Case Study Zero: why we built Trazer on our own leaking pipeline

We didn't set out to build a product. We set out to stop losing money in ways we couldn't see.

Bettroi ran the way most service businesses in the Gulf run, and if you run one, the next paragraph will feel uncomfortably familiar. Proposals were Word documents, duplicated from whichever old deal looked most similar and edited under deadline. Commercials lived in a spreadsheet that one person understood completely, two people understood partially, and everyone else treated as scripture. Client conversations lived in whoever's inbox the client happened to reply to. The pipeline existed as a weekly meeting and a shared sense of optimism. And the handoff from sales to delivery, the single most consequential moment in a project's life, was a meeting if we were disciplined that week, a forwarded email if we were busy, and occasionally nothing at all.

Here is the thing that took us years to admit: none of it felt broken. That is the trap. Every individual tool worked. Word produced proposals. Excel added numbers correctly. Email delivered messages. The failure never lived inside any tool, so no tool ever showed it to us. The failure lived in the gaps between them, and there is no report anywhere in that stack whose job is to describe a gap.

The proposal that never left

The moment that started the build was small and, in hindsight, embarrassing.

We were chasing a deal we genuinely wanted. The scoping had gone well, the client was engaged, and the proposal was written, reviewed, and polished. Then it sat in a drafts folder for eleven days.

Sales believed it had been sent, because in their memory the work was finished. The client, hearing nothing, assumed we had lost interest and quietly moved on to a competitor who answered faster. By the time anyone noticed, the deal was gone, and it had died of the single most preventable cause in commercial life: silence.

What made it sting was not the lost revenue. It was the realisation that our system, if you could call it that, had no way of even representing the problem. The proposal was not late in any system, because it did not exist in any system. It was a file. A file has no state. Nothing in our stack could know that this particular file's state was "written, approved, and never sent", so nothing could flag it, and no amount of being more careful next time would change that. We did not have a discipline problem. We had an architecture problem.

The audit we should have done years earlier

That loss prompted the exercise that turned into Trazer. We sat down and listed every leak we could remember from the previous two years, every deal that went quiet, every project that finished thinner than the quote promised, every awkward month-end surprise, and looked for the pattern.

The list was longer than anyone expected, and the pattern was almost monotonous. Scope agreed verbally on a call, never written anywhere, delivered later as a free favour because the client remembered it and we had nothing to say otherwise. VAT treatment, inclusive or exclusive, decided differently by whoever happened to build the quote, discovered at filing time by our accountant. A milestone delivered in week two of a project and invoiced never, because the person who delivered it was not the person who bills, and nothing connected the two. A follow-up that lived in someone's head through a holiday, and by the time the holiday ended so had the deal.

Different failures, identical shape. In every case, something important existed only in a person's memory, or in a file with no state, or in a system that could not see the system next to it. Not one failure was a person being careless. Every single one was work living somewhere that could not hold it.

That reframing mattered, because it changed what a fix could look like. You cannot train your way out of an architecture problem. You can only change the architecture.

What we actually built

The requirement, once we understood the problem, was almost boring to state: one record per deal, carrying everything, travelling the whole distance from first contact to paid invoice, with stages that cannot be silently skipped.

Concretely, that became eight governed stages: lead, qualified, discovery, NDA, proposal, won and PO, delivery, and paid. Governed means the record enforces the path. A deal cannot drift into proposal without the compliance stage happening, because the stage is a gate, not a suggestion. When a deal is won, the project is created from it automatically, carrying the full history: the scope as sold, the emails, the call notes, the documents, the VAT treatment. Delivery starts from what was actually promised, not from a reconstruction.

Documents stopped being files and became outputs. A proposal in Trazer is generated from the deal's own data into our approved template, with a document number issued by the database itself, which means two people generating documents in the same second cannot collide on a reference, and nobody ever retypes a client's name or TRN onto something that ends up in a tax file. The eleven-day drafts-folder failure is now structurally impossible: a proposal is either generated and sent, or the deal is visibly sitting in the proposal stage, going stale, on a screen everyone looks at every morning.

That morning screen was the piece we did not know we needed until we had it. Every deal carries a score built from five visible signals, how recently anything happened, how fast it is moving through stages, its value, how many contacts we hold, how often they respond, and anything silent for fourteen days gets flagged. The score is deliberately not machine learning. It is arithmetic we can inspect, which means when it says a deal is dying, we can see exactly why, and argue with it if it is wrong. It rarely is. Deals do not announce that they are dying. They just go quiet, and now quiet is a number.

What changed, honestly

We are careful with claims, so here is precisely what we can say from our own operation.

Month-end changed in kind, not just in effort. It used to be an investigation: three people reconstructing what had been delivered from memory, email, and the spreadsheet. It is now a list. Unbilled work, the quietest leak in any project business, stopped being something we discovered and became something we cleared.

Scope creep changed shape. When a client asks for something outside the baseline, there is now a written baseline to point at, so the request becomes a priced change request rather than a favour absorbed in silence. The conversation is easier, not harder, because it is about a document instead of about memories.

And the follow-up problem, the one that started everything, simply stopped recurring. Not because we became more disciplined, but because the system now holds what memory used to drop.

Why this is our pitch

We tell this story instead of showing customer logos for a straightforward reason: we do not yet have customers with published, verified numbers, and in a market as relationship-dense as the Gulf, one fabricated claim costs more than a hundred honest posts. What we have is a system that was built against real leaks, in a real business, in this region, with its VAT regime and its NDA-first deal culture and its milestone-billed projects, and that runs that business today, end to end.

That is Case Study Zero. We are our own first customer, and the product exists because the alternative was continuing to lose money in ways we could not see.

One more thing we learned, which surprised us: the discipline the system imposed changed how the business felt to run, not just how it performed. When the pipeline is a screen instead of a weekly meeting, the meeting becomes a conversation about decisions rather than a reconstruction of facts. When month-end is a report instead of an investigation, finance stops being the department that asks everyone what happened. And when nothing depends on any one person remembering anything, holidays became actual holidays, for the first time in the company's history. We built Trazer to stop losing money. The quieter benefit was that it made the company calmer, and we would struggle to say which of the two we would give back last.

If your deal flow looks like ours did, the diagnosis costs nothing: bring your messiest deal, and we will map it through the system live. Where it would have been flagged, where the handoff would have been caught, where the invoice would have raised itself. Thirty minutes. If it turns out you do not need this, we will say so.