Why Did Your Progress Score Drop Even Though You Improved?

Why did your progress score drop even though you improved?

Most likely, nothing got worse. What moved is noise, not you. Every measure, from a manager's rating to a stopwatch time to a sales figure, carries some built-in variation that has nothing to do with real change. Clinical and rehabilitation researchers call the threshold for telling the two apart the "minimal detectable change," the smallest difference that is bigger than the measure's own error (PMC, 2018). Below that threshold, a dip or a rise is simply the measure wobbling. Above it, something has probably actually shifted.

This matters for anyone tracking a professional goal, because a single bad week can look like failure when it is only measurement doing what measurement does.

What counts as measurement error, in practice

Measurement error comes from several ordinary sources, not from anything going wrong with you:

None of these are measuring whether you got better. They are measuring the measure.

A worked example: the founder watching conversion rate

A founder sets a goal to improve discovery-call-to-signed-contract conversion, tracked monthly. Month one: 22 per cent. Month two: 19 per cent. It feels like a step backward, and it is tempting to abandon the new call structure she has been practising.

But with roughly 15 calls a month, a swing of three points is well within the kind of sampling noise you would expect from such a small base, long before you get near anything resembling statistical confidence. The sensible response is not panic, and not celebration either. It is to keep the measure running for another cycle and look at the trend across three months rather than two adjacent points.

A worked example: the marksman and the scorecard

Competitive shooters know this problem intimately. A shooter's score can vary across a range session due to wind, light, fatigue or a slightly different stance, independent of any change in underlying skill. Coaches typically look at rolling averages across several sessions, not single scores, before concluding that technique has genuinely improved or declined. The same logic applies to a sales target, a 360 rating, or a writer's weekly word count: one data point is a sample, not a verdict.

A simple rule for deciding what a score change means

1. Note the size of the change (for example, a rating moved from 3 to 4, or conversion moved from 22 per cent to 19 per cent).
2. Ask what is known about that measure's typical noise. If nothing formal is known, treat anything less than roughly 10–15 per cent of the scale, or one check-in's normal swing on a numeric metric, as noise until proven otherwise.
3. If the change is inside that band, record it as held steady. This is a legitimate result, not a non-result.
4. If the change is clearly outside that band, and it repeats on the next check-in, treat it as real and worth a response, whether that is more practice hours, a change of approach, or a celebration.
5. Resist judging on a single reading. Two or three consistent check-ins carry more weight than one striking one.

Locke and Latham's long-running review of goal-setting research found that specific, measurable goals outperform vague ones precisely because they give you something concrete to track (Locke & Latham, 2002). A specific measure is only useful, though, if you also know when it is telling you something and when it is just breathing.

Where this leaves self-ratings

Self-ratings are useful for noticing effort and confidence, but as a sole measure of change they are among the noisiest instruments available, partly because the standard you are rating yourself against keeps moving as you learn. If your self-rating disagrees with harder evidence such as output, feedback or test scores, the evidence usually deserves more weight. We go into that disagreement in detail in why your self-rating doesn't match the evidence, and in how often to even check in at how often should you reassess a development goal.

How Jobulary handles this

Jobulary's check-ins take about three minutes, and growth is only reported as a gain when the change is larger than the measure's known error; otherwise the honest result shown is "held steady", for the reasons set out above and explored further in measuring professional growth honestly. Each live goal carries its own measure sized to a stated weekly hours budget, following a practice-first 70/20/10 path across 90, 180 and 365-day cycles, so a single noisy week never gets mistaken for a verdict on a year's work. The plan, including this history of held-steady and genuine-change results, belongs to the person and moves with them if they change employer; a sponsoring organisation never sees the reflections behind it.