App promotion
CPI is one division, and your creative can only touch one side of it. Here is which side, why the testimonial format is unusually good at it, and the number that quietly collapses while you optimize.
CPI stands for cost per install. It is a single division: everything you spent running a campaign, divided by the number of app installs that campaign produced. Spend $2,000, get 1,000 installs, and your CPI is $2. That is the whole calculation, and it is the number most app advertisers watch first because it is the first one that exists.
It is also the number people try hardest to lower, usually by changing the creative. That only works if you know which part of the equation creative can reach.
Every tactic you have ever read about is one of these two.
You pay less for the same impressions. Bidding, targeting breadth, placement mix, seasonality, auction competition. Most of it is media buying, and some is just what the market charges that week. Your video has almost no influence here.
Same impressions, more conversions, so the cost spreads across a bigger denominator. This is the half creative owns. Every improvement a video makes to CPI arrives through this door and no other.
Reduce it far enough and a person scrolling past your ad is answering one thing: does this actually work, for someone like me? Not is this app good, not is this a real company. Does it work, for me, for the thing I am dealing with.
The store page is normally where that gets answered, badly. A listing is written by the company, so the viewer discounts it automatically, and by then the attention is already spent. Anything the ad settles beforehand is a question the store page no longer has to survive.
A testimonial-shaped video settles exactly that, because structurally that is what a testimonial is: one person reporting an outcome in the first person. It is not more persuasive because it is more polished. It is more persuasive because the claim arrives attached to a situation the viewer can check themselves against — the ones who match tap, and the ones who do not keep scrolling.
With no physical product to show, a person's account of what changed is most of what you have. The other half is the screen, covered in how to promote an app with UGC-style ads.
Four parts, in this order. Remove any one and the other three weaken.
The situation the person was in, concrete enough that someone in it recognizes it. This is the qualifier: it decides who is still watching at second four, and you want that group small and correct.
What is different now, stated as a fact about their life rather than a property of the software. "Now I check once on Sunday and I'm done." Outcomes are believable in proportion to how ordinary they are.
Real capture of the screen that produced the outcome. This stops the video being an assertion — someone who has watched the interface do the thing is deciding whether to continue, not whether to begin.
One line naming a limit or a cost. "The free version does the tracking; I pay for the export." Not modesty — credibility. A claim with a boundary reads as a report; one without reads as an ad.
The same benefit written two ways. The right column is the one that converts.
| Big and vague | Why it slides off | Small and specific |
|---|---|---|
| "This completely changed how I manage money." | Attached to no situation. The viewer cannot tell whether their case is covered, so they assume it is not. | "I stopped losing track of subscriptions I'd forgotten about." |
| "Saves you hours every week." | A quantity with no method behind it. Reads as a marketing line because it is one. | "Logging a receipt takes about ten seconds now, at the till." |
| "The best language app out there." | A claim about a market, not about the viewer. Nobody installs an app for winning a category they were not shopping in. | "I can order food and understand the answer. That's as far as I've got." |
| "Effortless and intuitive." | Describes an interface, not a result. Every app in every category says it, which makes it worth nothing to any of them. | "I set it up on the bus. I didn't have to read anything." |
A modest claim invites verification; a large one invites doubt. You want the viewer checking, not appraising.
Almost everything that gets called a testimonial ad is really a demonstration: a person to camera naming a problem, showing the product solve it, and saying what changed. The claims are about your product, you can substantiate them, and the format is the one doing the persuading. That is exactly what we make, and it is the version that scales — ten scripted variations of a demonstration cost around $50 each and land inside 48 hours.
The narrower thing is an endorsement: a named individual attesting to their own results. In the US the FTC's Endorsement Guides treat a testimonial as a statement about an actual person's actual experience, and substantiation sits with you as the advertiser; rules differ by market and get revised, so check what applies where you advertise. Practically, that is a line in the script rather than a limit on the format.
So write claims about the product, not autobiography about the speaker. "This files a receipt in four seconds" is a product claim you can stand behind and we can perform. "I've been using this for six months" is a claim about a person, and it belongs to a real customer — worth collecting from your users anyway, because you will want both. AI UGC vs real UGC covers where each one fits.
Write scripts that make claims about your product, not claims about the speaker. "This is the screen where you split a bill" is yours to say. "This app saved my marriage" belongs to whoever it happened to.
The most useful thing about genuine testimonials is that real users describe your product in words you would never have written. Harvest the language even when you cannot use the person.
CPI is easy to lower if you stop caring what an install is. Broaden the targeting, promise more than the app does, put the word free in the first two seconds. Installs rise, the division gets kinder, and the dashboard looks like progress.
Underneath, you have changed who installs. People arriving on a misread expectation open the app once and leave. Retention and activation fall, the revenue that was supposed to follow does not, and CPI improves throughout.
The second-order version costs more. Your ad platform optimizes toward whatever event you feed it. Point it at installs and it gets better at finding people who install apps — a real behavior, and not the one you want. Point it at a deeper event, like a completed signup, and it learns something closer to your business. The constraint is volume: Meta ad sets need roughly 50 conversions a week to leave the learning phase and stabilize, so a deep event has to fire often enough to sustain that. Platform mechanics change, and their own documentation is the authority.
Never read CPI alone. Put it beside retention and cost per activated user, per creative. A creative with a higher CPI and double the day-7 retention is the better ad, and only the combined view shows you that.
There is no portable benchmark — anyone quoting one is describing their own category, geography and season rather than yours. The only useful comparisons are against your own numbers over time and against what an install is worth to you.
Only through one route: a larger share of the people you already reach choosing to tap. That lever is real and often substantial, but bounded. If your CPI problem is an expensive auction, no video fixes it.
A demonstration is a standard format — a person to camera, talking about a problem, showing a product solving it. What creates exposure is framing a scripted performance as a specific person's verified experience. Keep the claims about the product and you stay on the right side of it.
Enough distinct before-states to find which problem your buyers actually have. Volume beats polish, because you cannot pick the winner in advance. How many UGC ads to test sizes a first round against a fixed budget.
Around $50 a video, quoted per customer based on volume, delivered in 48 hours with script writing, caption variants, one revision round and full commercial rights included. A real creator commonly runs $150 to $500 and takes weeks — what that premium buys is someone who can attest to real experience.
Send us the problems your users describe in their own words. We'll write and produce a video for each, back in 48 hours.
Get a quote