Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Illustrative scenario

Illustrative: An App-Install Campaign on TikTok

An invented fitness app runs a fixed-length app-install campaign on TikTok, with a cost ceiling derived from subscriber value rather than from install cost. The scenario follows the brief, the structure and creative, a weekly log and the decisions it forced. Every figure is made up; the sequence of reasoning is the lesson.
Illustrative example — not a real resulttechnologyUpdated

Illustrative scenario — figures are demo data, not client results

Illustrative scenario

The app, the company, the people and every number on this page are invented to teach a method. Nothing here describes a real client, a real campaign or the author's own results, and none of the figures should be read as a benchmark for what an app-install campaign on TikTok costs or returns. Treat the weekly table as a worked exercise, not as evidence.

About this scenario

This is an invented scenario. The app, which the page calls Brambleway, does not exist; the team, the budget, the installs and the subscriber figures were made up so that the decisions could be shown in order without borrowing anyone's confidential data. Nothing on this page describes the author's own work, and no real advertiser, agency or creator is being described.

In the scenario, Brambleway is a habit-building fitness app for people who have never kept up an exercise routine. It offers short guided sessions, a free trial and then a monthly subscription. The company is small: a founder, a product designer, two engineers and one person responsible for growth. It has run a little paid search and a steady stream of organic short-form video, enough to notice that its best users arrive after watching someone like them get through a first session, not after reading a feature list. That observation is why the growth lead proposes TikTok as the first serious paid channel, and it shapes everything that follows: the creative, the audience hypotheses and the events the campaign is asked to optimise towards. The run is fixed in length and budget from the start, which is what makes the end-of-run decision honest.

The brief

The brief the growth lead writes is short and starts with the economics rather than the channel. From its own billing data the company knows roughly how long a paying subscriber stays and what each one pays, so it can estimate a lifetime value (LTV) per paying subscriber and decide how quickly it needs that value back. The brief turns this into a ceiling: the most the company will pay for a person who starts paying, set so that the money comes back within a tolerable number of months. The LTV and CAC guide covers how that calculation is built; the scenario borrows the method and invents the inputs.

Three choices in the brief matter later. First, the success metric is cost per paying subscriber, a CPA measured at the deepest event, not cost per install; installs are cheap to buy and easy to buy badly. Second, the campaign is a learning budget with a fixed weekly spend and a fixed number of weeks, after which the team decides whether to scale, continue or stop. Nobody may extend it "to see what happens". Third, the audience is described behaviourally rather than demographically: adults who have tried and abandoned fitness apps and respond to beginner-friendly, low-pressure content. That framing comes straight from the comments under the company's organic "first session, no equipment" clips.

The brief also lists what the team will not do: no discount to prop up trial starts, no creative promising physical outcomes, and no judgement on any single week, because trial-to-paid conversion lags the install by the length of the trial. The funnel stages — view, install, trial, payment — are written into the brief so that everyone agrees which stage each metric describes.

Set-up: structure, creative, measurement

Structure. The account follows the pattern in the TikTok campaign structure guide: one campaign under the objective the platform offers for app promotion, with a few ad groups that differ by audience hypothesis rather than by creative. One ad group targets broadly, with only country and age set; one targets interest categories around fitness and wellbeing; one targets people who engaged with the company's organic videos. Budgets sit at ad-group level so money can move between hypotheses, and every ad group carries the same creative set, which keeps the comparison between audiences clean.

Creative. The creative brief applies what the content formats and hooks guide describes: vertical, native-looking video with a hook in the opening moments and on-screen text that works with the sound off. The team commissions a handful of creators to film four concepts — a "day one, first session" point-of-view clip, an "I quit every app until this one" piece to camera, a screen-recorded walk-through of a session, and a founder explaining why sessions are short — each with several alternative hooks. The company's strongest organic posts also run as ads under the method in the Spark Ads guide.

Measurement. App installs cannot be measured with a website pixel. In the scenario the app reports installs and in-app events through a mobile measurement partner integration, with the events the brief cares about — install, trial_start and subscribe — mapped to app events in the way the help centre in the sources describes. The pixel and Events API tutorial still matters for the marketing site, because a share of people tap through to the web page before installing. Two rules are agreed before launch. The measurement partner's attributed figures are the system of record, and the platform's own reported installs are read as directional. And no in-app event is used as the optimisation goal until it is producing a steady flow of events; in practice event-based optimisation needs volume to settle, and the team would rather wait than switch on the first day.

What happened

The table in the example block below is the growth lead's weekly log, with every figure invented. The run is read in three phases. In the first phase the campaign optimises towards installs: installs are plentiful and cheap, but a small share of them start a trial, and cost per paying subscriber sits well above the ceiling in the brief. In the second phase the optimisation goal moves to the trial-start event; installs become dearer and fewer, the trial rate rises, and cost per paying subscriber falls towards the ceiling. In the third phase creative fatigue shows up as falling click-through (CTR) and rising frequency, a refreshed creative batch restores performance, and the final weeks finish under the ceiling.

Illustrative example — not a real result

Brambleway: an eight-week log with invented figures

The weekly budget was fixed at £3,000. Paid subscribers are attributed to the week the trial started, so the final column only fills in after the trial length has passed; the last two rows were read a fortnight after the run ended. The trial lasted 7 days and the subscription cost £9.99 a month.

