For CFOs of growing companies in Europe
Compare what each data or AI investment costs per year with what it returns: work it replaced, revenue it added, risk it reduced or decisions it improved. Every return needs evidence rather than assumption, and where none can be shown, it is a candidate to stop.
Free. Two questions you can answer from memory: what you cut, and what you've added since.
This year's budget decisions
Illustration
Started
Stopped
Most companies have an owner and a process for starting things: a budget line, an approval, a kick-off. I keep seeing almost none for stopping them, so tools, hires and data work carry on after the reason for them has gone.
My view is that data, and the spend around it, is company capital, not a cost centre parked under technology. Capital gets reviewed, and some of it gets withdrawn. I make that withdrawal as legitimate as the original approval, with a euro figure beside each decision.
The work ends in decisions made. You get a list of what to stop, change, keep or start, and I stay until the first cancellations are made. Then I leave.
The idea in 67 seconds
Captions on. Nothing loads until you press play.
Most companies have a process for starting things. Almost none has one for stopping them.
Lior Barak
If two of these came up in your last budget meeting, the invoices aren't where the problem is.
Ask each owner what the tool replaced, added or protected, and put a euro figure on it. Lower risk and faster decisions count too, if the owner can show the evidence. If no return can be shown by the next renewal, stop it.
New hires inherit the maintenance and ad-hoc requests that fill the week, so the backlog doesn't shrink. Measure how much of the current team's week goes on keeping old work running before you approve the next hire. Fix that share first, then hire.
Because a cut removes the tool, not the demand behind it. The work moves to people, spreadsheets or a new subscription. Keep the saving where nothing replaced it, and stop the replacement where something did.
The hours spent checking and rebuilding numbers nobody trusts, plus decisions that wait. In one company a finance report was rebuilt by hand every Friday, checked twice, and decisions waited two days. Fix the source of the mistrust, then stop the rebuild.
Want the answer for your company?
Run the 2-question CFO checkIn one e-commerce company, the finance report was rebuilt by hand every Friday. Numbers were checked twice before anyone trusted them. Decisions waited two days for an answer.
None of it appeared on an invoice. I keep seeing the same thing: the cost sits in people's time, not in the ledger. Priced at their own payroll rates and cloud bills, it came to €930,000 a year.
One e-commerce company, one year
From a client engagement. Anonymised.
Get your own number.
Run the 2-question CFO checkFTE Debt
The hours other teams lose working around data, multiplied by their fully loaded hourly cost, gives a monthly euro figure.
PCU-V
What each product costs per month to keep alive: infrastructure, plus the incidents, support and changes that take people's time.
CDSI
How much of the data team's time goes on maintenance, incidents and ad-hoc work instead of new work.
Two questions, no lookups. You get a net position: whether last year's cuts are still savings.
Run the checkWe test what the free check can't see: whether the saving held, who absorbed the work, and whether people still trust the numbers. The fee is credited in full to the audit.
See the workshop →Where the money and people's time go, with a list of what to stop, change, keep or start, each with a euro figure. I stay until the first cancellations are made, then I leave.
Offered after the workshop
I'm Lior Barak. I spent fifteen years inside data organisations, including Zalando and idealo, deciding what to build, what to keep and what to shut down. I wrote What Data Really Costs, a business fable about the money and time a company loses without seeing it.
About Lior →