Before a team launches paid acquisition or raises a budget, someone has to answer a plain question: what will this money buy? Forecasting is the part of a mobile user acquisition strategy that answers it with numbers the team can defend in a budget meeting.
A mobile user acquisition forecast estimates how a media budget may translate into installs, activated users, customers, and revenue using historical or benchmark assumptions for acquisition cost, conversion, retention, and LTV.
A forecast does not promise a result. Every input is an estimate, so a useful forecast shows a range of outcomes and names the assumption behind each number. This guide builds one step by step. By the end, you can estimate installs, activated and paying users, CAC, revenue and payback for a planned budget. You will also see how far each number moves when costs rise or conversion slips.
What is a mobile user acquisition forecast?
A mobile user acquisition forecast is a planning model that turns a budget into expected installs, activated users, paying customers, revenue and payback. It covers one cohort over a defined period, and each output is a range tied to a written assumption.
Five words get used as synonyms in planning meetings, and the confusion is expensive. A target is what the business wants, such as 1,000 new subscribers a month. A budget is what it is willing to spend. A benchmark is a reference point from outside the current campaign: a category report, a partner’s data or last year’s cohorts. A forecast is what the team expects the budget to produce under its assumptions. The actual is what the campaign delivered.
Tuning a forecast until it hits the target defeats its purpose. The gap between ambition and evidence then stays hidden until the actuals arrive. Keep the target in its own cell and let the model show whether the budget can reach it.
The model also has a short shelf life. Each new batch of campaign data replaces an assumption with an observation, so it is rerun as cohorts mature.
Table 1. Forecast terminology
| Term | Meaning | Example use |
|---|---|---|
| Forecast | Expected range based on assumptions | Planning |
| Target | Desired business result | Goal setting |
| Budget | Spend the team has approved | Model input |
| Benchmark | External or historical reference | Initial assumption |
| Actual | Observed campaign result | Model recalibration |
What inputs do you need before building the forecast?
A mobile UA forecast needs five groups of inputs: the media budget and channel mix, acquisition cost assumptions, post-install conversion rates, retention, and revenue or LTV for a defined window. It also needs the measurement settings that define them: attribution windows and the organic baseline. Note the source of each one.
Table 2. Input checklist
| Input | What to collect | Preferred source |
|---|---|---|
| Budget and channel mix | Total media budget, split by channel, platform and geo; other costs in or out | Finance plan, media plan |
| Acquisition cost | CPM, CPC, CPI and CPA by channel and spend level | Ad platforms, MMP cost data |
| Post-install conversion | Install-to-registration, activation, trial, first purchase | Product analytics, MMP events |
| Retention | Retention rate at the checkpoints your model needs (D1, D7, D30) | Cohort reports |
| Revenue and LTV | Net revenue per customer for a stated window | Store reports, billing data |
| Measurement settings | Attribution window per source; organic baseline | MMP configuration |
Media budget and channel mix
Start with the total media budget for the forecast period, then split it by channel. Each channel needs its own cost and conversion assumptions, because social, in-app networks and Apple Ads rarely deliver users at the same price or quality.
Decide early whether management fees, creative production and technology costs (measurement partner, analytics, fraud tools) sit inside acquisition spend, then apply that choice to every period. Decide as well whether the model covers paid acquisition alone or blended growth; a paid forecast keeps organic installs on a separate line.
Acquisition cost assumptions
Cost enters the model at several levels. CPM (cost per mille, the price of 1,000 impressions) and CPC (cost per click) describe the media. Cost per install (CPI) prices the install, and cost per acquisition (CPA) prices a defined post-install action such as registration.
To forecast mobile user acquisition cost, start from historical cost by channel, express it as a range, and adjust that range for the spend level you plan to reach. A single blended number hides channel differences: a cheap source may stop scaling at a fraction of the budget while an expensive one carries most of the volume. Metric definitions and target-setting are covered in our guide to app user acquisition cost.
Post-install conversion assumptions
List every step between install and revenue that the model needs: install-to-registration, registration-to-activation, trial start, first purchase or subscription. Then name the meaningful product event behind your activation rate, such as a first completed workout within three days. Reuse the definitions your analytics team already reports, as described in our mobile app metrics framework.
