After a bad mock: error taxonomy, spaced returns, and stopping the shame spiral

1. Separate knowledge gaps from exam-craft gaps

Sort errors: blank retrieval, misread stem, method slip, time overrun, presentation. Each class needs a different next session. Drilling definitions does not fix time allocation; timing drills do not fix conceptual blanks. Taxonomy first — otherwise you “revise harder” in the wrong dimension.

2. Repeat the smallest unit that failed until it changes under time

This is the same wound-first logic as honest past-paper practice: full papers are for integration; micro-drills are for repair. Retrieval attempts with feedback are the lever (Karpicke & Roediger, 2008).

3. Space returns — sooner for fragile, wider for stable

You cannot optimise every interval by hand; you can obey directionality from spacing research: fragile knowledge needs shorter gaps initially; stabilised knowledge can tolerate longer gaps (Cepeda et al., 2006). A simple rule: calendar the next return before you feel fully fluent on the weak topic.

4. One full mock only after micro-skills stop collapsing

• → If timed sections still disintegrate, another full paper mostly remeasures the same wound.
• → Re-enter full papers when single-type performance holds under clock pressure.

What Offload aims to do (without implementation detail)

We want weak signals from attempts — mocks, questions, cards — to feed a schedule you can trust: what returns when, and how the week moves when something slips. We are not publishing scoring or scheduling internals here; the intent is to align product behaviour with retrieval and spacing evidence instead of leaving students to rebuild the plan from guilt.