The Loop Methodology

The Loop's first principle is also its simplest to say and its hardest to adopt: measurement should be continuous, not annual. But “continuous” is a posture, not just a schedule. It changes how you build a measurement system, not only how often you look at it. This chapter is about that posture — why designing the whole system up front keeps failing, and how starting with one workflow and adding a step at a time gets you to something connected that actually holds.

The cycle, in one paragraph

Collect clean at the source, so every form, survey, note, and document enters as connected data instead of a pile to reconcile later. Analyze on arrival, so the moment a response lands it is read — open text classified, risk flagged, scores applied, each tied to the person's own words. Improve while you can still act, because the reading happened now: you fix a confusing question, catch a drop-off, or change a decision this week. Then the cycle runs again over the next wave, a little sharper each time. That is the whole method. Everything else is applying it well.

Why designing it all up front fails

The instinct is understandable: map every question, build the perfect form, configure the CRM, and then just run it for the year. But programs are moving targets. A funder adds a requirement, a cohort surfaces a need you didn't anticipate, a question turns out to be worded badly. Each change becomes another column in the spreadsheet, another custom field, another one-off form — and six years later the pieces still don't connect. Intake lives in one tool, the survey in another, the notes in a third, and no one can line up a single person's beginning against their end. When the configuration debt gets too heavy, teams do the thing that feels like relief and is actually the most backward move available: they start a clean spreadsheet and send one more email survey. The reset throws away the only thing that carries meaning — the thread connecting one person's data over time.

The experiment: one data-collection step at a time

The Loop asks for the opposite posture — an experimentation culture. You don't commit to a complete design; you run a small experiment, keep what works, and extend it. Start with the single data-collection step where you already get value. Customize that one workflow until it clearly earns its keep. Only then add the next step — and because everyone rides one persistent ID, the new step connects to what came before automatically. Nothing is thrown away, and you are never blocked waiting for the perfect form, because changing how you score re-reads everything already collected against the new standard.

Start with the step that already pays

Application or intake is almost always the right place to begin, because the value is immediate and self-funding. You already collect it; reading it on arrival — scoring against a rubric, classifying open text, flagging what needs a human — removes reviewer bias and saves real hours in week one. That early, obvious win is what earns you the room to add the next step. You are not asking anyone to trust a grand plan; you are showing them a working piece.

An experiment, not a rollout

The old model was a rollout: months of requirements, a long build, a dashboard delivered at the end — by which point the questions had moved and the team had never built the muscle to run it themselves. The Loop replaces that with an experiment the program owner can run directly. A leader at the Open Play Foundation is the clearest example: rather than waiting for an evaluation team, he started with one workflow, asked the assistant his own questions, and figured it out himself. He was surfacing insights daily and catching drop-off signals within weeks — and, just as important, he kept the capability in-house instead of renting it for the length of a project.

What you're actually building

Step by step, you are building a single connected record per stakeholder that gets richer every wave and never resets. That is the quiet payoff of the methodology: because you never throw work away, the system compounds. A year in, you are not staring at five disconnected exports — you have one queryable history per person, and the report becomes a question you ask rather than a project you survive.

Next: Principle 2 · Reliability: the same answer twice →  ·  Back to The Loop →

‍