A monthly report can show lower acquisition cost while leaving the buyer unable to decide whether to renew an agency, fund another creative cycle or increase media investment. The missing piece is often the relationship between a reported number, its definition and the action it supports.
Useful mobile app marketing agency reporting makes five things reviewable: the decision requested, the definitions behind the numbers, the evidence that supports the interpretation, the work completed, and the next action with an owner. Evaluate those reporting obligations separately from business performance. A reproducible report can describe poor results honestly; an attractive result can still be unsupported. Neither a polished dashboard nor a positive platform return is sufficient evidence of agency effectiveness.
This article proposes acceptance criteria for an established agency relationship. They are editorial recommendations, not a claim about an industry standard, a legal obligation or a proprietary Sharply Labs methodology. Agree the scope with the people responsible for media, product, analytics and finance before using the criteria to judge delivery.
Start with the decision the report must support
Write the pending decision before requesting another chart. “Review October performance” is an agenda item. “Decide whether to fund the next acquisition and creative cycle under the current scope” identifies a decision and a boundary.
The report should state the available choices, the evidence needed to distinguish them and the consequence of waiting. Some decisions are reversible operational changes. Others commit production capacity, require a release or expand exposure to uncertain acquisition economics. Those differences should determine the evidence requested.
Keep three questions separate. Did the agency deliver the agreed work? What happened to the business? What can reasonably be attributed to the work? Each question may have a different answer. A delayed product release can prevent a well-prepared campaign from launching. A promotion can improve reported revenue without establishing that an agency change caused the improvement.
If the engagement is still being purchased, use the mobile app marketing RFP guide to specify reporting deliverables. If access, definitions and decision rights have never been agreed, the agency onboarding framework is the earlier problem to solve.
Five gates for accepting a recurring report
Treat acceptance as permission to use the report for its stated decision. A report that fails a gate can remain useful for a narrower purpose. Record the restriction instead of marking every page either approved or rejected.
Gate Evidence to request What prevents acceptance --- --- --- Decision Named choice, scope and accountable reviewer A performance narrative without a requested decision Definitions Source, period, event, value basis and relevant exclusions Similar labels covering different populations Reproducibility Saved export or documented query and extraction time Numbers that depend on an unexplained manual adjustment Interpretation Observation, possible explanation and remaining uncertainty Causal conclusions from an uncontrolled comparison Action Owner, dependency, acceptance check and review point Recommendations that no one can authorize or verify
Do not turn these gates into an elaborate monthly scoring exercise. Their purpose is to expose the first missing condition that could change the decision. When a definition is wrong, another ten slides of analysis do not repair it.
Gate one: name the decision and its limits
The executive summary should say what the reviewer is being asked to approve. It should also say what the evidence does not yet support. A report may support continuing a bounded production test while remaining insufficient for a larger budget commitment.
Ask the agency to identify the strongest evidence against its preferred recommendation. That request helps reveal whether the proposal depends on a fragile assumption. It should produce a useful limitation, not a ritual statement that more data would be helpful.
Gate two: make definitions available beside the numbers
Maintain a compact metric dictionary. For each decision-critical measure, record the source, event, unit, population, time basis and exclusions. State whether a value represents a platform-attributed event, a product event, a customer, a transaction or a financial amount.
The dictionary should survive changes in report design. A new dashboard should not silently change what a paying user means. If the definition changes, show the effective date and whether the old and new series remain comparable.
Gate three: preserve a reproducible evidence path
Request an export, saved report or documented extraction that another authorized reviewer can inspect. Preserve the filters and extraction time with it. A screenshot can communicate a result, but it may omit the settings needed to understand the result.
Reproducibility does not require distributing raw customer records. Use aggregated or redacted evidence appropriate to the decision and the agreed access boundaries. The report should identify an authorized owner for restricted evidence rather than copy sensitive data into a presentation.
Gate four: distinguish observation from explanation
An observation describes the number that changed. An explanation proposes why it changed. A causal claim says the change would not have happened without the intervention. The report should make those levels visible.
For an observational comparison, list the material changes that occurred alongside the agency work. These may include product availability, price, promotion, event definitions, market mix and acquisition budget. Do not ask the writer to manufacture certainty when the decision can be made with a clearly bounded interpretation.
Gate five: make the next action testable
Every material recommendation needs someone who can authorize it and someone who can execute it. State the dependency, the evidence expected after completion and the condition that would change the plan.
“Improve onboarding” does not assign a task. “Product reviews the first-session drop-off evidence and accepts or rejects a scoped investigation before the next acquisition expansion” gives the recommendation a decision owner. The exact timing belongs to the engagement, rather than a universal rule in a reporting template.
Use a disagreement register instead of forcing totals to match
Different reports can describe related activity without sharing the same counting rules. A good agency report should identify whether a discrepancy is expected, unresolved or caused by an actual implementation defect.
Google provides a concrete example. Its App campaign asset-reporting documentation says auto-generated assets are excluded, so summed asset spend should not be expected to equal campaign spend. That establishes a reporting limitation; it does not establish the cause or size of a particular account's gap. (Google Ads Help: App asset reporting)
Google's enhanced reporting documentation also cautions that component-level metrics can overlap, asset reports may show served rather than billed cost, and some auto-created asset costs may be absent. It directs advertisers to campaign totals for overall performance and spend tracking. Preserve the distinction when presenting operational spend and financial reconciliation. (Google Ads Help: enhanced asset reporting)
For each disputed number, record both source definitions, the affected decision, a possible explanation, the evidence still needed and the responsible owner. Close the disagreement when its implications are understood, even if the two tools are not supposed to become numerically identical.
