Description
Stage 1 of the forward-thinking loop: pair CPM-GOMS time predictions with mental-model interviews to produce falsifiable hypotheses and decision bars before any testing.
Prediction front-loads insight: before recruiting participants, the researcher models task structure and estimates where bottlenecks will occur. The quant lens is CPM-GOMS with operator estimates (cognitive about 1.2s, perceptual about 0.4s, motor 0.2–0.3s) summed along the task to predict ideal completion time and the cost of misroutes. The qual lens is 4–8 quick contextual interviews asking where users would look first and why, sketched as an 'expected IA' and compared with the actual IA. Combined, they produce a hypothesis such as 'users will misroute to Notifications when creating Alerts, leading to 2–3× longer task time than predicted'. Rigour comes from setting a decision bar up front and logging qual with the first two Miles & Huberman steps (compiling, disassembling) plus a recursive evidence log of expected loop points. The stage is explicitly hypothesis generation, not proof.
Attributes
- Discipline
- Community-Proposed Frameworks
- Loss type
- Community-proposed
- Tested
- Not yet tested
- Stage
- Design
- Data
- Mixed
- Effort
- Days
Steps
- Decompose the target task into CPM-GOMS operators: cognitive ≈ 1.2s, perceptual ≈ 0.4s, motor ≈ 0.2–0.3s (e.g. 5 characters × 0.2s = 1s of typing).
- Sum the operators to predict ideal completion time (≈ 5.5–6.0s for the price-alert example) and estimate added time per misroute (two wrong tabs add 6–10s).
- Set a decision bar up front, e.g. predicted ideal 6s; an observed median above 15s is classified as a significant breakdown.
- Run 4–8 contextual interviews asking 'If you wanted to set a price alert, where would you look first? Why?' and sketch each participant's expected IA.
- Compile verbatims and sketches into one worksheet (participant, exact quote, expected-IA notes, initial hunch) and code each utterance with short low-granularity labels.
- Keep a recursive evidence log: capture the described order (e.g. Settings → Back → Markets), mark expected loop points and flag them as risk loops.
- Triangulate: compare CPM-GOMS time if the expected location is correct against added misroute time, and write a falsifiable hypothesis for Stage 2.
Sources
- Vivek Gopinath (2025), Forward-Thinking UX Research Playbook: A Mixed-Methods, Decision-Ready Approach from Prediction to Validation
- Vivek Gopinath (2025), Tree Testing and Error Taxonomy: A Complete Chapter for UX Researchers
- Bonnie E. John & Wayne D. Gray (1995), CPM-GOMS: An Analysis Method for Tasks with Parallel Activities, CHI '95 Conference Companion
Also known as
- Stage 1: Predict with CPM-GOMS and Mental Models
- Predictive task modelling
Categories
- Community-Proposed Frameworks