The short answer
Compare matching date windows, sources and metric definitions. A complete prior year versus an incomplete current year is not a fair growth calculation. If either report has limited collection, describe the observed samples instead of claiming an account-wide trend.
What you’ll take away
- Match the window before calculating a percentage.
- Keep each edition and its source evidence intact.
- A change describes a difference; it does not explain its cause.
Start with two comparable windows
Suppose your saved 2025 recap covers January through December, while your 2026 recap stops on October 31. Comparing their totals silently gives one year two extra months. You can show both totals with clear scope, but you cannot call the difference a like-for-like annual change.
For a fairer review, use January–October in both years from a source that actually contains those periods. Do not assume a recap interface can reconstruct a filtered historical export. Wrapped.li preserves supported saved editions; that is not a promise of a complete year-over-year analytics dataset.
Fill this worksheet before doing the math
Keep this table next to the reports. A mismatch is a reason to qualify or stop the comparison, not a reason to quietly normalize unknown values.
| Check | Period A | Period B |
|---|---|---|
| Start and cutoff | Record actual dates | Use the same window if available |
| Source | Public collection or owner analytics | Same source and access |
| Collected posts | Count plus completeness | Count plus completeness |
| Metric | Exact numerator and denominator | Unchanged definition |
| Observation age | When counts were collected | Comparable age where possible |
Worked example: more output, lower typical response
Imagine two complete, equally scoped samples. Year A has 20 posts, a median of 10 reactions plus comments and a total of 260. Year B has 40 posts, a median of 8 and a total of 420. Total response rose by about 61.5%, while typical post response fell by 20%. Both statements can be true.
The useful conclusion depends on your question. For publishing volume, 40 versus 20 matters. For the typical observed conversation, 8 versus 10 matters. Neither establishes more customers, better writing or algorithmic preference. Record those as separate questions.
Handle zero and missing values differently
If a real baseline is zero, ordinary percentage growth is undefined because the formula divides by zero. Report the absolute change: “From 0 to 5 observed comments.” If the baseline was never collected, even that sentence is unsupported.
A partial sample also complicates per-post averages. Collecting only recent or popular posts can change the distribution. Dividing by the number collected does not magically remove that selection effect. Say what was observed and retain the limitation.
Write the conclusion before the celebration
Use a sentence with four parts: window, source, result and limitation. For example: “Across complete January–October exports, I published twice as often and received more total responses, while median response per post was lower. This comparison does not isolate the effect of format or timing.”
Keep the source files or saved editions unchanged. Then choose one question for the next period, such as whether a smaller number of deeper explanations creates better conversations. A fair review should help you choose what to do next, not just manufacture a larger percentage.
Sources & product references
Product details reflect the publication date. External features and interfaces may change.
