KLM timing with log-normal correction

Description

Extends Keystroke-Level Model predictions from single point estimates to log-normal time distributions, giving predicted percentiles rather than only expert mean times.

The KLM predicts expert, error-free task time by summing operator times: K (keystroke, about 0.2-0.28 s), P (point, about 1.1 s or via Fitts), H (home hands, 0.4 s), M (mental preparation, 1.35 s) and R (system response). It yields a mean, but real operator times are right-skewed, and the summed task time is better described by a log-normal distribution. The correction assigns each operator a log-normal distribution with its KLM value as the median and an empirically set dispersion, then sums them by Monte Carlo simulation to produce predicted percentiles. This shows how often a task will exceed a time limit, which matters for time-critical work. Rigour requires calibrating dispersion from observed data and documenting M placement rules.

Attributes

Loss type
Latent
Stage
Design
Effort
Days

Steps

  1. Write the KLM operator sequence for each task variant using Card, Moran and Newell's M placement heuristics.
  2. Assign median times to operators (published values or local calibration) and a log-scale dispersion, typically sigma 0.3-0.5, from pilot data with 5-8 users.
  3. Simulate at least 10,000 task executions by sampling operator times and summing them.
  4. Report the median, 90th and 95th percentile task times and the probability of exceeding any time limit.
  5. Compare variants on both median and tail, and validate against observed task times; aim for prediction within about 20% of observed medians.
  6. Use the tail estimates to set design targets for time-critical tasks.

Sources

  • Goldstein, E. B., Cognitive Psychology: Connecting Mind, Research and Everyday Experience
  • Card, S., Moran, T. & Newell, A., The Psychology of Human-Computer Interaction (1983) and GOMS/KLM literature
  • Sauro, J. & Lewis, J. R., Quantifying the User Experience (2nd ed., 2016)
  • Cognitive Psychology (Goldstein)
  • GOMS/KLM literature
  • Quantifying the User Experience (Sauro & Lewis)
  • The Inquiry Series, Ch 4 — Time, Rhythm & the Three Timescales of Experience

Also known as

  • Keystroke-Level Model with distributional correction
  • Stochastic KLM

Categories

  • User Research
  • Task Analysis & Cognitive Modelling