A promotion is due to start, the team expects more purchases, and the media plan includes a seasonality adjustment. Before approving it, ask what is expected to change: the conversion rate for the campaign's actual objective, the number of people arriving, or the amount each customer spends. Those are different forecasts.
Google App campaigns can use seasonality adjustments in beta, according to Google's current documentation. Consider one for a substantial, temporary change in conversion rate backed by relevant evidence. Verify account access and campaign or ad-group scope, define the forecast's denominator and event window, and approve spending exposure separately. A promotion calendar alone does not justify an adjustment.
This guide helps an app growth team evaluate a proposal before it becomes a campaign change. The decision framework and fictional calculations below are editorial recommendations, not a Sharply Labs client result or a prediction of how an individual account will perform.
Are seasonality adjustments available for Google App campaigns?
Google's seasonality overview currently includes App campaigns in beta. It describes a tool for communicating anticipated conversion-rate changes around future events. Google recommends reserving it for major changes because Smart Bidding already handles ordinary seasonal patterns; its guidance favors events lasting one to seven days and warns that longer periods, such as more than fourteen days, may work less well.
Treat the beta label as an access check. An agency should demonstrate the supported configuration in the relevant account before including the feature as a committed deliverable. A public help page establishes documented capability, not access for every advertiser, subtype or proposed configuration.
The setup instructions document specific ad-group selection for App campaigns and say device selection is not applicable to them. Ask for an explicit inventory of affected campaign and ad-group IDs. Avoid an account-wide scope merely because it is easier to configure.
A promotion available to existing subscribers may have little relevance to an install campaign aimed at new users. A country-specific offer may not justify a common forecast for several markets. These are reasons to examine scope and objective; they are not claims that the platform automatically separates every such population for you.
Choose the control that answers the actual problem
Start with a sentence describing the proposed change in customer behavior. “We expect a higher purchase rate among comparable eligible arrivals during this offer” is a testable hypothesis. “It is a busy week” leaves the mechanism unresolved.
Observed or expected situation First decision to examine Evidence needed before acting --- --- --- A short promotion plausibly changes the optimized event's conversion rate Whether a seasonality forecast is justified Comparable event rates, offer eligibility and supported scope More demand is expected, with similar conversion propensity Spending capacity and commercial limits Demand evidence, marginal economics and exposure approval Reported conversions fall after a measurement defect Incident diagnosis and repair Event-path evidence and an affected-population register Users take longer to complete the outcome Evaluation timing and reporting maturity Relevant conversion lag and a consistent observation horizon Revenue rises because average order value changes Value economics and the actual bidding objective Value definitions, price effects and net contribution
These are suggested review priorities, not mutually exclusive platform states. A promotion can change several things simultaneously. Record each assumption separately so that one control is not expected to solve every part of the plan.
Use the App campaign budget-change framework when the unresolved decision is spending allowance or target economics. Use the data-exclusion recovery framework for a damaged measurement feed. A forecast about future behavior does not repair missing historical events.
Name the event before estimating the uplift
Google's App campaign setup guide distinguishes goals such as installs, in-app actions and pre-registration. Its bid-strategy guidance connects the available strategy choices to campaign goals and conversion inputs. The business outcome in a promotion deck must therefore be mapped to the configured objective before a forecast is useful.
Suppose the offer is a discount on an annual subscription. The team may forecast more paid subscriptions among trial users. That does not establish that the install rate among ad visitors will increase by the same proportion. Nor does a higher purchase count establish better net value if the discount reduces contribution per buyer.
Write the conversion definition in ordinary language. Include whether the event represents a first purchase, a renewal, a trial start or another action; whether repeat actions count; and which population is eligible for the offer. Match the forecast to the outcome that the campaign actually uses, rather than whichever metric produces the strongest slide.
For a value objective, separate event frequency from value per event. A promotion can improve one and weaken the other. A team should not enter a conversion-rate estimate simply by copying its revenue-growth target. If the connection between the proposed input and the campaign objective is unclear, resolve that connection before authorizing a change.
Build a forecast that another person can challenge
A useful forecast has a baseline, a proposed event-period rate and a reason for the difference. Its credibility depends on comparability, not the number of decimal places in the calculation.
For an internal planning calculation, define the rate as qualifying events divided by a clearly named eligible population over a fixed observation period. This planning denominator must be explained; do not present it as Google's internal model or assume that a backend population reproduces the platform's attribution population.
Compare like with like where possible. A previous promotion involving returning customers may be weak evidence for a new-user campaign. A different offer, market, acquisition source or product version can change the meaning of the comparison. Keep those differences visible instead of silently treating every previous sale as a reusable experiment.
Separate three evidence classes in the working document: directly comparable observations, directional analogies and unsupported expectations. The first can inform an estimate; the second needs explicit limitations; the third should remain a hypothesis. No universal minimum uplift or sample size is supplied here because those decisions require the actual data, uncertainty and exposure.
If the historical event was accompanied by a tracking repair, a major creative change or a different audience mix, identify that confounder. You may still learn from it, but you cannot confidently attribute the entire observed difference to the offer. A forecast with a wide uncertainty range can be more useful than a precise number built from incompatible cohorts.
Four gates before approving the adjustment
The following approval gates are an editorial operating framework. They help a growth owner distinguish an available feature from a defensible use of it.