Retention, revenue, and LTV assumptions
The retention window has to match the product model: a monthly subscription needs its retention rate for each billing cycle, while a game funded by ads needs daily engagement curves. An LTV (lifetime value) figure needs its calculation window, because 180-day and three-year LTV can differ several times over for the same users.
Prefer your own historical cohorts to generic benchmarks. A cohort is a group of users acquired in the same period through the same source, and its observed curve is the most defensible input available. Use revenue net of store fees. On the US App Store, a developer receives 70% of a subscription price in the subscriber’s first year and 85% after a year of paid service. Google Play charges a 10% service fee plus a 5% billing fee on auto-renewing subscriptions for US users from June 30, 2026.
The basic mobile UA forecasting formulas
The model follows one cohort down the funnel, multiplying each step by a rate. Every arrow in Figure 1 is an assumption to test separately. The examples below use the base case of one US fitness subscription app with a $50,000 monthly media budget; the worked example later adds the other two scenarios.
Figure 1. Acquisition funnel, base scenario of the fitness subscription app example
Forecast installs
Forecast installs = media budget ÷ expected CPI
To forecast app installs, divide the media budget by the expected CPI, and run the calculation for a CPI range because CPI is an assumption. Example: $50,000 ÷ $4.00 = 12,500 installs. With several channels, calculate installs per channel and add them up. Dividing the total by a simple average of channel CPIs misstates volume whenever budgets or prices differ.
Forecast activated users
Activated users = forecast installs × activation rate
To forecast activated app users, multiply forecast installs by the activation rate observed in comparable cohorts. Here, 12,500 × 40% = 5,000 activated users.
Forecast paying users
Paying users = forecast installs × install-to-purchase rate
The install-to-purchase rate is the activation rate multiplied by the activated-to-paid rate: 40% × 12.5% = 5%, so 12,500 × 5% = 625 paying users. Modeling both steps keeps activation inside the chain, so weaker onboarding shows up in revenue.
Forecast acquisition cost per customer
Forecast CAC = total acquisition spend ÷ forecast new customers
Example: $50,000 ÷ 625 = $80. Customer acquisition cost (CAC) prices the customer, so keep it apart from CPI and CPA in every report.
Forecast revenue
Forecast revenue = forecast acquired customers × expected revenue per acquired customer
With $96 of net revenue per customer over 180 days, 625 × $96 = $60,000. That is a 180-day ROAS (return on ad spend) of 120%.
Three rules keep these formulas honest. Every rate must describe the same cohort and time period: a D7 purchase rate multiplied by twelve-month revenue describes no real group of users. Any comparison of D7 conversion with lifetime revenue needs an explicit note about the gap. Finally, management fees and production costs are either tracked separately or written into total acquisition spend. Adding $8,000 of creative and fees in the example lifts the fully loaded CAC from $80 to $92.80.
Build three scenarios instead of one prediction
A forecast should use multiple scenarios because cost and conversion inputs are uncertain, and small errors in each one multiply through the funnel. Three scenarios show the range a budget can realistically produce and make the risk visible before money is committed.
The base scenario uses the expected value of each input. The conservative scenario takes plausibly weak values for each input; the aggressive one takes plausibly strong values. The media budget stays the same across all three, so the difference in outcomes comes from assumptions alone.
The width of each range should come from your own data. One practical method: take the last eight to twelve comparable weekly cohorts, use the median for the base case, and use the weaker and stronger quartiles for the other two.
When history is thin, document the reasoning behind each range so reviewers can challenge it. Write the scenario inputs next to each other in one table, as below, so a reviewer can see at a glance which assumptions differ and by how much. Then compute every output for all three columns with the same formulas.
Table 3. Three-scenario structure
| Input | Conservative | Base | Aggressive |
|---|---|---|---|
| Media budget | Same | Same | Same |
| CPI | Higher | Expected | Lower |
| Activation rate | Lower | Expected | Higher |
| Purchase rate | Lower | Expected | Higher |
| Retention | Lower | Expected | Higher |
| Revenue/LTV | Lower | Expected | Higher |
Avoid building the conservative case by pushing every input to its historical worst at once. Inputs that are independent rarely fail together, so stacking extremes produces a case nobody will plan around. Some inputs do move together: cheap inventory often brings weaker conversion, so a low CPI and a high purchase rate in the same scenario deserve a second look.
