Your editors release new episode cuts, buyers shift spend between two short-drama titles, and the next report shows better return on ad spend. Before commissioning more cuts in the same style, check which titles received the money. A portfolio result can improve even when performance within every title stays unchanged.
Quick take: To assess short drama title mix performance, compare the same titles in the same markets and language versions before interpreting the total. Show each title’s spend share alongside its results. If the improvement disappears when you hold the mix constant, the aggregate does not support a claim that the episode cut improved. If improvement remains within comparable titles, investigate the cut alongside other changes before assigning credit.
Separate portfolio performance from episode-cut performance
A short-drama title is the series being promoted; an episode cut is a particular advertising edit of its footage. Title mix describes how delivery or spending is distributed across those series. Changing the edit and changing the share of spend assigned to a title are different changes, even when they happen in the same launch batch.
The business consequences differ. A more favorable title mix may justify reviewing allocation, subject to revenue quality and repeatability. Evidence for a better cut may justify another production test. Confusing the two can send editors into a new production cycle without evidence that the editing choice helped.
XMP’s discussion of the short-drama creative iteration workflow identifies a practical problem: creative performance often lacks enough context to explain the result. Title mix is one piece of that context. Before extracting a reusable creative lesson, establish whether the observed change happened inside a title or accompanied a shift between titles.

Record what changed in both periods
Start with a side-by-side record of the earlier and later periods. Keep titles that entered or left the portfolio visible. Otherwise, a report filtered to current titles can silently change the comparison population.
Use one row per title, market and language version, with separate cut identifiers underneath where the data supports them. Add channel, operating system or destination splits when those differences affect comparability. The following is a recommended analysis record, not a claim that every reporting system supplies every field.
| Record | What to enter for each period | Why it matters |
|---|---|---|
| Promoted content | Stable title ID, episode, cut version and active dates. | Separates a replacement title from a replacement edit. |
| Market and language | Country or region, subtitle and audio versions. | A title can remain the same while its audience context changes. |
| Delivery conditions | Channel, OS, destination, audience, optimization goal and material bid or budget changes. | Defines the comparable scope and remaining alternative explanations. |
| Actual mix | Spend and spend share by title; impressions or clicks when relevant. | Shows what delivered, beyond the intended budget allocation. |
| Outcome definition | Metric source, event meaning, revenue basis and attribution window. | Prevents different outcomes from sharing one label. |
| Time and readiness | Date range, timezone, cohort age and reporting status. | Prevents newer cohorts from being compared with more mature ones. |
Preserve the link between assets and delivery records. XMP’s creative production workflow guide explains why asset usage, campaign assignment and performance need to remain connected. For this review, that means being able to identify which cut actually ran for which title, market and period.
How the total improves when neither title improves
Illustrative arithmetic only: The following invented figures demonstrate aggregation; they are not customer results or performance benchmarks. Assume one app, market, language version and measurement basis across two equal reporting periods. Revenue is attributed cohort revenue measured through the same cohort age, with all cohorts mature enough for that window. Spend and revenue are in USD.
| Title | Period A spend / revenue | Period B spend / revenue | ROAS in both periods |
|---|---|---|---|
| Title A | $8,000 / $4,000 | $2,000 / $1,000 | 0.50 |
| Title B | $2,000 / $2,000 | $8,000 / $8,000 | 1.00 |
| Portfolio | $10,000 / $6,000 | $10,000 / $9,000 | 0.60 → 0.90 |
Title B’s share of spend rises from 20% to 80%. Neither title’s ROAS changes, yet portfolio ROAS rises from 0.60 to 0.90. The entire aggregate movement in this constructed example comes from the changed spend mix.
The calculation is straightforward: portfolio ROAS equals the sum of each title’s ROAS multiplied by its share of spend. Period A is 0.80 × 0.50 + 0.20 × 1.00 = 0.60. Period B is 0.20 × 0.50 + 0.80 × 1.00 = 0.90.