Gate one: a temporary behavioral mechanism. State why eligible users should become more or less likely to complete the selected event, and why that change should end. An indefinite product improvement or a general seasonal narrative is a different planning problem from a short offer with a defined expiry.
Gate two: credible measurement. Establish that the baseline event and value definitions are usable. Reconcile material gaps before relying on their rates. If the rate changed because an event was duplicated, the forecast would encode a measurement error rather than a commercial opportunity.
Gate three: bounded implementation. Confirm beta access, affected IDs, start and end times, account time basis and the responsible operator. Review what is excluded as carefully as what is included. Have a second person compare the saved scope against the approved inventory before the event begins.
Gate four: separately approved economics. Name the spending exposure the business accepts, the evidence that would trigger intervention and who can make that decision. A conversion-rate forecast does not authorize a budget increase by itself. The business owner should understand the combined proposal, including any concurrent target or creative changes.
If a gate is unresolved, record a specific next action. That might be obtaining account confirmation, repairing an event definition or narrowing the population. “Wait for more data” is incomplete unless the team knows which observation would change the decision.
A fictional calculation: relative change is not percentage points
Consider a simplified internal planning example. A comparable baseline contains 10,000 eligible visits and 400 qualifying purchases over a defined observation horizon. The observed rate is 4%. The proposed promotion forecast is 600 purchases from the same 10,000 eligible visits, or 6%.
The difference is two percentage points. The relative increase is 50%, calculated as 6% divided by 4%, minus one. Those expressions describe the same hypothetical rates in different ways. Confusing them would produce a materially different input.
Now change the population assumption. If the promotion produces 15,000 eligible visits and 600 purchases, the rate remains 4%. Purchase volume increased by 50%, but conversion rate did not. Increased volume alone cannot justify copying a 50% rate increase into a planning proposal.
Add a separate commercial assumption: each purchase contributes $20 before acquisition cost in the baseline, while the discount leaves $12 during the event. Baseline contribution is $8,000; the event's 600 purchases contribute $7,200. More purchases can coexist with less contribution before advertising spend. This arithmetic omits fixed costs and many business-specific details; it illustrates why the forecast and the commercial approval must remain distinct.
None of these figures is a benchmark, recommended adjustment or estimate of incremental sales. The event population could include buyers who would have purchased later at full price. A before-and-after rate comparison cannot resolve that counterfactual on its own. Use the example to inspect assumptions, not to select a setting.
Keep a promotion register that survives the handoff
Record the offer, eligibility, qualifying event, baseline population, forecast basis and uncertainty in one document. Add the affected account and campaign identifiers, approved scope, timing and the person responsible for each operational step. Keep customer-level data and credentials out of a broadly shared planning packet.
Preserve the configuration before the change and the saved configuration afterward. Include other changes planned for the same interval: pricing, creative, product releases, measurement repairs and budget or target edits. This does not eliminate confounding; it prevents the next analyst from having to reconstruct it from memory.
Treat timezone as an operational acceptance item. The advertising interval, in-app offer expiry and reporting extracts need an explicit relationship. A date written without a timezone may describe different hours to the product team, the agency and the finance reviewer.
Set a review time that reflects the outcome's maturity. The conversion-window guide explains why the event clock matters. A promotion ending does not mean every eventual purchase has already appeared in the selected report. Preserve the extraction time and distinguish provisional observations from later, more mature figures.
What should happen after the event ends?
Google describes the adjustment as a scheduled forecast and says a compensating negative adjustment is unnecessary when the promotion finishes. That platform description is not a guarantee that your observed CPA, revenue or conversion rate returns to a particular level. See the official explanation of how the adjustment works.
Verify the scheduled interval has ended and the customer-facing offer has behaved as intended. Then review the quality of the forecast, the operational execution and the commercial outcome as separate questions. A correctly configured event can still have a poor forecast; an accurate rate forecast can still produce unattractive contribution.
Compare the forecast with observations using the same event definition, population boundaries and horizon wherever possible. Record where comparability broke. If a product release changed the checkout flow midway through the offer, preserve that fact rather than attributing the full difference to the bidding adjustment.
Do not call a favorable report proof that the adjustment caused incremental growth. A stronger causal claim needs an appropriate experimental design and the assumptions that support it. When no such design exists, report the observed results and limitations plainly. That still creates a useful planning record for the next promotion.
What should an agency proposal include?
Ask for a forecast memo, scope inventory, change record and evaluation plan. The memo should explain why a temporary rate change is expected and how uncertainty affects the recommendation. The inventory should identify the campaigns or ad groups affected and show that the proposed scope is supported in the account.
The commercial plan should state the combined exposure across settings, rather than approving each change in isolation. The evaluation plan should define which evidence supports continuing, intervening or deferring judgment. Assign a named decision owner for each response, including outside normal working hours if the event requires it.
Reject promises of guaranteed beta access, precise bid behavior, stable acquisition costs or automatic incremental profit. Also question a proposal that cannot distinguish increased traffic, conversion propensity and value per customer. Those distinctions matter more to the buying decision than an elaborate screenshot of the settings panel.
For established app teams evaluating a short promotion, Sharply Labs' mobile app growth work and paid acquisition services provide the relevant service context. Bring a redacted forecast, the intended optimization event and the decision your team cannot yet authorize to an initial conversation. The useful first output is an agreed view of evidence gaps, owners and a scoped validation plan; detailed audit or implementation work is separately agreed. No beta access, approval, CPA, revenue or campaign outcome is promised.