Work Measurement & Learning Lab.
A stopwatch says a job takes about five minutes. So how much time should you plan for it, and how many people do you need for the day’s orders? People also get faster with practice, then lose some of that speed in the days between repeats. Change each setting and watch three things move: the planned time, the crew size and the time per job after 100 repeats.
How long a job takes, with practice and breaks
With a 1-day gap, 92% of the practice built up so far survives each wait. After 100 repeats, the practice left is worth 14.0 repeats, not 100. That’s close to the most this gap allows (14.0), so the time per job has levelled off. The 100th job takes 3.33 min, 1.23 min (59%) longer than with no gap. Compare the 7-day and 30-day cards below.
Timing the job
Demand
Practice and forgetting
Pace rating: 100% is a typical pace. Rating the timed worker above 100% says they were quick, so the planned time gets stretched to suit a typical worker. The allowance is the share of planned time set aside for rest and delays nobody can avoid. With no gap, a 85% learning curve works like this. Every time the repeats double, the time per job becomes 85% of what it was.
The forgetting rate is a “what if” setting, not a measured fact. Before planning with it, tune it to records of real people doing the task. Click the surface above to pick a gap and a repeat. The repeat axis uses a log scale, so 1 to 10 gets as much room as 10 to 100.
Time for the 100th repeat, by gap
85% learning curve · 1-day gap · 92% of practice kept per gap · repeat 100 is 1.23 min slower than with no gap
Time per job at the selected gap, in minutes
| Gap | Practice | Time | % of repeat 1 | vs no gap |
|---|---|---|---|---|
| 1-day gap | 14.00 | 3.33 min | 53.9% | +1.23 |
| no gap | 100 | 2.10 min | 34.0% | — |
From stopwatch to planned time
normal = 5.00 × 105% = 5.25 min
standard = 5.25 ÷ (1 − 0.15) = 6.18 min
Observed is the stopwatch average. Normal adjusts it to a typical worker’s pace. Standard adds the allowance for rest and delays. That’s the time used for planning.
The ten stopwatch readings
spread: standard deviation 0.24 min (4.7% of the average). That’s about how far a typical reading sits from the average. Bars start at zero on a 0–6 min scale.
Planned time vs takt · crew size
takt = 480 ÷ 80 = 6.00 min
6.18 min work ÷ 6.00 min takt = 1.03 → 2 people
each person busy = 1.03 ÷ 2 = 51%
At this planned time, demand from 78 to 155 units a day needs 2 people. At 156 a day, it takes 3. Takt is the pace the day’s orders set: one finished unit every so many minutes. More orders shorten the takt, but the job itself takes just as long. Pace rating and allowance change the planned time, which can push the crew up or down by a person. Crew size uses the planned (standard) time, not the faster times people reach with practice.
How the model works, how it’s checked and what it leaves out
How it works
The ten timed cycles average 5.00 min. That average times the pace rating gives the normal time. Normal time is then divided by (1 − allowance), with the allowance counted as a share of the total time allowed. Takt is 480 ÷ daily demand. Crew size is ceil(standard ÷ takt), where ceil means round up to a whole person. It’s simple math on the average workload, with no waiting lines and no ups and downs.
Learning follows the unit curve Tₙ = T₁·nᵇ, which tracks the time for each single job. Here b = log r ÷ log 2, where r is the learning curve as a decimal (0.9 for 90%), and T₁ = standard time. Gaps are tracked with one number, experience E. Before each repeat, the experience above a total beginner’s level gets multiplied by R = exp(−k·gap), a number from 0 to 1 (k is the forgetting rate). Then the repeat adds one. So Eₙ = 1 + (Eₙ₋₁ − 1)·R + 1, and Tₙ = standard·Eₙᵇ.
The surface works this out for every whole-day gap from 0 to 30 and n = 1 to 100. It uses the current rating, allowance, learning rate and k.
How it’s checked
k = 0 or a zero gap gives R = 1 and Eₙ = n exactly, so the selected-gap curve matches the classic curve. A 100% curve gives b = 0 and a flat surface at the standard time. Tₙ never drops when the gap or k goes up, so the surface never slopes down toward the back.
Eₙ matches the exact formula 1 + (1 − Rⁿ⁻¹) ÷ (1 − R), and never goes above 1 + 1 ÷ (1 − R). Rating and allowance scale every Tₙ by the same amount, so they don’t change the shape of the surface.
What it leaves out
The forgetting rate is an example “what if” setting, not a fixed fact about human memory. How much people keep depends on the task, how well they knew it before, their tools and the cues around them.
Not included: a separate long-term memory that never fades, and faster relearning after long breaks. Also uneven gaps, differences between workers, and learning curves based on the running average of all jobs so far. And any effect of learning on crew size.