The Audit Room and the Kitchen
What statutory audit taught me about the gap between a business's story and its numbers.
The high floors of a bank at midnight during audit season are quieter than you expect. Not empty — there are always three or four of us, and someone from the client’s team who cannot go home until we do. Quiet in the way a room is quiet when everyone in it is doing arithmetic.
That was where I learned there are two ways to find out whether a number is true.
You can test the controls — examine the machine that produced the number, and if the machine is sound, trust its output. Or you can test substantively — ignore the machine, go to the underlying records, and rebuild the number yourself.
Controls testing is cheaper. Substantive testing is certain. Every audit is an argument about how much of each you can afford, and the argument is never really about cost. It’s about how much you’re willing to believe the story a system tells about itself.
I left in 2009. I have spent every year since inside businesses that run on narrative, because hospitality is a narrative business. And I have never stopped running the same two tests.
What the discipline buys
I was a founding member of the team that brought Pita Pit to India. I opened the first three outlets myself, in landmark locations across Delhi NCR.
Opening a store is substantive testing in its purest form. You are standing in the room. You count the covers. You watch the closing checklist get completed because you are the one completing it. Nothing is being reported to you — you are the record.
Then I went international, and everything inverted.
Regional director across India, the UK, the UAE, Saudi, Singapore. I could not stand in those rooms. What I had instead was the playbook: the SOPs, the training systems, the quality controls, the reporting cadence. I stopped verifying outcomes and started verifying the machine that produced them.
This is controls testing, and it works. It is the only thing that works at that distance. Twenty-one units in India, franchise relationships across five geographies — none of it is possible if you insist on counting the covers yourself.
But something happens to you when you spend years trusting systems, and it does not announce itself.
What it costs
By the time I was signing franchise agreements in India, I had spent five years trusting playbooks. So I trusted the paperwork.
The choice is FOFO, FOCO, or JV. Franchisee-owned and operated. Franchisee-owned, company-operated. Or a joint venture where you take equity and share the downside. On a spreadsheet this is a clean trade — capital intensity against operational control, plotted against how fast you want to grow.
I read it as a capital question. FOFO is capital-light. The franchisee funds the build, carries the risk, and you scale on someone else’s balance sheet. Against a target of twenty-one units, with the money and the pipeline both in front of me, the model was obvious. I ran the substantive test on the thing I knew how to test: unit economics. Cover counts, rent as a percentage of sales, breakeven month, payback. I tested it properly. The numbers held.
They held because they were the wrong numbers.
A family investor with capital is not an operator. In India the two arrive together and look identical at signing — the capital is real, the enthusiasm is real, the family has run businesses for three generations. Nothing in the diligence pack distinguishes an operator from an investor who intends to hire one. And a franchisee who does not personally run the store does not notice, on a Tuesday in month four, that the closing checklist stopped being completed.
No sample size tells you this. This is a controls question wearing a substantive question’s clothes, and I tested the wrong one, and it cost us stores.
The auditor’s instinct is to widen the sample. The operator’s job is to know which test you are running. Learning when to stop testing took me a long time. Learning what to test took longer.
The synthesis
In 2020 I went to build cloud kitchens.
A hundred and twenty-two of them, three cities, under twenty-four months. Ninety per cent occupancy inside twelve, across more than a hundred enterprise F&B brands.
Those are the numbers. But 122 is not the number that matters. Ninety per cent occupancy is the number that matters, and it only means something if you know what it is a ratio of — how sticky the tenants were, how quickly a vacated kitchen refilled, whether occupancy was being bought with pricing.
A number without its denominator is a story wearing a number’s clothes.
That is the thing audit teaches, and the kitchen tests, and the two rooms are less different than they look.
The gap
Capital and execution speak different languages.
Capital speaks in multiples, cohorts, terminal value, comparable transactions. Execution speaks in covers, wastage, roster hours, the thing that broke on Saturday. Both are describing the same business. Neither can hear the other clearly.
Most businesses do not fail at strategy or at operations. They fail in translation between them — a plan built on numbers that operations cannot produce, or an operation generating numbers that capital cannot read.
I have spent two decades moving between those two rooms. I have never found a shortcut. What I have found is that the work of translation is real work, it is nobody’s job title, and it is where most of the value goes missing.
That is the gap I build in.