Avoid allocating an unexplained residual to a convenient category. “The report does not include every asset” does not justify assigning the entire difference to generated video. A reporting boundary is not an account-specific diagnosis.
A fictional example: an improving ratio that cannot support renewal
Consider a fictional subscription app reviewing two acquisition periods. These numbers are an arithmetic illustration, not a benchmark, client result or recommended target.
In period A, the team records $20,000 of media spend and 200 first-time paying customers under its agreed definition. The observed media cost per new paying customer is $100. In period B, the agency reports $24,000 of spend and 300 purchase events, producing $80 per purchase event.
The second number looks better, but the denominator changed. Suppose the 300 events comprise 180 first-time customers and 120 renewal events. Under the original customer definition, period B's observed media spend per first-time paying customer is about $133.33. That does not establish that acquisition deteriorated because of agency work; it establishes that the two headline ratios do not answer the same question.
The reviewer should first request a comparable first-time-customer series, then examine whether the acquisition populations and observation periods permit a useful comparison. If period B contains newer cohorts with less time to become paying customers, state that maturity difference. Do not silently treat incomplete observation as a permanent outcome.
Next, separate media spend from total engagement costs. Agency fees, production, software and other agreed costs may matter to the business decision, but they should enter an explicitly defined calculation. They should not appear selectively in one period's numerator.
The immediate decision may be to accept the delivery record, return the headline comparison for correction and defer the expansion request. A reporting correction alone is not a reason to declare the agency effective or ineffective. The mobile app ROAS guide provides the complementary framework for deciding which value and cost definitions a return calculation should use.
Ask what creative learning changed, not just what launched
Creative reporting should connect production to a decision. Ask which customer argument was tested, which execution changed, where delivery occurred and what the evidence now makes worth producing.
A count of new files can establish output, but it does not by itself establish differentiated concepts or learning. Conversely, limited delivery does not prove a concept failed. Record whether the uncertainty comes from insufficient exposure, incompatible comparison conditions or a genuinely weak observed response.
Google describes asset-level ratios such as CPA and ROAS as directional because the assets serve in combinations. They do not isolate a single asset's contribution. An agency should not present a ranked asset table as a controlled experiment. (Google Ads Help: asset-level metrics)
For portfolio construction and Google-specific interpretation, use the App campaign creative asset guide. The recurring agency report has a different responsibility: explain which production decision follows, what remains uncertain and who owns the next brief.
Separate agency responsibilities from shared dependencies
Define responsibility around decisions and deliverables the parties can actually control. The agency may own a campaign change record, a creative brief or an evidence interpretation. Product may own a release. Analytics may own an event definition. Finance may own the treatment of revenue and costs.
For a missed action, request the original commitment, the dependency, the date the dependency became known and the proposed resolution. This makes the discussion more useful than assigning every disappointing number to media execution or treating every product dependency as an excuse.
Use the same discipline for successes. A stronger business result should prompt examination of the conditions that produced it. It should not automatically become an agency case study or an instruction to repeat every change made during that period.
At renewal, ask whether the current operating arrangement can resolve the next important constraint. The question may be one of scope, product readiness or measurement ownership rather than vendor quality. Changing agencies does not automatically remove those constraints.
Match the review cadence to the decision
Do not assume a monthly presentation can carry every operating responsibility. Agree which incidents require prompt escalation, which checks belong in routine operations and which decisions need more mature evidence.
For example, a broken destination or misleading creative claim should have a correction path without waiting for the next monthly review. A longer-term value question may require an agreed observation period. The report can show progress and uncertainty without prematurely closing that question.
Keep a decision log between reviews. Record what was approved, the conditions attached, the owner and the evidence expected. At the next review, begin with those commitments. This prevents each report from becoming a fresh narrative that ignores the previous month's unresolved recommendation.
If an automated summary is used, require a human reviewer to verify the source, definition and recommendation. Automation can format evidence; responsibility for approving a business action remains with the named decision owner.
Renew, narrow, repair or defer?
Use the reporting assessment to define the next commercial conversation rather than manufacture a universal agency score.
Evidence state Appropriate decision to examine Required next evidence --- --- --- Report reproducible; work useful; next scope clear Continue the agreed scope Next-cycle acceptance and business constraints Report sound; business outcome weak Reassess the constraint or scope Competing explanations and feasible next test Outcome attractive; reporting unsupported Repair evidence before expansion Comparable definitions and reproducible exports Material data or product dependency unresolved Narrow or defer the affected commitment Named owner and acceptance condition
These are conditional choices. They do not prescribe termination, contractual remedies or a minimum performance threshold. Use the actual agreement and appropriate professional advice for legal or financial decisions.
Bring one disputed decision to the conversation
Sharply Labs' mobile app growth practice and paid-acquisition and growth services are relevant to app founders and growth teams whose agency reporting leaves a production, measurement or acquisition decision unresolved.
Bring a redacted recent report, the definitions behind its main numbers, available source exports and the decision currently waiting. A scoped conversation can identify the evidence gap, responsible owner and next proving check. Any detailed audit or implementation scope should be agreed separately.
The useful output is a clearer decision and an evidence plan. The conversation does not promise lower acquisition cost, improved return, complete attribution, a causal verdict on the current agency or a particular renewal outcome.
Evidence and limitations
The three official Google sources cited above were checked on November 20, 2026. They support narrow reporting definitions and limitations, not findings about an individual account. Interfaces and available reports can change. No private client results, agency benchmark, search-volume estimate or first-party practitioner case is presented.
The acceptance gates, disagreement register and renewal matrix are original editorial recommendations. Apply them to the agreed scope, business model and available evidence. Their purpose is to make the next decision reviewable, rather than make every dataset agree or turn uncertainty into a performance claim.