Even if new cuts launched in Period B, these figures would provide no evidence of an improvement within either title. They also would not establish profitability or prove that Title B could sustain the same return at further scale. This is short drama UA aggregation bias in practical terms: interpreting a change in the combined result as a change in the performance of its components.
Compare within titles, then hold the mix constant
First inspect each title within the same market and language version. Identify whether the selected outcome improved, deteriorated or remains uncertain. If a title moved into a different market at the same time as the edit changed, split the markets where possible. A title-level total can hide a market-mix change just as a portfolio total can hide a title-mix change.
Next, use a fixed mix as a diagnostic. For ROAS, choose and disclose a reference set of spend shares, such as the earlier period’s shares among titles present in both periods. Apply those same shares to each period’s title-level ROAS. In the example, applying Period A’s 80%/20% split to Period B still produces 0.60. The improvement disappears.
This answers a limited question: how would the summary differ if the observed title-level rates received the same weights? It does not predict what would happen if you actually moved the budget. Delivery, audience composition and returns may change when spending changes.
Use the denominator appropriate to the metric. ROAS uses spend weights; CTR uses impression weights. Calculate aggregate CPI as total spend divided by total installs, rather than averaging title CPIs using spend shares. Retain the underlying totals so another analyst can reproduce the result.
Keep newly launched and discontinued titles in a separate contribution view. Do not invent an earlier-period rate for a new title. Report how much of each period’s spend is covered by the matched comparison, so readers know whether its findings describe most of the portfolio or only a narrow slice.

State what the evidence can support
Use the conclusion-boundary framework in the broader guide to paid UA confounding factors when several inputs changed together. For a title-mix review, keep the report’s wording as specific as its evidence.
| Observation | Supported interpretation | Conclusion to withhold |
|---|---|---|
| The total improves; comparable title results stay flat. | The mix change explains the aggregate movement if the arithmetic reconciles. | The new cuts improved acquisition performance. |
| Comparable title results improve, but bids or audiences also change. | Performance improved within the stated scope. | The edit alone caused the improvement. |
| Only new titles use the new editing approach. | The new title-and-cut combinations have observed results. | The editing approach outperformed independently of title choice. |
| Cut-level outcomes are allocated estimates. | The breakdown offers directional context. | Each cut generated the exact assigned revenue or conversions. |
| Comparable rows contain few outcomes or immature revenue. | The comparison remains unresolved. | A confident winner or failure. |
The distinction about estimates matters when several titles or cuts share one advertising object. XMP’s automation and reporting explanation states that, where exact creative-level MMP data is unavailable, creative outcomes can be approximated using weighted higher-level data. Confirm how your report was produced before treating its apparent detail as direct measurement.
If title and cut changes cannot be separated, preserve the combined observation and design a narrower follow-up. For example, compare old and new cuts of the same episode within a consistent market and language version, using a controlled comparison where feasible. A before-and-after comparison still leaves timing and delivery changes as possible explanations.
Choose the next action from the evidence
- Validate measurement. Confirm title-to-asset mapping, metric definitions, attribution basis and whether outcomes are directly measured or allocated. Resolve unexplained gaps before interpreting the mix.
- Check maturity and sample sufficiency. Align cohort ages and examine outcome counts within comparable rows. If the available evidence cannot distinguish a useful difference from noise, record the result as unresolved.
- Connect acquisition cost with business value. Review comparable viewing, payment or revenue outcomes that your business actually measures. Keep attributed revenue distinct from internally confirmed revenue, and avoid treating cheaper acquisition as sufficient evidence of better economics.
- Select the action that matches the finding. Review allocation when mix explains the result; investigate a cut when improvement remains within a comparable title; repair missing mappings when the comparison cannot be made. Record the scope, remaining uncertainty, owner and next review condition.
A useful production handoff names the title, market, language, cut versions and observation window, then says exactly what changed. That gives editors a specific hypothesis to investigate without turning a portfolio movement into an unsupported creative rule.