How many check-ins before you can trust a trend?
You need at least three to four check-ins before treating a direction as real. One reading tells you almost nothing, because every measurement — a self-rating, a manager's note, a performance number — carries some error. Research on minimal detectable change shows that a single data point can shift up or down for reasons that have nothing to do with actual progress, so the smallest honest unit of evidence is a pattern across several readings, not one of them.
This matters because plans die for the wrong reason. Someone has a flat week, assumes the plan has failed, and abandons it before the noise has had a chance to settle.
Why a single check-in can mislead you
Every measure has a margin of error built in. A self-rating depends on mood, recent setbacks, and who you compared yourself to that morning. A quantitative measure — calls closed, words translated, sprint time — depends on conditions that vary day to day: sleep, client mix, weather, equipment.
The research on minimal detectable change, developed mostly in clinical and rehabilitation settings, makes a useful general point: unless a change is larger than the error range of the instrument, you cannot say it happened at all. The same logic applies to a development plan. If your measure naturally wobbles by, say, two points on a ten-point scale, a one-point dip after a single week is not evidence of decline. It is the instrument breathing.
This is also why Jobulary reports growth only when it is larger than measurement error, and says "held steady" otherwise. A flat reading is not a failure signal. It is often the correct, honest description of a short window.
A rough guide to how many readings you need
| Check-in frequency | Readings before a trend is trustworthy | Typical use |
|---|---|---|
| Weekly | 4–6 | Habit-level actions (e.g., calls made, reps completed) |
| Fortnightly | 3–4 | Skill practice with variable conditions |
| Monthly | 3 | Slower-moving capabilities (e.g., negotiation, writing) |
| Tied to a 90-day cycle | 1 cycle-end comparison, plus the one before it | Reassessing the goal itself |
These are working rules of thumb, not fixed thresholds — the right number depends on how noisy your particular measure is. The general principle from the measurement-error literature holds regardless: more readings reduce the influence of any single bad or good day, and a reassessment interval shorter than that risks mistaking noise for signal, something covered in more detail in how often should you reassess a professional development goal.
Two examples
A sprinter working on her block start tracks reaction time off the gun. One session she is 0.03 seconds slower — within the normal variation for that measure, affected by the surface, her warm-up, even the starting pistol's calibration. Her coach does not flag it after one session. After four sessions, if the times cluster consistently higher than before, that is a real signal worth addressing, not a bad Tuesday.
A founder tracking her own skill in running investor conversations rates herself after each meeting. The first meeting post-training feels like a step backwards — she stumbled over a question she usually handles well. Taken alone, that reading would suggest the training hasn't worked. Taken alongside the next three meetings, where her rating recovers and climbs past baseline, the single dip reveals itself as noise around a genuine improvement. This kind of mismatch between one reading and the fuller pattern is explored in why did your progress score drop even though you improved.
What to do while you wait for enough readings
- Keep the measure fixed. Changing how you score something mid-stream destroys the ability to compare readings at all.
- Log the conditions, briefly. A line noting "client was difficult" or "slept badly" helps you later tell noise from signal without relying on memory.
- Resist reacting to any single week. Locke and Latham's review of goal-setting research found that clear, specific goals with feedback improve performance over time — but the feedback needs to accumulate before it is acted on, not be treated as final after one reading.
- Set the next review date before you start, so you are not tempted to check and react daily.
This is also why a development plan benefits from a fixed horizon rather than constant tinkering — see measuring professional growth honestly for more on why "held steady" is itself a legitimate result, not a non-result.
Where this fits in a development plan
A Jobulary goal going live sets a measure and a first step up front, and check-ins of about three minutes track that measure across a 90, 180 or 365-day cycle. Growth is only reported once it clears measurement error — otherwise the plan correctly shows "held steady," which is a different thing from "not working." Because the plan belongs to the person rather than the employer, this pattern of readings stays with them even if they change jobs, so the trend line is never lost at a transition.
If you are deciding how to structure the goal itself before you start measuring it, how to write an Individual Development Plan that survives week six is a reasonable place to begin.
Frequently asked questions
Is one bad check-in a sign the plan isn't working?
Not on its own. A single reading includes normal measurement error. Wait for the next two or three check-ins before deciding whether the direction has actually changed.
What is minimal detectable change?
It is the smallest change in a measure that is larger than the error built into how you're measuring it. Research on minimal detectable change shows that readings below this threshold cannot be distinguished from noise (see sources).
Does a longer gap between check-ins make a trend more trustworthy?
Up to a point. A 90-day cycle usually gives more reliable signal than a weekly one, because short-term swings average out. Beyond 365 days, though, you risk missing a plan that has quietly gone off course.
Sources
Topics: Measurement · Individual Development Plan · Goal setting