Worked example: forecasting results from a mobile UA budget
The fitness subscription app from the formulas section plans one month of paid acquisition in the US across iOS and Android. Activation means completing a first workout within three days of install. A paying user is one charged after a seven-day trial, within 30 days of install. Revenue is net of store fees and counted over 180 days. Every input below is an example assumption chosen to show the method.
The base column repeats the calculation from the formulas section. The other two columns run the same formulas on weaker and stronger inputs. User counts in this and the following tables are rounded to whole users, and CAC and revenue are calculated from the rounded counts.
Payback comes from the cumulative revenue curve. In the base case, net revenue per customer reaches $30, $48, $62, $74, $86 and $96 at the end of months one to six. It first exceeds the $80 CAC in month five, so the projected payback period is five months. Payback here compares net revenue after store fees with media-only CAC. Teams that measure payback on contribution margin also subtract variable costs such as payment processing, infrastructure and support, which moves payback later.
Table 4. Worked forecast: fitness subscription app example
| Metric | Conservative | Base | Aggressive |
|---|---|---|---|
| Budget | $50,000 | $50,000 | $50,000 |
| CPI | $4.80 | $4.00 | $3.40 |
| Installs | 10,417 | 12,500 | 14,706 |
| Activation rate | 35% | 40% | 45% |
| Activated users | 3,646 | 5,000 | 6,618 |
| Purchase rate (install to paid) | 4.0% | 5.0% | 6.0% |
| Paying users | 417 | 625 | 882 |
| CAC | $119.90 | $80.00 | $56.69 |
| Net revenue per customer, 180 days | $80 | $96 | $110 |
| Forecast revenue | $33,360 | $60,000 | $97,020 |
| 180-day ROAS | 67% | 120% | 194% |
| Projected payback | Beyond 6 months | Month 5 | Month 3 |
No single input moves by more than half between the conservative and aggressive cases, yet paying users more than double and revenue nearly triples. That is the compounding effect of a multiplicative funnel.
The input with the most influence is the one with the widest plausible range. Here the purchase rate spans 4% to 6%, a 50% spread, while CPI spans $3.40 to $4.80, about 41%. The purchase rate is also the input confirmed last: CPI is visible within days, purchase conversion only after trials end.
Look at the conservative column again. It still delivers more than 10,000 installs, which reads as a healthy month in an install report. Yet those users return $33,360 on $50,000 of spend within 180 days and do not pay back inside the window. Install volume cannot show whether growth funds itself, which is why buying installs is no longer enough.
Download the forecast template (XLSX) to run these calculations with your own inputs. It contains the three scenarios, spend tiers, a sensitivity grid and a forecast-versus-actual sheet, filled in with this example.
How to account for rising acquisition costs as you scale
Scaling often raises CPI because larger budgets reach beyond the most responsive audiences and into less efficient inventory. The effect is common, but it is neither universal nor linear, and only your own spend-level data shows its shape.
Several forces drive it. Audience saturation means the best-matched users have already seen the ads. Diminishing returns follow as each extra dollar buys fewer installs. Campaigns expand into broader placements, the channel mix shifts toward sources that were second choices, and creative fatigue lowers click-through rates. Higher bids raise competition in the same auctions, geographic expansion adds markets with different costs, and the conversion quality of marginal users often declines.
Two numbers need to be kept apart here. Blended CPI is total spend divided by total installs. Marginal CPI is the cost of the installs bought by the next increment of budget. Blended CPI moves slowly because it averages new, expensive installs with the cheap ones already bought.
Table 5. Spend tiers instead of one CPI (example values)
| Spend tier | Expected CPI range | Reason |
|---|---|---|
| Initial budget ($0–50k) | Historical efficient inventory: $3.80–4.40 | Proven audiences |
| Expansion budget ($50–80k) | Moderately higher uncertainty: $4.40–5.40 | Broader inventory |
| Scale budget ($80–120k) | Wider range: $5.20–6.80 | Saturation and channel expansion |
Figure 2. Marginal and blended CPI by spend tier, fitness subscription app example
In the example, raising the budget from $50,000 to $80,000 buys 6,122 extra installs at a $4.90 marginal CPI, and blended CPI only moves to $4.30. With a 4.5% purchase rate on those marginal users, the extra $30,000 acquires 275 customers at $109.09 each, while blended CAC reads $88.89. A further $40,000 at $6.00 buys 6,667 installs and 267 customers, a marginal CAC of $149.81. Approve each tier on its marginal CAC, and expand one source or geo at a time so the effect of each step stays readable.
