Keystroke-Level Model (KLM)

Description

Predicts expert task time by summing standard operator times for keystrokes, pointing, homing, mental preparation and system response.

KLM is the simplest member of the GOMS family. A task is written as a sequence of operators: K (keystroke, about 0.2-0.28 s), P (point with mouse, about 1.1 s), H (home hands between devices, about 0.4 s), M (mental preparation, about 1.35 s), B (button press) and R (system response). Heuristic rules specify where M operators are placed. Summing the sequence gives a prediction for an expert, error-free performer. The treatise's worked example, copying a paragraph between documents ([H mouse][P paragraph][K Ctrl][K C][P destination][K Ctrl][K V]), yields roughly 3.8 s with 1983 parameters. With modern validation, predictions usually fall within 20-30% of measured times, an accuracy few contemporary UX methods approach, and the model needs no user testing to produce a falsifiable prediction.

Attributes

Loss type
Extinct
Stage
Design
Effort
Hours

Steps

  1. Define the task and the design alternative precisely, including the starting state.
  2. Write the physical operator sequence (K, P, H, B) for the expert method.
  3. Insert M operators using the Card-Moran-Newell placement heuristics.
  4. Add R for system response times where the user must wait.
  5. Sum operator times, using updated parameters for touch or mobile where relevant (for example Fitts-based P).
  6. Compare alternatives and, where stakes are high, validate against timed expert trials.

Sources

  • Lost Methods in UX — A Complete Treatise (Gopinath)
  • Card SK, Moran TP & Newell A (1980) The keystroke-level model for user performance time with interactive systems. CACM 23(7)
  • Kieras D (2001) Using the Keystroke-Level Model to Estimate Execution Times

Also known as

  • KLM-GOMS
  • Keystroke Level Modeling

Categories

  • User Research
  • Task Analysis & Cognitive Modelling