Hey {{first_name|there}},
This is email 5 of 6 in a series. New here? Start from the beginning:
1. Everyone's talking about AI. Few are using it.
2. My AI disagreed with my EBITDA. It wasn't wrong.
3. The CFO skill nobody admits they're bad at
4. Revenue grew 10%. So why did contribution margin fall?
Last email, I told you about a client who doubled his revenue and made less money doing it.
He called me insane for telling him to slow down.
He didn't slow down. His next plan: grow another 30%.
The numbers below are illustrative, to protect him. The story is real.
$40M of revenue. EBITDA at 5%, down from 15% when he was half the size.
Next year's P&L looked fine. $52M of sales. EBITDA up from $2.0M to $2.6M.
But the new customers paid in 90 days, not 45. And serving them meant holding 120 days of stock across the whole business, not 50.
So the extra $12M of sales would tie up about $10M of cash:
$3M more in receivables
$8M more in inventory
minus about $1M of extra supplier credit
Borrowed at 8%, that's $0.8M of interest a year.
The new sales added $0.6M of EBITDA.
The growth cost more to finance than it earned.
You find that kind of information in a model.
What a model is really for
A P&L shows the sale. A full model, with the P&L, balance sheet and cash flow linked together, shows what the sale does to your cash.
That's when the real questions start:
Do we have the balance sheet to carry $10M more?
Do we renegotiate the operating line before we commit?
Do we put capital in, or wait for the growth to prove itself?
Or do we grow 15% with customers who pay on time?
Each question is a scenario. Each scenario can change the decision.
The costly mistake is approving a plan you only tested once.
Why most teams test one scenario
Built by hand, a model like this takes close to a month. 15 to 20 working days.
So most teams build one version: the plan. The hard questions get asked after the money is committed.
In email 1, I told you about the board model I built in 27 hours over one weekend.
That weekend taught me something. The method was the real asset.
So I turned it into a skill: written instructions that an AI tool can pick up and follow. I've run it in both Claude and ChatGPT.
Now the next model doesn't start from a blank spreadsheet. Scenarios switch on and off, so you see each option's impact side by side.
And adding a new scenario takes a few tweaks.
I still review every assumption. The judgment stays mine.
What changed: I answer the board's "what if" before the meeting, instead of the usual “Let me get back to you on this”.
What's next
On Wednesday, November 11, at noon Eastern, I'll show you, live, the dashboard I built with AI, and how to keep AI's bill from surprising you. It’s free, and it’s only 45 minutes.
P.S. What’s the main challenge for finance teams when it comes to financial models? Hit reply. I read every answer.
Disclaimer:
This content is for informational and educational purposes only and should not be construed as financial, legal, or professional advice. Always consult with a qualified advisor before making any business or financial decisions. The author and publisher disclaim any liability for actions taken based on this content.
Talk soon,
