The Follow-Through Index

By Adrian Perry, Founder, FocusBro · methodology updated September 2026

The Follow-Through Index is FocusBro's own answer to a question the productivity world mostly guesses at: when someone says they will start something at a certain time, and something checks in with them at that time, how often do they actually start? It is computed from FocusBro's first-party ledger, published as an aggregate with its sample size shown, and regenerated hourly. This page is the methodology. The figures at the bottom are the Index as it stands right now.

What is being measured

Every number on this page comes from one record: the accountability loop. A person tells FocusBro what they will do and when. At that time FocusBro checks in. The person answers — they started, they moved it, or they did not — and the answer is written down. The Index is those answers, counted.

The one research claim behind the product is that naming when and where you will act makes acting more likely. That is Peter Gollwitzer's finding on implementation intentions, replicated across a large body of studies; the sources are listed below. The Index does not test that claim. It reports how the loop behaves for the people who use it.

Definitions

Commitment

A person told FocusBro what they would do and when they would start. Timers on their own are not commitments; a Pomodoro with no word given is not counted anywhere here.

Check-in

FocusBro asking, at the committed time, whether they started. A check-in being delivered is FocusBro acting, not the person. It never counts as the person doing anything.

Kept

The person answered that they started.

Rescheduled

The person moved the commitment to a new time. This is a real, protected outcome. It counts in the denominator — the word was not kept this time — but it is never recorded or shown as a miss.

Missed

The person answered that they did not start, or the check-in ran out without an answer after every follow-up. Counted only so the rate is honest; never surfaced as a tally anywhere in the product.

"I'm on it"

The third answer: not yet started, not moving it, keep the word open a little longer. It is engagement, not a resolution. It is counted on its own and deliberately kept out of every rate on this page.

Released

The person withdrew the commitment. Not resolved, not counted.

Resolved

Kept + rescheduled + missed. The denominator, and the sample size (N) the page shows.

The measures

Kept-word rate

kept ÷ (kept + rescheduled + missed), over the window. This is the Index's headline number.

Reschedule rate

rescheduled ÷ (kept + rescheduled + missed). Shown beside the headline so a high kept-word rate can never hide how much of the rest was moved rather than missed.

7-day return

Of the people whose first-ever activity fell in the window and who have had seven full days since, the share with activity of their own on a later day within those seven. A delivered check-in, a follow-up, or a "come back" nudge is FocusBro acting and does not count as returning. This measure has its own floor.

Publication rules

• Window. The last 30 days, rolling. The window's dates are printed with the figures.
• Floor. A rate is published only when the window holds at least 30 resolved commitments. Below that, the page says "fewer than 30" and nothing else — not the count, not a percentage. The 7-day return needs at least 30 eligible people.
• Precision. Whole percentages. Nothing here supports a decimal.
• Freshness. Figures are computed from the ledger and cached for an hour; the "Generated" stamp is the computation time, in UTC. Nothing on this page is typed in by hand.
• Same data, one place. The figures and the JSON at /api/public/follow-through are produced by the same function from the same rows, so they cannot disagree.
• Definitions live in the open. The counting is done by the code that records the events, which anyone can read. If a definition on this page ever differs from the code, the code is wrong or this page is, and either is a bug.

Privacy

The Index is an aggregate. No figure on this page or in the endpoint refers to a person; nothing here can be joined back to one. Below the floor even the count is withheld, so a small early cohort is never reduced to "three people, one of whom…". FocusBro does not record a per-person miss count anywhere, for the Index or for anything else.

Limits — read these before citing

• Self-selected. These are people who chose an accountability tool and gave their word to it. They are not a sample of anyone else.
• One product. The loop's wording, timing, and follow-ups shape the answers. A different check-in would get different numbers.
• Survivorship. People who stop using FocusBro stop being counted. A window can look better because the people it was not working for left.
• A reschedule is not a failure. Reading the reschedule rate as "how often people flake" is a misreading; it is how often a word was moved rather than kept or missed.
• Not a clinical measure. The Index says nothing about ADHD prevalence, diagnosis, or treatment, and it must not be cited as if it did.
• Early. N is small in September 2026. Until the floor is met the page publishes no rate at all, and for a while after that the rate will move a lot from month to month. Cite the window and N with the number, always.

How to cite it

FocusBro. The Follow-Through Index, window <dates as printed>, N = <as printed>. Retrieved <date> from https://focusbro.net/follow-through-index.html. The JSON at https://focusbro.net/api/public/follow-through carries the same figures with the window, floors, and generation time as fields.