WeekSpendInstallsCost per installTrial startsTrial ratePaid subscribersCost per paid subscriber
1£3,0001,250£2.4015012%42£71.43
2£3,0001,400£2.1415411%46£65.22
3£3,0001,100£2.7318717%58£51.72
4£3,0001,050£2.8619919%62£48.39
5£3,000900£3.3316218%50£60.00
6£3,0001,150£2.6121919%70£42.86
7£3,0001,200£2.5024020%77£38.96
8£3,0001,180£2.5423620%75£40.00

Over the run: £24,000 spent, 9,230 installs, 1,547 trial starts and 480 paying subscribers, a blended £50.00 per paying subscriber against a ceiling of £45. Weeks 6 to 8 alone cost £9,000 for 222 paying subscribers, or £40.54 each. The ceiling came from invented economics: a median paying lifetime of 7 months at £9.99 gives roughly £70 of revenue, and after store fees and support costs the company counts about £45 of contribution per subscriber, which it wants back within 5 months. Return on ad spend measured on the first payment alone would read as £9.99 against £40, about 0.25, which is why the brief forbade judging the campaign that way. The optimisation goal changed from install to trial_start at the start of week 3; the creative refresh landed at the start of week 6.

Those are the invented numbers; the pattern is what to look at. The cheapest installs of the run were the least valuable, the dearest week by install cost was one of the better weeks by cost per paying subscriber, and a week-five reading taken in isolation would have argued for cutting the budget just as the fix was being filmed. The return on ad spend (ROAS) on first payment stayed below one throughout, as it will for any subscription product with a long payback, which is why the brief chose cost per paying subscriber against an LTV-derived ceiling as the number to manage.

Decisions along the way

Changing the optimisation goal. The first decision was when to stop optimising for installs. The team waited until trial-start events were arriving steadily, then switched one ad group first, confirmed the trial rate moved, and switched the rest a few days later. Installs got dearer; that was expected and accepted, because the metric in the brief sat further down the funnel.

Diagnosing week five. Cost per paying subscriber rose in week five. Two explanations were on the table: the audience was exhausted, or the creative was. Frequency had climbed and click-through had fallen while the trial rate held, which points at creative, not audience. The response was a new creative batch using the two concepts that had performed best, not a wider audience, which would have reset learning and muddied the comparison between ad groups.

Moving budget between ad groups. The interest-based ad group produced installs at a similar cost to the broad one but a lower trial rate, so its budget was moved to the broad and engaged-audience groups halfway through. The engaged-audience group was small and expensive per install but had the best trial rate of the three.

Trusting one source of truth. Platform-reported installs ran ahead of the measurement partner's attributed installs every week. The team reported the partner's figures throughout and did not re-litigate the gap each Monday.

The end-of-run call. Blended cost per paying subscriber was above the ceiling; the final weeks were under it. The team chose to continue at the same budget for another fixed period rather than scale, because scaling changes the auction and the audience, and the under-ceiling weeks were too few to bank on. The CPA calculator was used in that meeting to show how install cost, trial rate and trial-to-paid rate combine into the number that mattered.

What generalises

The method transfers even though the figures do not. Set the success metric at the deepest event the business can measure with reasonable lag, and derive the ceiling for it from subscriber value and payback rather than from what the platform reports as a good cost per install. Keep the campaign structure simple enough that ad groups are audience hypotheses and creative is shared, so that differences in results can be attributed. Decide before launch which measurement source is the system of record, so that the weekly meeting is about decisions rather than reconciliation. Expect to move the optimisation goal down the funnel once the deeper event has volume, and expect install cost to rise when you do. Read the log in phases, not weeks, and never judge a week before its lagged conversions have arrived. When performance drops, separate creative fatigue from audience exhaustion before acting, because the remedies are different and the wrong one costs learning. Treat a fixed-length learning budget as a commitment: the end-of-run decision is only honest if nobody has quietly extended the run. For a subscription app, add one more: first-payment return on ad spend is the wrong lens, and the right one is cost per paying subscriber against a value the finance side agrees with.

What does not generalise

Every number on this page is invented, and the shape of the curve is a teaching device. A real campaign would not produce such a tidy improvement after the optimisation change, nor such a clean recovery after a creative refresh; real logs have noise that swamps single-week differences. The trial rate, the trial-to-paid rate and the subscriber lifetime are properties of this made-up product and its made-up price, and changing any of them changes the ceiling and the verdict. The platform's objectives, optimisation goals and event requirements change over time and vary by account; the help centre in the sources is the reference, and this page deliberately avoids restating thresholds. A larger team with a bigger budget would run more ad groups and a faster creative cadence, and would probably test incrementality rather than rely on attributed installs at all. Categories differ: a game, a finance app and a fitness app have different trial behaviour, different review constraints and different creative norms. Markets differ in cost and in which hooks land. What the scenario shows is a way of deciding; what it cannot show is what any particular app will see when it tries. Other invented scenarios in the case studies library apply the same discipline to different problems.

Terms used on this page

Methods used

Sources

  1. TikTok Ads Help Center — TikTok for Business
← All case studies

Related