Mobile app marketing agencies are usually priced in one of five ways: a fixed project fee, a monthly retainer, a percentage of managed media spend, a performance-linked fee tied to an agreed outcome, or a hybrid that combines a base fee with a variable component. The fee model tells you how the agency is paid. It does not tell you what work is included, who owns the accounts and data, or what you will still have to do yourself. Those three answers decide your real cost.
Why two proposals with the same number are rarely the same purchase
A proposal is a bundle. It contains labor, decision rights, tooling, creative production, measurement work, store work, and a set of assumptions about what your team will supply. When two agencies quote the same monthly number, they are almost never selling the same bundle. One may include four new video concepts a month and store-listing experimentation; the other may include media buying only, with creative billed separately and analytics treated as your responsibility.
That is why "cheaper" is the least reliable word in a vendor comparison. Before you compare fees, you have to make the bundles comparable. This guide is about that normalization step. If you are earlier in the process and still deciding which agencies belong on the shortlist at all, start with our guide to choosing a mobile app marketing agency, which covers diligence, references, and fit; this page picks up where pricing and scope economics begin.
One caution before the models: this article deliberately publishes no market rate ranges. Any number presented as a typical fee is a claim about a market that varies by geography, seniority, channel mix, app category, and scope. You do not need a benchmark to evaluate a proposal. You need a consistent way to compare the proposals actually in front of you.
The five fee models, compared on the same axes
Fee model How you are billed Where it fits Main risk to the buyer What must be defined in writing --------------- Fixed project One price for a defined deliverable and end date Audits, account rebuilds, measurement implementation, store-asset overhauls Scope creep, or a deliverable that ends before results can be observed Deliverable list, acceptance criteria, revision count, end date, what happens after handover Monthly retainer Recurring fee for an agreed scope and team Ongoing operations with steady workload Paying for capacity you do not use, or quiet scope shrinkage over time Named roles, hours or output commitments, response times, what counts as out of scope Percentage of media spend Fee scales with the budget the agency manages Programs where workload genuinely tracks budget across many campaigns Incentive to grow spend; workload and spend can diverge in both directions Percentage tiers, minimum fee, what spend counts, who approves budget increases Performance-linked Fee tied to an agreed outcome (installs, activations, trials, purchases, revenue) Situations where the outcome is cleanly measurable and mostly influenced by media Definition and attribution disputes; outcome moved by factors neither side controls The exact event, the source of truth, the attribution window, dispute and audit process Hybrid Base fee plus a variable component Most mature engagements Complexity; two things to renegotiate instead of one Base scope, trigger for the variable part, caps, and review cadence
Read the matrix as a set of tradeoffs, not a ranking. Each model is legitimate. Each becomes a problem when it is applied to a program whose workload, measurability, or decision structure does not match it.
Normalize the proposal before you compare the fee
Use the same worksheet for every vendor. The purpose is to move every cost out of the headline number and into a named line, so that what remains is genuinely comparable.
Total annualized cost of the engagement = agency fee (12 months, including any onboarding or setup fee) + media budget you intend to run + creative production not included in the fee (concepts, editing, UGC sourcing, licensing, localization) + measurement and tooling (MMP, analytics, dashboarding, BI, creative-management tools) and who pays for them + store work (screenshot and icon production, localization, listing experiments) + taxes, pass-through costs, currency handling, and any markup applied to media or third-party invoices + internal client time, costed honestly (approvals, engineering for events and deep links, design, legal, finance) − any work the agency removes from an existing vendor or contractor you can now cancel
Then, for each proposal, write down four numbers next to that total: what is fixed, what varies with spend, what varies with outcome, and what is not covered at all. The last column is the one that decides most engagements. A proposal with a low fee and a large uncovered column is not a cheap proposal; it is a partially staffed program with the remainder assigned to you.
Two normalization rules save the most arguments later. First, if a cost is conditional, record the condition, not the optimistic case. Second, if a line is "included," ask what volume is included — "creative included" means very little without a monthly count and a definition of what counts as a new concept versus a variation.
The capability and ownership map
Fee models describe payment. The capability map describes coverage. For each capability, mark who does the work, who approves it, and who owns the asset or account afterwards.
Capability Typical work Ownership question to settle --------- Paid user acquisition Campaign build, bidding inputs, budget pacing, structural changes, testing cadence Who holds the ad accounts and billing, and who can revoke access Creative Concepting, production, editing, iteration, versioning, asset libraries Who owns source files, footage, and creator licensing after the contract ends ASO and store conversion Metadata, screenshots, icon, app preview, listing experiments Who has store console access, and at which permission level Analytics and attribution Event definitions, SDK and API instrumentation, conversion setup, reconciliation, reporting Who owns the MMP and analytics contracts and the raw event data Lifecycle Onboarding, push, email, paywall messaging, win-back Whether lifecycle is in scope at all, or assumed to be yours Product dependencies Event implementation, deep links, paywall or onboarding changes needed for tests Whose engineering time is committed, and on what timeline
Two capabilities are worth extra attention because they quietly determine whether the rest of the program can work at all.
