Hey Reader 👋

Last week, while hiking through the British countryside with my phone blissfully on "Do Not Disturb," I spotted this horse crossing warning sign.

No AI. No sensors. Just a simple metal warning to say: “Something happens here. Often.”

But here's the thing – while admiring that sign, I started thinking about a conversation I had at a networking event just days earlier

I was listening to a finance director confess that she hadn’t had to approve a single transaction for multiple quarters.

Sounded like a good problem - until it wasn’t.

The Warning Sign

Her company had strict £10,000 approval rules.

Decent segregation of duties.

Smart workflows.

On paper? Rock solid.

But reality had other plans.

Spend kept climbing. Vendors multiplied. Projects mushroomed.

Yet nothing - nothing - breached the approval line. And no approvals came to her.

Turns out, they'd stumbled into the approval blind spot.

Just like horses have learned the safest crossing spots, your vendors and internal teams have learned the safest spending spots.

Right under your approval threshold.

Let’s dig into what this looks like and how to surface it.

The Invisible Cavalry Charge

That £15K software license? Split into two £7.5K invoices.

That £12K consultancy project? Becomes a £9K initial phase plus a separate £3K "additional scope."

Those office supplies? Mysteriously (or rather shamelessly) always come to £9,999.

No approval needed. No visibility for you. But your budget? Still disappearing.

If no one’s breaking your thresholds anymore, it doesn’t mean the big spend stopped. It means the dodging got smarter.

Time for Better Warning Signs

Suddenly, you're not just catching the obvious stuff. You're seeing the whole herd.

This is exactly where data science shines. Instead of relying on static thresholds, we can build dynamic pattern recognition that learns and adapts.

Machine learning algorithms excel at spotting anomalies in spending patterns. They can identify vendor clustering, detect suspicious timing sequences, and flag unusual amount distributions.

But you don't need a full machine learning setup to start catching these patterns. Sometimes a well-crafted SQL query does 80% of the work.

Warning Signs You Can Script

🎯 SQL Query: Detecting Threshold-Dodging Transactions

It’s an elegant way to turn smoke into signals.

Want to go deeper?

Here are three variations:

1. Month-End Gaming (detects timing tricks):

2. The Bold 9999 Players (zero shame invoice gaming):

3. Cross-Project Splits (uncovers distributed dodging):

Mix and match these based on your data structure and what patterns you're hunting. The beauty is in adapting the logic to your specific threshold-dodging creativity.

Key Take-aways

✅ Repeated sub-threshold spend is rarely random.
✅ SQL is your first line of pattern defence - no AI needed (yet).
✅ Once detected, start flagging, not blocking - make it visible.

If a horse crossing gets a sign for being too frequent, maybe your vendors deserve one too.

If this gave you ideas, raise a flag. If it raised your eyebrows, forward it to a friend or invite them to subscribe here.
And if it made you open your SQL editor - well, that’s a pattern worth sharing.

Have a cracking week!

Tony

P.S. If one of those poll options hit a nerve, take 3 seconds and vote. It’s quick, unless you decide to reply with a war story. (Those are always welcome.)

Your click helps shape what lands in your inbox next.