Methodology

A matched pair is one person's Day 0 submission and Week 10 submission, joined by the private code they created (their birthday and their parents' initials — nothing that identifies them). Only complete pairs are used to compute the Day 0 → Week 10 change figures — a person who submitted only Day 0 contributes to the funnel above but not to any change measure.

A person is eligibleonce their Day 0 submission is ten or more weeks old — they've had the full window to also submit Week 10, whether or not they did. A person is in progressif they have a Day 0 submission but ten weeks hasn't elapsed yet. Every person with a Day 0 baseline is exactly one or the other.

The completion rate shown on the results page divides finished pairs by the eligible count — never by everyone who ever started. Counting people who haven't had the chance to finish yet as if they'd failed to would measure the calendar, not the program.

Each of the nine behavioral domains is scored from its two questions per submission, as a percentage of the maximum possible score (0–100%). A domain's Day 0 or Week 10 average on the results page is the group mean of that percentage across everyone in a matched pair, at that time point. A couple of items are worded in the opposite direction on purpose, and scored accordingly, so a higher raw answer always means more of the trait in the final percentage.

For each finished pair, we take that person's own average change across the nine domain scores, Week 10 minus Day 0. The threshold for each group is:

• Improved: +5 points or more
• About the same: less than ±5 points
• Declined: −5 points or less

Every finisher is shown in one of these three groups, not just an average that could hide who moved which way.

Two of the three life-anchor questions (life satisfaction, confidence in the next 12 months) are answered on a 0–10 scale and shown in that same unit. The mornings question is answered on the same 0–10 scale, for consistency with every other question in the instrument, but the question itself defines a proportional mapping onto real mornings out of 14 (a response of 10 means 14 of the last 14 mornings) — the results page converts to that everyday unit for display. The relative percent change is unaffected by this conversion.

A submission that shows a clear pattern of low-effort, same-answer clicking is excluded from every aggregate figure on this site. The raw submission itself is never altered or deleted — it stays in the permanent record, just outside the group calculations. The exact detection rule isn't published, so it can't be gamed.

Our own system tests submit real entries through the live form, using identifiers that no participant can have — email addresses on the domains reserved for testing (example.com, example.net, example.org) or a fixed test code. Those entries are left out of every figure on this site. Like every other submission they are never altered or deleted, and they remain part of the fingerprinted raw dataset.

Individual answers, ever. Private codes exist solely to match a person's Day 0 submission to their Week 10 submission and are never shown or used to contact anyone — there is no name, email, or contact detail to use.

This page covers the concepts, not the exact formulas, thresholds, and edge-case rules the nightly build runs — that level of detail, along with the underlying dataset itself, is available on request, at Growth Impact Group's discretion. Contact growthimpactgroup@protonmail.com. For automated verification without contacting us, see /verify.json.