Every vendor and every candidate will tell you they are data-driven, and the claim cannot be tested in a pitch. What can be tested is sequence: what they build first when handed a real budget. Whoever builds the scoreboard before touching spend becomes accountable to it. Whoever starts with campaigns stays unaccountable forever, whatever their dashboards say.
In my first two weeks as fractional CMO at FX Replay, a trading simulation software company, I shipped no campaigns, no creative, and no budget changes. I shipped four events.
Four customer actions, defined precisely, recorded server-side, agreed with the founders as the only numbers that would count: the event dictionary the whole engagement ran on. Every dollar of the next 18 months was judged against those four definitions, including the dollars that judged me.
That order of operations looks slow from the outside. It is the fastest thing I know how to do for a company that has already found demand, and this piece walks the logic so you can put the test to anyone you hire, me included.
Why the scoreboard comes first
A marketing engagement without a trusted scoreboard produces opinions at senior prices. Smart opinions, often. Still opinions, because nothing can settle them.
The left column is where growth companies live for years. Every question on it feels strategic, gets debated as strategy, and is actually a measurement question in a strategy costume. Seniority cannot answer it. Instrumentation can.
The first thing I build is the scoreboard, because every argument after it gets shorter.
There is a personal reason in it for me too, and it belongs in the open: building the scoreboard first is how I make myself falsifiable. A leader who instruments the business before touching it has volunteered to be graded. That is either attractive to you or it is a warning about the last people you worked with.
What should you measure before spending?
The event dictionary carries more weight than its size suggests. Four to eight events forces the argument about what actually counts as progress, and that argument, had once, in plain language, is worth more than most strategy decks. Signups or activations. Trials or the moment payment clears.
That session needs the founder in the room, because the argument it forces is a business argument wearing technical clothes. Whether a trial counts at signup or at first payment decides what every team optimizes toward for years. Delegate it to a vendor and the company ends up optimizing to a definition nobody actually chose.
The server-side layer exists because platform pixels are witnesses with a financial interest in the verdict; the signal problem has its own full treatment. Identity exists so the same customer stops being three different people in three systems, and it can start as a disciplined join key long before it becomes a platform, which I have built for about $30 a month when the time came.
The North Star ratio compresses the rest into a number a founder can carry into a board meeting. Mine, for multi-model businesses, is gross-profit-based and paired to acquisition cost, but the specific choice matters less than the founder being able to recite it.
Why vendors skip this
Not incompetence. Incentives. Measurement work is invisible in a monthly deck: no launch, no creative, nothing to screenshot. A vendor billing for activity will rationally spend the first month shipping visible activity.
And a clean scoreboard is a risk to whoever reports against it, because it can contradict the deck. Reporting follows billing, everywhere, which is why asking a vendor to grade their own work with instruments they built after the campaigns started produces the reporting you already distrust.
So put the sequence question in your vetting, right next to the receipts: given our budget, what do you build before you touch spend? The answer sorts the market faster than any case study.
You can also read the tell in proposals you already have. The typical agency proposal reaches campaign concepts by week two. The measurement line item, where one exists at all, sits in an appendix priced at zero, which tells you exactly what the model believes that work is worth.
What it costs, honestly
Less than one month of most retainers. The dictionary is a working session. The server-side layer is an engineering sprint in most modern stacks. Identity starts as discipline rather than software. The infrastructure bill is routinely two digits a month.
There is also a trust-building phase the timeline should include: a few weeks where the new scoreboard runs beside the old reports and every disagreement between them gets explained, one number at a time. Skipping that window is how technically correct measurement ends up politically dead, still contradicted in meetings by whoever preferred the old deck.
Payback tends to arrive embarrassingly early, and usually as subtraction. The first thing a working scoreboard finds is spend that was never doing anything, and cutting it has funded the build many times over in every engagement where I have run this sequence.
The expensive ingredient is judgment: knowing which four events, which join key, which ratio. That is what you are actually paying a senior person for in the first weeks, and it is why those weeks predict the entire engagement.
The build then feeds the machine that runs on it: the operating system and the weekly rhythm both assume a scoreboard nobody argues with. Built in this order, they get one.
The scoreboard build is the opening move of every audit I run. If your monthly numbers cannot currently settle an argument, let's talk.
