Compare progress with a historical reference

The period immediately before this one is not the baseline that matters. Compare the block you want to understand with an earlier equivalent period and get one cross-domain assessment without rebuilding the deltas yourself.

Name the period you want to understand first

The agent needs two sides with different roles:

1. Primary period — the period you want to understand.
2. Historical reference — an earlier equivalent period used as context.

That order determines every delta and assessment. “Compare Q2 2026 with Q4 2025” means Q2 minus Q4. The agent does not flip the pair to make the dates or signs look more convenient.

If you only say “compare two quarters” or name more than two periods, the agent asks for the missing anchor instead of guessing.

Choose an equivalent reference

Natural periods compare with the same calendar unit: week with week, month with month, quarter with quarter, semester with semester, and year with year. Rolling windows use the same number of days. Exact ranges use the same inclusive duration.

Clean requests look like this:

Compare July 2026 with April 2026.
Compare the first half of 2026 with the first half of 2025.
Compare the last 30 days with the 30 days ending March 31.
Compare August 1–10 with June 5–14.

Full calendar months remain full calendar months, so July can be compared with June even though they contain a different number of days. The report uses coverage and comparable metrics to keep the reading honest; it does not cut July down to 30 days.

Keep the reference in the past

The reference must end before the primary period starts. The two windows cannot share even one day.

Valid:   Aug 1–10 compared with Jun 5–14
Invalid: Aug 1–10 compared with Aug 5–14
Invalid: Aug 1–10 compared with Sep 1–10

verxion rejects an overlap, future reference, incompatible unit, or mismatched duration. It does not swap, clip, or normalize the windows silently.

What the report brings together

get_progress_report resolves both periods and returns one canonical reading across training, nutrition, body, activity, and recorded qualitative evidence. The same operation produces:

• exact dates and labels for both periods;
• current and reference values;
• deltas in the fixed direction primary - reference;
• data coverage and source warnings;
• metric comparability;
• domain and overall assessments.

This matters when the broad question is “Was this block actually better?” A higher training volume may describe more work without proving progress. A body-weight increase can point in either direction depending on the goal. Sparse nutrition data can weaken the overall conclusion even when training coverage is complete. The report keeps those qualifications attached to the numbers.

Read the result without flattening it

Positive does not always mean better, and a real value does not always support a valid comparison. Pay attention to the status beside each delta:

• comparable — the reference supports a direct delta;
• descriptive_only — the value is useful, but the delta does not prove progress;
• segmented — a routine, plan, or engagement change requires a split reading;
• not_comparable — there is no equivalent reference for that metric;
• insufficient_data — there are not enough observations or no baseline.

The overall assessment can legitimately be mixed or insufficient_data. The agent preserves that conclusion instead of turning every report into a positive story.

Use a narrower comparison when the question is narrow

This recipe is for a holistic period review. If you only want to compare two executions of one lift, ask specifically for that exercise or session. If you want the broad story across your logbook, name the primary period and its earlier reference and let get_progress_report keep the dates, deltas, coverage, and assessment in one place.