Measurement first. Google's documentation describes app conversion tracking covering installs, first opens, and in-app actions, with reporting that may include modeled conversions (Google Ads: About mobile app conversion tracking; Google Ads: Set up mobile app conversion tracking). Google also documents that app conversion counts can differ across Firebase, Google Play, and third-party sources because of settings, filtering, aggregation, and attribution logic (Google Ads: About comparing app conversions). Where an App Attribution Partner is used, Google documents that the integration can pass app events and consent status for attribution and optimization (Google Ads: Tracking app conversions with an App Attribution Partner). None of that documentation determines an agency's price or proves that an implementation was done well — but it does mean that "measurement" is real, scoped work with named owners, not a line item you can leave blank in a proposal.
Store work second. Apple documents that App Store Connect provides first-party acquisition and engagement analytics (Apple: App Analytics) and that teams can test store creative such as icons, screenshots, and app previews (Apple: Overview of product page optimization). Google Play documents store listing experiments run on real store traffic (Play Console: Run A/B tests on your store listing). Both are platform- and audience-specific, and traffic volume affects how quickly a conclusion is usable. If your proposals differ on whether store conversion work is included, they differ on far more than a fee — see our breakdown of store screenshots and paid UA economics.
Campaign operations sit on top of both. Google's App campaign guidance ties performance to correct conversion settings, bidding inputs, creative assets, deep links, and controlled changes (Google Ads: Best practices for App campaigns). Vendor best practices are not guaranteed outcomes, but they are a useful checklist for whether a proposal covers the inputs the platform itself says the work depends on. Our comparison of Google App Campaigns and Meta Ads for apps and the wider paid channel stack are useful when a proposal claims coverage across many channels at once.
A worked example, with clearly hypothetical numbers
The numbers below are illustrative only. They are not benchmarks, market rates, or claims about what anything should cost. They exist to demonstrate the arithmetic.
Two proposals for the same twelve-month program:
Proposal A: 10,000 currency units per month, described as "full-service app growth." Media managed separately. Creative: "up to 2 new concepts per month." Measurement: "supported, client owns implementation." Store work: not mentioned. Tools: client pays. Notice period: 60 days.
Proposal B: 14,000 currency units per month. Creative: 8 new concepts and 24 variations per month, source files transferred quarterly. Measurement: event map, MMP configuration, and monthly reconciliation included. Store work: two listing experiments per quarter where platform traffic allows. Tools: agency pays for creative tooling, client pays for MMP. Notice period: 30 days.
Annualized fee: A = 120,000; B = 168,000. B looks 40% more expensive.
Now normalize. Under A, you still need the missing capability. Assume — again, hypothetically — external creative production at 4,000 per month (48,000 per year), a contractor for measurement setup and monthly reconciliation at 2,000 per month (24,000), and store creative work at 3,000 per quarter (12,000). A's normalized annual total becomes 204,000 against B's 168,000, before internal time.
Then add internal time. If A requires roughly 20 additional hours a month from your team for briefing, coordination, and reporting, and you cost internal time at 60 per hour, that is another 14,400 a year. The headline-cheaper proposal is now materially more expensive and more fragile, because more of the delivery risk sits with people who have other jobs.
Run the same arithmetic with your own figures and the conclusion may reverse — if you already have an in-house creative team and a working measurement stack, the uncovered column under A shrinks toward zero and A becomes the better purchase. That is the point of the exercise: the answer depends on your gaps, not on the fee.
Performance-linked fees: the definition problem
Performance-linked pricing is not inherently better or worse. It is harder to define. The outcome an agency is paid on is influenced by product quality, pricing and paywall design, retention, organic demand, seasonality, attribution rules, and your own execution speed. When any of those move, the fee moves with them, and neither party can cleanly separate the causes.
That is a solvable problem if you settle five things before signing: the exact event being paid on and how it is defined; the single source of truth for counting it; the attribution window and model; how modeled or aggregated conversions are treated, given that platform reporting may include modeled conversions and can differ across sources (Google Ads: About comparing app conversions); and the process for disputes and audits. If you cannot answer all five, the model will generate an argument later. Our guide to attribution architecture beyond platform reporting covers the reconciliation work this depends on.
Percentage of spend: incentives and workload, not accusations
Percentage-of-spend pricing exists because for many programs, workload genuinely does scale with budget: more campaigns, more markets, more creative in rotation, more pacing decisions. Treat its problems as structural tradeoffs rather than bad faith.
