Prescribe Stage (Forward-Thinking Loop)

Description

Stage 5 of the forward-thinking loop: translate hazard ratios, SHAP drivers and quotes into mechanism-linked fixes placed on an effort × impact grid, each with a decision bar.

Prescribing turns statistical drivers and qualitative mechanisms into actionable fixes. The researcher reviews top quant drivers (SNR, mistake rate, retries), cross-checks them with qual evidence, drafts prescriptions in a 'because → therefore' structure, places each on an effort × impact grid and attaches decision bars. An evidence ⇒ prescription table records problem area, quant driver, qual mechanism, design prescription, effort and impact. QCA sufficiency results are used to bundle fixes: if Low SNR * High Mistakes is sufficient for failure, labelling and SNR uplift ship together. Mechanism memos and one quote per fix keep the mechanism salient in design review. Rigour comes from refusing insights without fixes, quantifying effort, and binding each fix to a threshold such as tree success ≥ 85%, OTP success ≥ 80% within 3 attempts, TLX frustration ≤ 40/100 and end-to-end completion ≥ 60%.

Attributes

Loss type
Community-proposed
Effort
Days

Steps

  1. Review the top quantitative drivers from Prove and Forecast (e.g. HR per mistake = 1.31, first-click 74% vs 49%, OTP 65% → 80% predicts +5 points).
  2. Cross-check each driver with qualitative evidence to name its mechanism (e.g. 'I looked under Notifications, not Markets').
  3. Draft prescriptions in 'because → therefore' form and record them in an evidence ⇒ prescription table with effort and impact ratings.
  4. Place fixes on an effort × impact grid: low effort/high impact now, high effort/high impact on the roadmap, low/low opportunistic, high effort/low impact deferred.
  5. Bundle fixes that QCA shows are jointly sufficient (do not ship labelling without SNR uplift if Low SNR * High Mistakes drives failure).
  6. Attach a decision bar to each prescription, e.g. first-click ≥ 75%, tree direct ≥ 85%, HR ≤ 0.80, OTP success ≥ 80%, TLX frustration ≤ 40/100.
  7. Attach one representative quote per fix for design review.

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

Also known as

  • Stage 5: Prescribe, from evidence to design action
  • Evidence ⇒ Prescription mapping

Categories

  • Community-Proposed Frameworks