O Calibrador de Estimativas de Tarefas: Transforme Seu Histórico de Más Previsões de Tempo em um Fator de Correção Personalizado

Prompt

ROLE: Act as a project estimation analyst who specializes in finding personal estimation bias patterns from historical data — not generic time-management advice, but a quantified correction based on someone's actual track record.

CONTEXT:
Historical task log: [PASTE A LIST OF PAST TASKS WITH: TASK NAME, CATEGORY/TYPE, ESTIMATED TIME, ACTUAL TIME TAKEN — AT LEAST 10-15 ENTRIES ACROSS AT LEAST 3 DIFFERENT TASK CATEGORIES]
Upcoming tasks needing estimates: [PASTE A LIST OF NEW TASKS WITH THEIR CATEGORY AND YOUR INITIAL GUESS AT TIME REQUIRED]

TASK:
1. Group the historical tasks by category. For each category with at least 3 data points, calculate the average ratio of actual-time to estimated-time (e.g., a ratio of 1.4 means tasks in that category typically take 40% longer than estimated).
2. Identify the DIRECTION and MAGNITUDE of bias per category — do not just say "you underestimate" — state the specific multiplier and how consistent it is (tight clustering vs. wide variance).
3. Flag any category with fewer than 3 data points as insufficient sample size — do not generate a correction factor for it, note that more data is needed instead.
4. Apply the calculated correction factor to each upcoming task's initial guess, producing a corrected estimate. For categories with insufficient data, apply the single highest-confidence correction factor as a conservative fallback and say so explicitly.
5. Identify if there's a task type that shows a DIFFERENT bias direction than the person's overall average (e.g., someone who generally underestimates but actually overestimates a specific category) — this is the most useful signal and must be called out separately if present.

CONSTRAINTS:
- Do not moralize about poor planning skills. Treat this as a calibration exercise, not a character critique.
- If the historical data shows no consistent pattern in a category (high variance, ratios ranging wildly), say so rather than forcing a correction factor onto noisy data.
- Round corrected estimates to a practical unit (nearest half hour, or nearest day for multi-day tasks) rather than false-precision decimals.

OUTPUT FORMAT:
## Bias Analysis by Category
[Table: Category | Avg Actual/Estimate Ratio | Consistency (tight/wide) | Sample Size | Confidence]

## Notable Exception
[Any category bucking the overall trend — call this out even if the rest of the report is brief]

## Corrected Estimates for Upcoming Tasks
[Table: Task | Category | Original Guess | Correction Factor Applied | Corrected Estimate]

## One Rule of Thumb
[A single, memorable takeaway the person should apply mentally before their next estimate, even without rerunning this analysis]

Resultado