Two mismatches are worth planning for. The incentive mismatch: the fee rises when spend rises, even when the right recommendation is to hold or cut. The workload mismatch runs both ways — a large, stable, mostly automated budget can require less work than a small, fast-moving portfolio of tests. You can manage both with tiered percentages that decline as spend grows, a floor and a ceiling on the fee, a written rule about who authorizes budget increases, and a periodic scope review that checks the fee against actual work.
What the cheapest quote often leaves out
Cheap quotes are usually cheap for identifiable reasons, and most of them are visible if you ask. Common exclusions: creative production volume, or the difference between a new concept and a resized variation; instrumentation and QA of events and deep links; store assets and listing experiments; senior time, where the person who wins the pitch is not the person who runs the account; reporting beyond a platform dashboard export; and the reconciliation work between platform reporting and your own analytics or MMP. Ask which of these are excluded rather than whether the price is negotiable. The exclusions are the price.
Contract and exit terms belong in the pricing conversation
Exit costs are part of the purchase. Settle these before you sign, not at the end:
Ad-account ownership. Accounts should be owned by you, with the agency granted access. Apple documents role-based access and campaign-group-limited permissions (Apple Ads: Invite users to your account) and account and API roles for third-party access that can be reviewed or revoked (Apple Ads: Use the Campaign Management API). Technical revocability is useful, but it does not replace contractual rights to the account and its history.
Raw data access. Exports of campaign, creative, and event-level data in a usable format, on request and at termination.
Creative and source files. Ownership of concepts, edits, project files, footage, and the terms of any creator or stock licensing.
Tool accounts. Which subscriptions are in your name; what happens to historical data in agency-owned tools.
Notice period and transition support. Length of notice, whether a paid transition window exists, and what handover includes.
Markups, rebates, and referral arrangements. Any markup on media or third-party invoices, and any rebate or referral income the agency receives from platforms or vendors.
Approval authority. Who can change budgets, launch campaigns, or alter measurement configuration without written approval.
Knowledge transfer. A documented account structure, event map, naming conventions, test log, and learnings — delivered as a condition of final payment.
When each model does not fit
Fixed project pricing does not fit ongoing optimization, because the work has no natural end and the deliverable becomes a report rather than a running system. Retainers do not fit highly seasonal or stop-start programs, where you pay for capacity in months you cannot use. Percentage of spend does not fit small or deliberately constrained budgets, where the resulting fee cannot staff the work, or very large stable budgets, where the fee outruns the workload. Performance-linked pricing does not fit apps whose key outcome is rare, slow to occur, heavily influenced by product changes, or measured differently by each system. Hybrids do not fit teams without the internal capacity to administer them, because two mechanisms mean two sets of definitions to maintain.
Agency, in-house, or hybrid
Pricing only makes sense against the alternative. In-house is usually preferable when the channel mix is stable, the volume of work is predictable, and you can recruit and retain specialists. An agency is usually preferable when you need several specialisms at partial capacity, or need to move before you can hire. Hybrid — in-house ownership of strategy, budget, and measurement, with external execution capacity for media and creative — is common because it keeps decision rights inside while buying throughput outside. The choice is about capability coverage and retention risk, not fee level. Our agency selection guide covers that decision in more detail, and our growth services and app growth practice describe how the work is structured when it is external.
The proposal comparison checklist
Run every proposal through the same list before you compare any numbers:
Is the fee model stated explicitly, with the base, the variable component, and any caps?
Is every capability in the map assigned to a named owner — including measurement, store work, and lifecycle?
Is creative specified by volume and type, not just "included"?
Are the tools listed, with who pays and in whose name the accounts sit?
Are the platform accounts owned by you, with the agency holding access only?
Is raw data exportable on request and at termination?
Are the people who will actually run the account named, with their time commitment?
Is the reporting cadence defined, including reconciliation between platform reporting and your own analytics?
Are markups, rebates, and pass-through costs disclosed?
Is the notice period, transition support, and handover package written down?
If the fee is performance-linked, are the event, source of truth, window, and dispute process defined?
What is the uncovered column — the work that remains yours — and have you costed it?
If a proposal cannot survive this list, the problem is not its price. If two proposals both survive it, you finally have a fee comparison worth making. Keyword- and channel-level scope questions can be pressure-tested with the same discipline, for example against our Apple Ads keyword strategy guide.
Where a conversation helps
If you are holding competing proposals and cannot tell which is genuinely cheaper, we run a scoped diagnostic conversation for app teams. It covers proposal normalization against the worksheet above, your measurement readiness and what it implies for any performance-linked term, the capability gaps that will remain uncovered whichever vendor you pick, and where the operating boundary between your team and an external one should sit. You receive a written summary of those four items and the questions to put back to each vendor. It is a scoping conversation, not a pitch, and no CPI, CPA, ROAS, ranking, or growth outcome is promised.