Measured with instruments, not memory: the method behind my systems

The diagnosis

In August 2026 I rebuilt the twenty projects in my portfolio from memory. The result was 138 data points marked as pending, spread across all twenty records without exception: users nobody counted, times nobody measured, dates nobody wrote down. The work existed; the evidence did not.

That is where the method's thesis comes from:

Documentation is written during execution; results are measured with instruments, not memory.

Five phases, three gates

Every project goes through the same five phases: framing, discovery, design, build and adoption. Between them there are three gates —kick-off, build and close— where I decide whether the project moves forward, moves forward with conditions, is reworked or is dropped.

A gate is not a meeting. It is a checklist run against the evidence and recorded with a date, a decision and its conditions. An unmet criterion doesn't invalidate the gate: it leads to a “conditional go”, with the condition written down. What invalidates it is not looking.

Why it isn't DMAIC

For a while I described my way of working as DMAIC. That was inaccurate. DMAIC contributes one specific piece to the method —define and measure before improving: the baseline— but not the full cycle. Each piece has its source:

• Stage-Gate: phases separated by formal decision points.
• ISO 21502:2020: life-cycle gates, transition into operations and benefits measured after closing.
• BPM life cycle (Dumas et al.): discover and analyze the process before redesigning it.
• DMAIC: the baseline and the measurement.
• ADKAR: adoption, one person at a time.

One piece was taken from each source and the rest was left out. Calling it my own methodology would be inaccurate; presenting it as a reasoned adaptation of recognized standards is defensible.

Five routes

A single method for everything has to be stretched. A client engagement, a capstone project and a prototype aren't approached the same way, so the method distinguishes five routes by the nature of the work:

• Process digitization: there is a paper form, an Excel file or an email someone fills in today. It is understood, redesigned and replaced.
• Client engagement: someone external pays. Scope becomes an agreement and acceptance becomes their call.
• Own system: nobody asks for it or pays for it. The method exists to force closure, with a public date.
• Academic: an institution sets the date and the criteria. You plan backwards and translate the rubric.
• Exploration: the outcome is learning. Concluding “it doesn't work” also counts as finishing.

Each route also gets an intensity —light, standard or full— that decides how much documentation it requires. Size tells you how much paperwork; the route tells you what to do.

Three principles

Process first, software second. Digitizing something nobody defined hard-codes a made-up practice. In the Leaders App that meant mapping the Standard Work and updating the procedures before writing the first screen: an AI agent built on outdated documentation spreads the error instead of fixing it.

Built is not adopted. The project ends when people use it, not when it works.

Measure the before, or state why it wasn't measured. No figure is claimed without an instrument; when there is no data, I write down that it wasn't measured and why, never an estimate dressed up as a measurement.

What this changes on this website

The projects on this site are no longer written by hand: they are generated from the portfolio records. A data point marked as pending is not published, and a year the record doesn't confirm doesn't appear on the card. That's why some cards say less than they could. It's on purpose.