There’s an old consulting story:

A machine breaks down. The factory calls in an expert. He listens, thinks, then taps the machine once. It roars back to life.

He sends an invoice:

• Tapping with hammer: $1
• Knowing where to tap: $9,999

It’s a perfect metaphor for the kind of work we do in capital markets technology — where the most valuable contributions often come not from effort alone, but from expertise applied with precision.

At Phi Partners, we see this regularly: small, high-leverage interventions that avoid major cost, performance, or risk issues before they escalate.

Some of our favourite real-world examples from the wider industry:

• A one-character code change at Amazon — swapping a copy() for a reference — reduced compute needs equivalent to 15,000 servers.
• A quant at a major bank fixed a pricing model bug with a single line of code — avoiding hidden arbitrage and potential regulatory exposure.
• A trading latency issue resolved by switching a system call — restoring competitiveness in microsecond-sensitive markets.
• A bug in TCA reporting double-counted spread costs — fixed with one if clause, preserving a key client relationship.

These aren’t stories about heroics — they’re stories about judgment, context, and knowing where to look.

---

We recently had our own “knowing where to tap” moment on a cloud infrastructure engagement with a global financial institution:

• The client’s CloudTrail was logging high-volume S3 data events — information that was already captured more efficiently elsewhere. These logs were ingested into CloudWatch at rates of up to 2–5 GB/sec, driving significant costs.
• Their CloudWatch Agent was also exporting up to 200 custom metrics per server, using default settings that weren’t aligned to business needs.

Our cloud consultants reviewed and reconfigured both:

• Disabling redundant event logging
• Optimising metric exports for relevance and efficiency

Resulting in estimated annual savings of $462k and significantly reduced operational overhead.

No new tools, no replatforming — just experience, applied in the right places.