Your data is not the problem.
The decision is.
Every e-commerce brand past £5M has more data than it knows what to do with, and a leadership team that still cannot say what to change on Monday morning. That gap is where the cash goes.
It lives in six systems that do not talk to each other
Shopify knows what sold. Meta and Google know what you paid to sell it. Klaviyo knows who opened the email. The ERP knows what is left in the warehouse. The warehouse system knows what is actually on the shelf, which is not always the same thing. Xero knows who has not paid you yet.
Six sources of truth. No truth.
So the MD asks a simple question. How much cash is tied up in stock that is not moving? Someone disappears into spreadsheets for two days. The answer lands on Thursday, and by Thursday it has already moved.
Centralising the data changes nothing on its own
So you buy a BI tool. Now all the numbers live in one place, and the decisions still do not get made. You look at it in the morning. You look at it again after lunch. Brendan calls this data voyeurism, and it is exactly the right phrase for it. Watching your numbers is not the same as running your business.
Forty charts. Every one of them accurate. Not one of them tells you whether to cut the TikTok spend this week or leave it alone.
More reporting produces more looking. It does not produce a decision.
So the four decisions that move cash never get made
The decisions that change a commerce P&L are always the same four. Kill the spend that stopped working. Move it to the spend that is working. Order more of what is selling. Discount what is not.
Every one of those answers is already sitting in your data. They do not get made, because making one properly means holding marketing, stock and cash in your head at the same time, and nobody does that at 7am on a Tuesday with a board pack due.
All three are free cash flow leaving the business. For a commerce company, free cash flow is the whole game.
This is not a competence problem
The MDs and CMOs we work with are sharp operators who know their category better than any consultant will. Hand the best of them six platforms, 6,200 SKUs and no joined-up view, and they will still miss the £27,000 sitting in a wool throw that is selling out whilst the budget goes somewhere else.
Nobody is short of intelligence here. They are short of a surface that does the joining, so the judgement they already have gets pointed at the right thing.
Not a talent problem. A surface problem.
A decision layer, not another dashboard
One surface with the marketing, the sales, the stock and the cash joined into a single view. That part is table stakes and Supermetrics does most of it out of the box.
The part that is not table stakes is what sits on top. Something that knows the margin on every category, the lead time on every supplier, which one of them runs late, and where the business is actually trying to get to. It reads the joined data every morning and comes back with what to do next, in order, with the number attached and the reason stated.
You stay the gatekeeper. Nothing moves without you approving it.
Stop reading your data. Start deciding on it.
That layer is Claude
A good chief of staff knows the business inside out, walks into the room having already done the reading, and puts the decision in front of you clean. They never take the decision. That is the job Claude does. Your marketing agents sit inside it, and it sits between your data and your Monday morning.
A dashboard knows your numbers. It does not know that Tableware carries your Christmas, that TerraStudio always runs a week late, or that the board agreed in January to hold margin above 50% even if it costs volume. Claude holds all of it.
It does not make the call. It makes the call clear.
Act one said this was a surface problem rather than a competence one. This is the surface. Claude does the part nobody can do at 7am, which is holding the marketing, the stock, the cash, the SOPs and the three year plan in mind at the same time.
What comes out is not an instruction. It is your decision, with the fog cleared, the number attached and the reason stated. You still say yes or no, and nothing moves until you do.
Your judgement. Now with something worth judging.
Here is the surface, running on a real trading month
Composite data from a 6,200 SKU homewares brand. Change the range, the channel or the category and every figure recomputes, because all of it derives from the same joined dataset. The queue at the bottom is Claude's brief on what that data means today.
