Return to the decision made before release.
Twenty-seven per cent of the views in Masoom’s 2026 comparison came from older episodes. If we credited that whole total to new releases, we would be reviewing the wrong body of work. Start by identifying what the result contains, then return to the decision you made before publication.
Separate three questions: what happened, what might explain it and what to do next. Numbers answer parts of the first. The content, audience and production context help with the second. The third requires an editorial choice whose uncertainty remains visible.
Define what the total contains.
The September 2026 Masoom case compares the show’s long-form catalogue from 1 January to 3 September in 2025 and 2026. The 2026 period recorded 4.66 million views. Of those, 1.26 million—27%—came from older episodes. The 35 episodes released in the current-year period contributed 3.40 million.
That split changes the interpretation. The total describes the audience’s use of the catalogue during the period. It cannot all be assigned to the new releases. Older work is still part of what the show is delivering, and a review of only the latest episodes would miss a substantial share of the viewing.
Before interpreting a release report, write down which question you are answering. How much viewing did the catalogue generate this year? How did newly published work contribute? How did two episodes perform in their first seven days? Each can be useful, but the comparison needs the corresponding set of videos and measurement window.
The Masoom comparison includes sponsored episodes and older long-form releases, and excludes personal uploads and Shorts. It records show outcomes from the combined work of Masoom and the team. It is not an equal-age comparison of individual videos and does not isolate the effect of a particular editorial change.
The exercise below addresses the separate equal-age question with invented data whose arithmetic can be followed in full.
An unequal-age comparison gives the wrong starting point.
Entirely invented teaching dataset. Video V01 has 80,000 views at thirty days. V02 has 18,000 at seven days. Calling V01 the stronger first-week release from those two totals would be wrong. Use the same measurement window before drawing that comparison.
| Release | Duration | First 7 days: views | First 7 days: average viewed |
|---|---|---|---|
| V01 · What the red patch measures | 6:00 | 12,000 | 3:48 · 63.3% |
| V02 · Where the shade moves | 6:00 | 18,000 | 3:18 · 55.0% |
| V03 · Who maintains the tree | 8:00 | 9,000 | 4:48 · 60.0% |
In this equal-age view, V02 has more views than V01, but a lower average viewing duration. V03 has fewer views and a longer average viewing duration; its longer runtime matters to that interpretation. The correct comparison is more informative than a single winner label.
Check the arithmetic and the context.
| Release | Views × average viewing seconds ÷ 3,600 | Watch hours |
|---|---|---|
| V01 | 12,000 × 228 ÷ 3,600 | 760 |
| V02 | 18,000 × 198 ÷ 3,600 | 990 |
| V03 | 9,000 × 288 ÷ 3,600 | 720 |
These aggregate teaching numbers reconcile exactly by construction. In real exports, rounding, different filters or different report definitions may create small differences; investigate material mismatches before analysing them. Do not merge snapshots with incompatible timezones or cutoffs.
V02’s synthetic external-traffic share is 22%, compared with 8% for V01. That difference may be relevant to who watched, but it does not establish that external traffic caused the retention difference. The topics also differ. This is an observational comparison, not a controlled test of one creative change.
Measurement update · 9 September 2026. YouTube changed public view counting on 24 August to count playback starts across formats. Before comparing periods across that date, record the exact report, metric definition and export date. Reconcile a comparable series where possible; otherwise label the break. The field name “views” alone does not establish comparability. YouTube’s explanation and our analysis of what the change means explain the distinction.
A complete postmortem decision.
The review record
Original intention: test whether a concrete observation can give unfamiliar viewers an accessible entry into a city-systems explanation.
What happened: V02 reached more viewers in the first seven days and generated more total watch time than V01, while its average viewing duration was lower. Its traffic mix differed.
What remains unknown: whether the difference came from topic, packaging, opening, audience mix or another factor. The totals cannot isolate those effects.
Decision: develop a comparable next explainer with a concrete opening, then inspect the transition into the main explanation. Keep the promise accurate and log the actual title, opening and distribution treatment.
Owner and next evidence: the editor records the change before publication; the analyst reviews the same first-seven-day window and relevant audience segments after data are available.
Include the work needed to make the result.
A release can perform well while depending on unsustainable access, overtime or revision. If those costs were measured, include them alongside the content result. If they were not, record the gap rather than inventing an efficiency improvement from a smoother-looking process.
Capture what the team learned about the brief, research, recording, edit and approval sequence. A missed answer may explain an awkward middle better than a stylistic critique. A late premise change may have consumed the time intended for packaging. Give each repair an owner and connect it to the next production plan.
Keep the conversation about decisions and conditions the team can change. A postmortem that ends with “try harder” has not produced a useful operating lesson.
Keep one record that the next release can inherit.
Save the release ID, exact treatment, publication time, data cutoff, cohort, observations, competing explanations and chosen action. Link the actual source data. Do not turn every review into a new disconnected report that nobody consults before the next commission.
The Roshan case is a real account of an audience-informed follow-up; its published cumulative snapshots are not the equal-age dataset used here. The synthetic data and blank postmortem worksheet make the comparison method inspectable without inventing a client experiment.
Questions about the method.
Can I compare two videos by their current total views?
You can describe those totals with their ages, but they do not establish equal-age performance. Use the same time since publication, consistent filters and relevant context for that comparison.
What should a content postmortem produce?
A verified account of what shipped and happened, plausible explanations with limits, and a concrete next decision with an owner and evidence to review. It should not end at a dashboard summary.
Tell us what you want to make.
Tell us about the channel or idea, the team you have and where you need help. We’ll discuss the scope on a 30-minute call.
Book a 30-minute call rishwajeet@machinehouse.media