AppLovin Discovery campaigns: Discovery vs Prospecting vs Universal

AppLovin Discovery campaigns optimize for purchases from people who have never interacted with your brand: no site visit, no product view, no add to cart. Prospecting optimizes for new customers, which can include people who have visited your site but never bought. Universal optimizes for every purchase, new or returning, valued the same. If your main goal is total ROAS or the lowest cost per purchase, start with Universal. If your goal is net-new demand that your other channels are not already reaching, Discovery is the strategy built for it.

Below we explain what each audience strategy in AppLovin Ads actually optimizes for, what the early data on Discovery shows (and what it does not), and the launch order we use when a DTC brand asks us which one to start with.

The short version

• Universal counts every purchase the same. Best for total ROAS or lowest cost per purchase.
• Prospecting optimizes for new customers. Some of its traffic can still be past site visitors who never bought.
• Discovery optimizes for purchases from people who have never visited your site. It launched in February 2026.
• You can run more than one strategy at the same time, and AppLovin says so in its own setup guide.
• Early third-party numbers on Discovery are promising but come from a small agency test. Run your own incrementality test before you move most of your budget.

What Universal, Prospecting and Discovery optimize for

Since June 2026 the self-serve platform is called AppLovin Ads, and Axon is the name of the AI engine underneath it. When you build a ROAS or Cost Per Purchaser campaign, you pick one of three audience strategies. You are not choosing interests or lookalikes. You are telling the model which purchases count as a win, and the model goes looking for more of them.

Here is how AppLovin itself describes each one, side by side.

The difference between Prospecting and Discovery is the one most people miss. A shopper who clicked a Meta ad last month, browsed two product pages and left is a "new customer" for Prospecting, because they have never purchased. For Discovery, that same shopper is not new at all. Discovery only gets credit for people your brand has never touched.

How Discovery knows who is new

To exclude people who have interacted with you, the model needs to know who those people are. That is why Discovery depends on historical data. According to AppLovin's February 2026 product update, advertisers already running Prospecting campaigns are set up for Discovery, new advertisers share historical purchase data through Ads Manager, and Shopify merchants are covered automatically through the integration.

In practice, the quality of that exclusion depends on your tracking. If the AppLovin pixel is missing events, or fires on some pages and not others, the model has a weaker picture of who has already visited. Before launching Discovery we check three things:

1. The pixel fires on every page view, product view, add to cart, checkout and purchase. AppLovin's launch checklist recommends validating this before launch.
2. The Conversion API is connected. The checklist calls it a source of "stronger optimization signal".
3. On Shopify, the AppLovin app is installed and syncing, so purchase history flows in without a manual upload. Our AppLovin Shopify setup guide covers this step by step.

None of this is specific to Discovery, but Discovery is the strategy most exposed to bad data, because its whole job is to separate "new" from "already seen".

What the early data on Discovery says

Discovery is new, so there is not much public data yet. Here is what exists, and how much weight we would put on each piece.

AppLovin's alpha test. In its launch post (25 February 2026), AppLovin said the share of new visitors more than doubled in many cases and improved significantly in all of its initial test cases. AppLovin did not publish the number of brands, the spend, or cost per purchase in that post. It is the platform describing its own product, so treat it as a direction, not a benchmark.

Common Thread Collective's four-month test. CTC, a DTC agency, published results on 26 May 2026 from four months of testing Discovery across four brands. They reported:

• 85% to 90% of Discovery traffic was new visitors, compared with 60% to 75% for Prospecting.
• Average measured incrementality of 284% for Discovery, or about 190% with outliers removed.
• A prior benchmark of roughly 140% to 150% incrementality for Universal and Prospecting.

CTC was clear about the limits. Their own team said the measurement is "not perfect" because of cookie tracking, results varied widely between brands, and they recommend an incrementality test for each brand. Four brands is a small sample, the brands are not named, and CTC did not publish CPA or ROAS for the campaigns. We read this as good evidence that Discovery does what it says (reaches people your brand has not reached) and early evidence that it can be more incremental. It is not evidence that it will beat your current Universal campaign on cost per purchase.

Fospha's cross-channel view. Fospha, a measurement company, reports from its 2025 client data that AppLovin drove a 39% new customer rate, 26% above the paid media average, and that last-click attribution captured just 31% of AppLovin's ROAS in its model. That data covers AppLovin overall, not Discovery specifically, but it explains why judging a new-customer campaign on last-click ROAS alone will undercount it. We covered this in more detail in our post on AppLovin attribution and incrementality.

Which strategy to launch with

There is no single right answer, but the choice follows from two questions: what is your main KPI right now, and how much of your growth already comes from people who know you? This is the decision guide we use.

