OKR Progress Tracker

What gets scored, and what does not

An objective is a sentence about a state you want to be true. Key results are the numbers that decide whether it is. The objective itself has no score — there is no way to compute one — so what any tracker reports is an aggregate of the key results hanging off it, treated as a proxy. That is the whole of what this does. Whether those particular numbers actually represent the objective is a judgement no calculation reaches, and it is the judgement that decides whether the score means anything.

The formula, and why not the obvious one

Each key result scores (current − baseline) ÷ (target − baseline) × 100. The tempting alternative — current divided by target — inflates everything that did not start at zero. Moving retention from 32% to a target of 50% with 41% on the board today reads as 82% attained on the naive formula, when you have in fact covered half the distance you set out to cover. Half is the honest answer.

Decreasing metrics need no special handling. When the target is below the baseline, numerator and denominator both go negative and cancel: response time going 45 → 15 minutes, currently at 21, gives (21−45)/(15−45) = 80%. There is no direction toggle in the form because the numbers already say which way the metric is meant to move.

Weights, caps and the awkward edges

Weights are relative, not percentages. A key result at 3 counts three times one at 1; they do not need to total anything. Omit the weight and it is 1. Set every weight to zero and there is nothing to divide by, so the tool falls back to an unweighted mean and says so rather than returning a silent NaN.

Baseline equal to target is the other degenerate case: no distance, no ratio. Those score 100% when current matches target and 0% when it does not. Overshoot is capped at 100% for the average by default, which stops one metric at 210% carrying three that are at 30%; the table and bars always show the uncaught real figure. Turn the cap off if you specifically want overshoot to count, and read the table before quoting the headline. Backsliding — current worse than baseline — shows as a negative in the table and floors at 0% in the average.

Against the clock

Give the tool a period start and end and it works out what fraction of the period has elapsed as of your reference date, then compares. Halfway through the quarter, 50% is the neutral mark. More than three points above reads as ahead, more than three below as behind, and the band between as level. The threshold is a display convention, nothing more.

The comparison assumes progress accrues evenly with time, which plenty of work does not. Anything front-loaded with setup shows as behind for weeks and then moves all at once; anything where the easy wins came first shows as ahead and then stalls. Read the direction the gap is moving week over week, measured the same way each time, rather than the level on any one day.

Two things about scores in practice

First, the scale. This page reports 0–100%. A lot of organisations score on 0–1.0, where 0.7 is often treated as a good outcome for a deliberately ambitious objective, and others use red/amber/green bands with their own cutoffs. These are conventions, not standards, and they disagree with one another — 70% is a strong result under one and a failure under another. Whatever your organisation uses is what governs; convert this figure into it rather than presenting it raw and letting people apply the wrong scale.

Second, the failure mode everyone eventually meets. If scores feed performance reviews, people write key results they are confident of hitting, targets get set just past where the metric was already heading, and every objective lands near 100% while nothing changes. At that point the number is measuring the target-setting, not the work. The tell is a full set of key results clustered at the top with no misses anywhere — a slate that never scores badly is a slate that was never asking for anything.

Questions people ask