Run a sensitivity analysis before approving the budget
A sensitivity analysis changes one input at a time and shows how far the outcome moves. It answers a simple question for the person approving the budget: which assumption, if wrong, breaks the plan?
Table 6. What happens to CAC and expected revenue if CPI rises while conversion remains unchanged
| CPI | Installs | Paying users | CAC | 180-day revenue | 180-day ROAS | Payback |
|---|---|---|---|---|---|---|
| $4.00 (base) | 12,500 | 625 | $80.00 | $60,000 | 120% | Month 5 |
| $4.40 (+10%) | 11,364 | 568 | $88.03 | $54,528 | 109% | Month 6 |
| $4.80 (+20%) | 10,417 | 521 | $95.97 | $50,016 | 100% | Month 6 |
| $5.20 (+30%) | 9,615 | 481 | $103.95 | $46,176 | 92% | Beyond 6 months |
If finance requires payback within six months, the base plan survives a 20% CPI increase but not a 30% one. That turns a vague worry into a concrete question for the media team: how likely is CPI to exceed $4.80 at this budget?
Table 7. A 10% adverse move in each input, base case
| Input moved by 10% | Paying users | CAC | 180-day revenue | Payback |
|---|---|---|---|---|
| CPI up to $4.40 | 568 | $88.03 | $54,528 | Month 6 |
| Activation rate down to 36% | 563 | $88.81 | $54,048 | Month 6 |
| Activated-to-paid down to 11.25% | 563 | $88.81 | $54,048 | Month 6 |
| Retention: 10% fewer paid months | 625 | $80.00 | $54,000 | Month 6 |
| Revenue per paid month down 10% | 625 | $80.00 | $54,000 | Month 6 |
In a multiplicative model, a 10% adverse move in almost any input costs about 10% of revenue. The paths differ. Cost and conversion inputs raise CAC, while retention and revenue per user leave CAC unchanged and stretch payback. The inputs to watch most closely are the ones that are both volatile and slow to confirm.
How to forecast a new app with little or no historical data
A new app can forecast user acquisition without historical data, but the forecast starts as a wide range built from outside references. A controlled validation budget then narrows it, so treat the first campaign as a measurement exercise.
- Use ranges, not point estimates, for CPI, activation, purchase rate and revenue.
- Separate known inputs (price, store fees, budget) from assumptions (CPI, conversion, retention).
- Use category benchmarks cautiously, and only with their country, platform, category and date.
- Run a controlled validation budget large enough to produce enough paying events to measure conversion.
- Update assumptions after the first cohorts mature, starting with CPI and activation.
- Avoid copying another app’s CPI or LTV without its market, audience and monetization context.
- Document attribution and observation windows for every source.
Windows differ by default. AppsFlyer applies a seven-day click and one-day impression lookback when an attribution link carries no window parameter. Apple’s AdAttributionKit counts click-through installs within 30 days and view-through installs within 24 hours. If an app already has organic users, their behavior is a useful starting point, which is why product readiness before scaling acquisition belongs in the first planning conversation.
At ROCKAPP, the inputs agreed before a test launch usually include the target event and the platform and geo split. They also cover attribution settings, any historical or organic cohorts, and the cost scope of the budget.
Without data, the main result of the first campaign is reduced uncertainty: a narrower CPI range, a measured activation rate and the first points of a revenue curve.
How to compare forecast and actual acquisition results
Update a UA forecast whenever meaningful new campaign or cohort data arrives. Cost and volume assumptions get reviewed every week or campaign cycle; revenue assumptions get reviewed when a cohort window such as D30 matures.
The review compares planned spend with actual spend, forecast CPI with actual CPI, and forecast installs, activation and customer counts with observed ones. It also compares forecast CAC with actual CAC and predicted revenue or LTV with observed cohort value. Forecast variance is the actual divided by the forecast, minus one.