1. You are new to AppLovin and need the channel to prove it can pay back. Start with Universal. It has the widest pool of purchases to learn from, so it usually exits learning fastest, and it is the fairest read on what the channel can do for total revenue.
2. Your main KPI is new customer CAC, or you sell a subscription. Start with Prospecting. AppLovin's setup guide names growing the customer base and acquiring new subscribers as the fit for this strategy.
3. You already spend heavily on Meta, Google or TikTok, and you suspect AppLovin is taking credit for buyers those channels found. Start with Discovery, or add it next to Universal. Excluding everyone who has touched your site is the cleanest way to see whether AppLovin is bringing in demand you were not already paying for elsewhere.
4. Your brand is small, with little traffic history. Discovery and Prospecting behave more alike here, because most of the people AppLovin reaches have never visited you anyway. Universal is usually the simpler place to start.
5. You have a long consideration cycle and a high AOV. The audience strategy matters less than the conversion window. AppLovin recommends Day 7 optimization for products like furniture or mattresses, and Day 0 for most other brands.

Whatever you start with, give it enough budget. AppLovin's setup guide suggests a budget that would drive about 10 purchases a day on your existing social channels as a solid starting point. A Discovery campaign starved of budget will not tell you much. Our post on what AppLovin ads cost walks through how to size that number.

Running more than one strategy at the same time

AppLovin's launch checklist says you can run multiple audience strategies at once to balance total purchases, new customer acquisition and net-new visitor volume. Most mature accounts end up doing this. The question is how to split the money.

CTC described one structure for mature accounts: the majority of spend in Discovery, with around 10% in a Prospecting or Universal campaign layered on top. That is one agency's setup after four months, not a rule. We prefer to earn our way into a split like that, in this order:

1. Launch one Universal campaign and let it stabilize at your target ROAS or cost per purchaser.
2. Add a Discovery campaign with its own budget, sized to reach roughly 10 purchases a day at your expected CPA.
3. Use the same creative sets in both at first, so creative is not the variable you are testing.
4. After two to four weeks, compare new visitor share, new customer CAC and your blended numbers (total revenue against total ad spend across all channels), not only in-platform ROAS.
5. Shift budget toward whichever strategy moves the blended numbers, then confirm with a holdout or geo test before going further.

One practical point: expect Discovery to report a lower in-platform ROAS than Universal in many accounts. Universal is allowed to count returning customers and warm visitors, who convert cheaply. Discovery is not. A lower platform ROAS on Discovery is not, on its own, a sign that it is failing.

How to judge whether Discovery is working

Because Discovery is built to find people who have never seen you, the usual dashboard metrics can mislead. These are the numbers we watch, roughly in order of importance.

Also keep an eye on creative. CTC reported that 81% of AppLovin spend in their accounts went to videos over 30 seconds, and that adding many new videos at once caused performance dips, so they add two or three at a time. Cold audiences have never heard of you, so videos that explain the product from scratch tend to matter more here than in a retargeting-heavy campaign. See our AppLovin creative best practices for formats and specs.

Setting up a Discovery campaign, step by step

1. Check tracking. Validate the pixel and connect the Conversion API. On Shopify, confirm the AppLovin app is syncing orders.
2. Confirm purchase history is shared. If you have never run Prospecting, share historical purchase data through Ads Manager so the model can tell new users from past ones.
3. Create the campaign. Choose the ROAS or Cost Per Purchaser buying option, then select Discovery as the audience strategy.
4. Pick the conversion window. Day 0 for most brands, Day 7 for high-AOV products with a longer decision.
5. Set a realistic target. Enter ROAS goals as a percentage (a 2x goal is 200%) and start from your breakeven, not from your best Universal number.
6. Load creative sets. Each set needs at least one video and one interactive or image ad, and can hold up to 10 of each.
7. Budget for learning. Aim for enough daily budget to reach about 10 purchases a day at your expected CPA.
8. Plan the test. Decide before launch how you will measure incrementality, so you are not choosing a method after you see the numbers.

If you want the full pre-launch list beyond audience strategy, our AppLovin launch checklist covers the rest. And if you would rather have a team set up and run the test, that is what we do on our AppLovin management service.

FAQ

What is an AppLovin Discovery campaign?

It is an audience strategy in AppLovin Ads that optimizes only for purchases from people who have never interacted with your brand, meaning no site visit, no product view and no add to cart. AppLovin launched it in February 2026.

What is the difference between AppLovin Prospecting and Discovery?

Prospecting optimizes for new customers, which can include people who visited your site but never bought. Discovery is stricter: it only counts purchases from users who have never visited your site at all.

Can I run Discovery and Universal campaigns at the same time?

Yes. AppLovin's own launch checklist says you can run multiple audience strategies at once to balance total purchases, new customers and net-new visitors. Give each campaign its own budget and compare them on blended results.

What do I need before launching Discovery?

A working AppLovin pixel, ideally the Conversion API, and historical purchase data shared with AppLovin. Advertisers already running Prospecting are set up already, and Shopify stores are covered through the AppLovin app.

Why does Discovery show a lower ROAS than Universal?

Universal can take credit for returning customers and warm visitors, who are cheaper to convert. Discovery only gets credit for people who have never seen your brand, so in-platform ROAS often reads lower even when the campaign is adding more new revenue. Judge it on new customer CAC, blended MER and an incrementality test.