Guidelines
Suggest edit

Compute aggregates at the task interval for average comparisons

For average-comparison tasks over grouped time intervals, use discrete per-interval aggregate encoding on time-series charts to improve fidelity and mitigate errors caused by continuous aggregation for viewers comparing grouped results.

  • purpose:refine
  • basis:empirical
  • task:compare
  • scope:grouped-result
  • time:time-interval
  • data:temporal
  • quality:fidelity
  • lever:encoding

advice

Discrete interval aggregates

Compute and display the summary at the same interval the reader must compare. For example, add one average bar per month in a composite graph instead of relying on a continuous moving-average line when the question asks which month has the highest average.

reason

Why task-aligned aggregation works

Average comparison becomes easier when the chart already matches the grouping used in the question. Readers no longer need to mentally segment a continuous summary into monthly chunks before comparing them.

Mechanism: Discrete per-interval summaries align the visual unit of comparison with the task unit of comparison, reducing the extra work of extracting monthly values from a continuous smoothing line.

Evidence: For average comparisons, encodings that discretely aggregated the data per month outperformed the remaining encodings, and the paper highlights composite graphs as better matched to this task than a continuous moving-average display (Albers et al., 2014).

context

Use when the interval is known in advance

  • User Goal: Decide which time interval has the highest average.
  • Task: Compare summaries across fixed blocks such as months.
  • Data: Quantitative time series with a known grouping level.
  • Chart Setting: A chart where the designer can compute interval summaries before display.
  • Success Criterion: Higher accuracy on interval-average comparisons.

exceptions

Do not use when the needed interval can change

Break it when: The reader’s comparison granularity is not known ahead of time or may shift across weeks, months, or other scopes. Why: The paper notes that discrete aggregation requires prior task knowledge and can misalign the chart with the reader’s actual scope.

costs

Costs of discrete aggregation

Sacrifice: You give up some flexibility for other temporal scopes and some raw local detail. Risk: A summary computed for months may not help with week-level questions. Mitigation: Keep the raw series visible underneath the interval summary when both local detail and interval averages matter.

mistakes

Common failure with this lever

Mistake: Add only a continuous moving-average line when the reader must compare averages for discrete time blocks. Why it fails: The reader still has to estimate where each block begins and ends and mentally recover a block-level average from a continuous curve.

check

How to test the choice

Failure Sign: Reviewers trace along a smoothing line instead of directly comparing one summary mark per interval. Quick Check: Show the same data once with per-interval averages and once with a moving average, then ask which interval has the highest average. Stronger Test: Repeat the A/B check across several datasets with different distractor intervals to confirm that the discrete version remains more accurate.

fix

What to change

  • Replace a continuous moving-average overlay with one summary mark per task interval.
  • Use bars or other discrete marks to encode the average for each interval.
  • Overlay the raw line beneath the discrete averages if readers still need the original series.

References

Albers, D., Correll, M., & Gleicher, M. (2014). Task-driven evaluation of aggregation in time series visualization. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 551–560. https://doi.org/10.1145/2556288.2557200