Half a million users. Millions of transactions. One truth.
MultiTankCard CEO Patrick Roozeman on a column swap that turned a hundred-euro invoice into ten thousand, an audit that depended on suppliers, and why volume is not the enemy of quality.

A customer opens an invoice for charging an electric car. It should say about a hundred euros. It says ten thousand. Nothing crashed and no system raised an alarm. A column was swapped upstream, and every system after it did exactly what it was told.
That is one of the stories Patrick Roozeman told Peter Engel in the first episode of High Volume Accounting Live. Patrick is CEO of MultiTankCard, the largest business mobility platform in the Netherlands. Its card covers fuel, EV charging, parking and public transport. He has more than 30 years in transaction processing, including a background at American Express. The conversation is about the moment a CEO stops accepting numbers that are almost right.
Nobody notices until it breaks
Patrick’s first lesson comes from his years in payments. "The most important lesson is nobody cares about it until it goes wrong." Transaction processing is invisible work. As long as it holds, nobody asks how. When it breaks, it is the only thing anyone talks about.
At MultiTankCard that pressure is constant. "We have over 500,000 end users", and every one of their transactions has to be right.
The ERP records what it gets
The invoice story shows why. In Patrick’s words:
"We have had a situation where the kilowatt-hour column was swapped with the euro column. And it sounds too trivial. But of course for a customer that means that he pays a hundredfold. So what would normally expect a bill of a hundred euros, you suddenly get a bill for ten thousand euros."
An ERP writes down what it receives and passes it on. It does not check whether a kilowatt-hour is a euro. If validation only happens after the ERP, the customer is the one who finds the error.
When suppliers know your volume better than you
The most candid moment of the episode is about his first audit at MultiTankCard. The transaction data was incomplete.
"I had to phone certain oil companies to ask them what they thought our volume was. If you don’t have that control, that seems very strange. And it was strange."
If the only way to confirm your own volume is to ask your suppliers, you cannot prove completeness to an auditor. What changed it was receiving the data at source, one transaction at a time. "Only late last year, we have been able to receive 99% of our transactions individually and in real time."
Volume is not the problem
The switch from fuel to EV charging shows how fast volume can grow without the business changing. A combustion car refills its tank roughly once a week. Charging an electric car happens far more often, and Patrick’s conclusion is that "you already have a transaction volume that is fourfold". Same driver, same car, four times the transactions.
Yet his closing line points the other way: "volume is not the stumbling block for quality". Volume makes whatever is already in your foundation bigger. With validation at the source, more transactions give more signal. Without it, they give more errors.
Read the data, not just the euros
Patrick also has a view on what this asks of finance people. Counting money in and out is only the surface of the job. What produces the numbers is the data underneath. Finance teams, he said, "should be far more data minded than Euro minded".
He goes one step further on the close.
"The technology should allow you to close on a single-day basis."
That only works when every transaction arrives validated, so there is little left to rebuild at month end.
What to take from it
Three questions from this episode are worth asking in your own team.
- Where in your chain could a column swap travel all the way to an invoice without anyone noticing?
- Could you prove your transaction volume to your auditor today, from your own data?
- If your volume quadrupled next year with the same customers, which step in your process would break first?
If one of those answers starts with "probably", the Actuals scan helps. It takes 15 minutes and shows where your transactions and your books first stop matching.
In practiceSix factors decide whether your numbers hold.


