

Count the full cost of a failed low-cost build. The fee. Then the year of attention. Then the market window. Then the internal credibility of the next digital initiative, which will now be judged by this one. The cheap build is rarely cheap. It is deferred, compounding expense.

Encodex was structured against these incentives deliberately. Two engagements per month, chosen by application. Evidence before scope. A working prototype before any commitment of capital — free, so the advice owes nothing to the invoice. And the standing obligation to say "do not build yet" when the evidence says so. That is not generosity. It is the only structure under which product advice can be trusted.