Creative Asset Management for Mobile Games: Naming, Tags and Learning History

A mobile game team finishes a gameplay video, translates its captions, changes the opening scene, and exports versions for several channels. Weeks later, a media buyer asks which opening deserves another production round. The files are available, but the report contains ad IDs, the folder contains filenames, and the original brief describes only the concept.

Creative asset management for mobile games connects those records. It lets a team trace a result back to the exact creative, understand what changed, and decide which learning applies to the next brief.

Quick take: Give each exported asset a stable ID, describe it with consistent concept, hook, language, platform, and version fields, and map it to every ad that uses it. Keep performance conclusions in dated learning records with their test conditions. If a result cannot be traced to an exact version and comparable delivery context, treat it as unresolved before commissioning more variations.

Connect source material, exported assets, and ads

The business benefit of an organized library is being able to answer a production question: should the team reuse this gameplay sequence, change the hook, or stop developing the concept? A folder location alone cannot answer that question. You need a relationship between the production inputs, the delivered creative, and its results.

The XMP creative asset management guide describes creative libraries, categorization, and performance tracking. The practical method below adds a proposed record structure for mobile game teams; these fields and version rules are workflow recommendations, not a claim that XMP implements this exact schema.

Use three connected record types. A source record identifies gameplay capture, editable projects, voiceovers, or other production inputs. An asset record identifies one exact exported file. A deployment record identifies where and when that file was used, including platform, account, and ad ID.

Keep the relationships flexible. One source can feed many exports, and an export can combine several sources. One export can also run in multiple ads. Where an ad combines a video, end card, and copy, record every component and its role. An ad-level result does not automatically reveal which component caused the outcome.

Retain platform asset IDs separately from internal asset IDs. An upload into another account may create another platform reference without creating a new creative idea. Use platform and account alongside ad ID when matching reports, rather than assuming an ID or display name is unique everywhere.

Build a mobile game creative naming convention around stable fields

A filename should help someone recognize an asset during handoff. Structured metadata should make the library searchable and comparable. Keeping both means a renamed file does not erase its identity, and a long filename does not become the only place where important context survives.

Start with the following creative asset taxonomy. The examples are illustrative values for a fictional puzzle game, not platform requirements.

A readable filename could be pzl_c017_h02_en-US_snapchat_9x16_v002_a104.mp4. Here, c017 groups the concept, h02 identifies the hook, and a104 identifies the exact export. Define the version sequence within a family such as game, concept, hook, language, platform, and format. The asset ID removes ambiguity when several families each have a v002.

Choose one approved spelling per tag value. For example, map “fail opener” and “failed move” to the same agreed hook term if they mean the same thing. Assign an owner to maintain that dictionary, and allow an explicit unknown value during intake instead of guessing. Resolve unknown fields before using them to group results.

Keep language separate from target market: an English asset can run in multiple countries, and the market belongs to the delivery context. Similarly, keep descriptive tags separate from conclusions. “Failed-move” describes content; “winner” is an evaluation that needs a date and evidence.

Store format requirements without crowding the filename

A platform tag helps route an export, but it does not establish launch eligibility. Record the intended ad format and the specification source checked during review. For example, Snapchat's official ad formats page lists different requirements for Single Image or Video Ads, Commercials, and AR formats. That distinction makes a format field more useful than a generic “Snapchat-ready” label.

Keep dimensions, duration, encoding, and review date in metadata. Check the current requirements for the specific format before deployment; a previous approval should not silently become approval for another format.

Use ad creative version control to preserve what actually ran

When a file changes after launch, overwriting it can leave yesterday's results attached to today's creative. Preserve the previous export and create a new asset record for the replacement. The history should explain both what changed and why the team made the change.

Illustrative version example: A puzzle team has a rescue concept with a failed-move opening. It corrects a subtitle, then produces a successful-move opening for comparison. No performance result is assumed.

A new concept ID is appropriate when the core creative proposition changes. A new hook under the existing concept is appropriate when the opening changes but the central proposition remains. Write that boundary into the team's tagging dictionary so different producers classify similar edits consistently.

If an ad's creative changes while its ad ID remains the same, preserve dated deployment intervals for the old and new assets. Avoid attributing a combined reporting period entirely to either file. When a report cannot resolve the transition, mark that period as mixed.

Record reusable conclusions and the conditions that invalidate them

The learning history should tell a producer what to do next and where the evidence stops. A label such as “bad creative” hides whether the asset had little delivery, attracted low-value installs, or ran under conditions that differed from the comparison.

For each review, attach a dated learning record to the exact assets and deployments. Include these fields:

• Question: State the decision being tested, such as whether a different opening merits further production.
• Context: Record game, market, OS, channel, placement where available, audience, objective, and delivery dates.
• Evidence: Link the report and record its source, timezone, currency, attribution basis, metric definition, denominator, and cohort age.
• Observation: Describe what the report shows, separating recorded facts from interpretation.
• Decision: Choose reuse, revise, retire, or unresolved, with the reason and responsible owner.
• Recheck conditions: Specify changes that require another review, such as a new market, game build, offer, placement, or measurement setup.

Sample sufficiency depends on the question and outcome being assessed. Record spend, delivery, and relevant event counts alongside the conclusion. A creative with limited exposure should remain unresolved when the evidence cannot support the decision. A mature install cohort and a recent cohort should not receive equivalent revenue judgments simply because both appear in the same report.

Illustrative learning note: “Question: can the successful-move opening replace the failed-move opening? Assets: a104 and a105. Observation: the variants ran in different markets. Decision: unresolved for the hook comparison. Next action: test under comparable conditions. Recheck: retain the game build and localization context.” This records a useful limitation without inventing a winner.

When reviewing concept groups, retain the underlying deployments so the team can inspect differences in delivery. Calculate aggregate rates from compatible underlying totals rather than averaging percentages. If reporting only supports a bundle or campaign result, keep the conclusion at that level instead of assigning it to an individual asset.

Put the system to work on the next production batch

Start with active assets and the next batch of exports. Give the creative producer responsibility for concept, hook, source, and revision records; give the media buyer responsibility for deployment mapping; and give the analyst responsibility for evidence definitions and learning notes. One person can cover several roles, but each handoff needs an owner.

Before launch, check that the approved file matches its asset record and that required metadata is present. After launch, reconcile platform references and actual ad usage. During creative review, open the linked evidence before writing the next brief. Preserve retired files and their history while removing them from the ready-to-use selection.

XMP's creative production automation overview describes folders, tags, creative libraries, and reporting as parts of a connected workflow. Use those capabilities to support retrieval and review, and maintain any additional lineage or learning fields in a linked record where needed. The taxonomy remains a team responsibility.

Close each review with a practical decision sequence:

1. Validate the mapping and measurement. Confirm that the report refers to the intended export, deployment interval, and outcome definition.
2. Check maturity and evidence sufficiency. Identify immature cohorts, sparse delivery, or mismatched conditions before making a creative judgment.
3. Connect cost with business value. Consider acquisition cost alongside the relevant retention, revenue, or post-install outcome, plus the effort required for another production round.
4. Choose the next action. Reuse a supported idea within its recorded conditions, revise a defined element, retire an unsuitable asset, or keep the conclusion unresolved. Link that decision to the next brief.

Qu Jun Xing

AdOps & Growth Notes

Ad Ops AI & SaaS Automation

10+ years in mobile growth and AdTech operations. I build practical systems across AdOps, automation, SEO/SEM, GEO, and lifecycle marketing.

in LinkedIn 🗂️ More notes ⚡ Request an XMP Trial

Questions or updates? Reach out on LinkedIn or use the “Give feedback about this article” button.