Operating principles

Every system I run is held to the same standards: a measurement before and after each change, counters and alerts on every data flow, a monthly report written for the owner. Every claim in that report carries one of four labels.

  1. I know.

    Backed by a query, a log entry or a test. The number comes with its source.

  2. I don't know.

    Said immediately, with what it takes to find out and how long that takes.

  3. I'm estimating.

    A range with its assumptions. Replaced by a measurement as soon as one exists.

  4. I think.

    An opinion, labelled as one, with a way to verify it.

Operating practice

The standards every system I run is held to.

Baseline before every change

The number a change should move is recorded before release and read again after thirty days.

Counters on every data flow

Each run logs rows read, written and rejected. A run that writes nothing raises an alert.

Performance from the query plan

Slow queries are diagnosed in the database, indexes follow from the plan, timings before and after are recorded.

Vendors and APIs monitored

Error rates, quotas and spend tracked daily per vendor. Budget limits stop the pipeline before they are exceeded.

Monthly reconciliation

Orders, settlements and invoices matched across all channels. The margin figure holds up to an audit.

One-page monthly report

What changed, what it earned or saved, what failed and how it was fixed, what is next. Numbers first.

Non-negotiable

Each of these has cost a company money. Declining them is part of the service.

  • No pricing change without a before-and-after measurement.
  • No migration declared complete before every row is verified.
  • No deadline without a range and the assumptions behind it.
  • No AI message sent to a customer without human approval.
  • No figure reported without its source.
  • No opinion presented as fact.

Want your system reported on this way?

Get in touch