HOW TOCKLOG'S AI SCORES YOUR PRODUCTIVITY

A number that grades your week is only useful if you know how it was produced. Opaque scores invite one of two failures: you either trust them more than they deserve, or you dismiss them entirely. So here is the actual mechanism, without the marketing gloss.

The first thing to know is that TockLog shows you two different scores, produced in two completely different ways. People conflate them constantly, so we will separate them properly.

SCORE ONE: THE LETTER GRADE (A FORMULA, NOT AN AI)

The large number and letter at the top of the Analysis tab are not generated by a language model. They are computed on your device from a fixed, published formula over the current week's sessions. Three components, weighted:

• Consistency — 40 points. The share of elapsed days this week on which you logged anything at all.
• Volume — 35 points. Total tracked minutes measured against a 30-hour weekly target. Thirty hours caps the component; more does not score higher.
• Diversity — 25 points. How many distinct categories you used, against a target of four.

The three are summed and clamped to 0–100. The letter follows on fixed thresholds: 90 and above is A+, 80 is A, 70 is B+, 60 is B, 50 is C, 40 is D, and below that is F. Nothing is rounded in your favour and the boundaries do not move.

Because it is arithmetic rather than a model, it is fully predictable. The same week of sessions always produces the same grade, and you can work out in advance what would move it.

The grade is not an opinion about your work. It is a measurement of how consistently, how much, and how broadly you logged — and nothing else.

WHAT THAT FORMULA DELIBERATELY IGNORES

Read the weights again and the blind spots are obvious. Consistency is worth more than volume, so five short days beat two long ones. Volume stops rewarding you at thirty hours, so a ninety-hour week scores no better than a thirty-hour one. And a quarter of the score is category spread, which means a week of nothing but deep Work is capped below a balanced one.

Those are opinions baked into the design, and reasonable people would weight them differently. We would rather show you the weights than pretend the number is objective.

SCORE TWO: THE DAILY AI REPORT

The daily report is the part that actually uses a model. At report time TockLog sends the day's logged sessions — title, category and time window, the same information visible in your own history — for analysis. It has no access to your screen, your keystrokes, your location or your files. It reads the log you wrote.

What comes back is structured rather than prose:

• A daily productivity score and a consistency score, each 0–100. These are the model's own numbers and are shown separately inside the report — they are not the letter grade above.
• Category time and percentage splits.
• Peak focus hours, and the low-productivity zones opposite them.
• Time leaks — activities that consumed more of the day than they appear to have earned.
• Key insights, habit patterns, distraction patterns, and actionable suggestions.

These are model outputs, so unlike the formula they are not perfectly reproducible. Two similar days can land a few points apart.

THE PART THAT IS ACTUALLY USEFUL

In our view both scores are the least interesting thing on the screen. They are single scalars, and single scalars are easy to game and easy to obsess over.

The patterns are where the value sits. Peak focus hours tell you when to schedule the work that matters. Time leaks name the specific recurring activity quietly eating your week. Habit and distraction patterns only emerge across many days, which is exactly the kind of observation nobody makes reliably from memory.

If you read one thing in the daily report, read the suggestions and the leaks. If you find yourself checking the number first and closing the app, the feature is working against you.

LIMITS WE ARE NOT GOING TO HIDE

• Neither score can verify your log. Record a two-hour deep work session you did not do and both will score the fiction.
• Neither has any notion of the importance of your work — only its duration, category and distribution.
• The letter grade is a weekly measure. Judging a single day by it will mislead you, especially early in the week when few days have elapsed.
• The AI's pattern analysis needs weeks of history, not days, before the trend fields say anything real.

None of that makes the report useless. It makes it a measuring instrument with a known error bar, which is the only honest way to present a number about something as complicated as a working week.