Table 8. Forecast versus actual review, run 30 days after the last install of month one (example)
| Metric | Forecast | Actual | Variance | Likely cause | Next action |
|---|---|---|---|---|---|
| Spend | $50,000 | $48,200 | -3.6% | Delivery capped on one source | Check bids and caps first |
| CPI | $4.00 | $4.35 | +8.8% | Fatigue on the lead creative | Rotate new concepts |
| Installs | 12,500 | 11,080 | -11.4% | Higher CPI, lower spend | Recalculate at the new CPI |
| Activation rate | 40% | 37% | -7.5% | Onboarding change mid-cycle | Compare cohorts before and after |
| Paying users (D30) | 625 | 512 | -18.1% | Fewer installs, lower activation | Hold budget flat until activation recovers |
| CAC | $80.00 | $94.14 | +17.7% | CPI and conversion both weaker | Update both assumptions |
| Net revenue per customer (M1) | $30 | $31 | +3.3% | Plan mix as expected | Keep the curve; recheck at M3 |
Check timing before drawing conclusions. A cohort whose window has not closed is incomplete, and AppsFlyer, for example, marks such periods as partial data in cohort reports. Platform dashboards and your measurement partner may also count conversions under different attribution windows, so pick one system of record per metric.
iOS needs its own reading. AdAttributionKit splits the first 35 days after launch into three conversion windows (days 0–2, 3–7 and 8–35) and sends a postback after each one closes: 24 to 48 hours after the first, 24 to 144 hours after the second and third. Fine-grained conversion values appear only in the first postback and only at higher crowd anonymity tiers; otherwise the value is coarse or absent. A complete iOS picture of a cohort therefore arrives around day 36 to 41, later than the D30 review in Table 8. Until then, part of the observed iOS result is aggregated or modeled. Reconcile it with server-side revenue, and judge iOS variance at channel or campaign level, where volumes are large enough to read.
Treat a forecast error as information. Move the assumption that caused it, note the change and its date, and let the next forecast start from better inputs. Editing the forecast column to match the actuals erases exactly what the review was meant to teach.
Common mobile user acquisition forecasting mistakes
Most forecasting errors come from a short list of habits. Each one makes the plan look more certain than the evidence allows:
- Treating a forecast as a commitment. A range presented as a promise invites pressure to hide misses.
- Using one fixed CPI at every spend level. The plan overstates installs at exactly the budget levels where most money is at risk.
- Forecasting installs without user quality. A plan can hit its install goal and still lose money.
- Mixing data from different attribution windows. A source measured on a 30-day click window looks stronger than one measured on 7 days, with no change in real performance.
- Ignoring organic uplift or double-counting it. Ignoring it undervalues paid media; counting it twice inflates it.
- Using lifetime revenue with short-term costs without explanation. The plan looks profitable on revenue that may arrive years later, if at all.
- Relying on generic benchmarks instead of product data. Another app’s numbers reflect its own audience, pricing and onboarding.
- Omitting creative-production and technology costs. CAC jumps the moment finance adds them back.
- Assuming all channels scale equally. The cheapest source usually saturates first, so its CPI cannot carry the whole budget.
- Failing to update the model after launch. Decisions keep resting on assumptions the campaign has already disproved.
Mobile user acquisition forecast checklist
Before the budget goes for approval, confirm each item:
- Objective defined: the business result the budget must support, kept separate from the forecast
- Forecast period defined, for example one calendar month of spend
- Paid and organic users separated, with any expected organic uplift on its own line
- Budget states explicitly which additional costs it includes or excludes
- Channel assumptions documented with their sources and dates
- CPI range used for each spend tier
- Activation and customer events defined in the same way on iOS and Android
- Revenue window defined and net of store fees
- Conservative, base and aggressive scenarios created
- Scaling effect considered with marginal CPI
- Actual-versus-forecast review scheduled, with an owner for each assumption
A useful forecast is a decision model, not a promise
Uncertainty cannot be removed from a forecast, only measured. The model earns its place when it helps a team compare scenarios, decide what to test first and choose which budget tier to approve. Its accuracy comes later, from actual cohort data feeding back into every assumption.
If your team needs to connect forecasting with channel planning, media buying, creative testing, and post-install quality, explore ROCKAPP’s mobile user acquisition approach.
