Skip to content

Ad Platforms ยท AppLovin

Last Updated: August 11, 2026

Interactive production

How Tierra builds the Interactive (step 2): the Interactive Builder workflow, format choice, relaunch math, and the hero template. Read alongside the playbook for the platform model and vocabulary. Contribution and lift numbers live in reporting and tracking.

The Interactive is the tappable asset that fires after the video plays. It carries a headline and a call-to-action, and a tap sends the shopper to the landing page. Don't confuse it with the end card (step 3), which shows the catalog or branding.

What the Interactive contributes

On one account, the click-to-purchase path split across the three steps like this:

Step Click share Purchase share
Video 24% 52%
Interactive 65% 36%
End card (catalog) 10% 12%

Interactives drive most of the clicks; videos drive most of the purchases. Both matter. Treat the Interactive as your main lever for clicks and click quality.

Format options

Input formats and their specs come from AppLovin's ad specs and requirements.

  • HTML interactive: best for a multi-step interaction (quiz, swipe, advertorial, presell). It needs a per-account whitelist toggle from AppLovin, and it's harder to mass-produce.
  • Video MP4 Interactive: best for high-production brand storytelling. It treats the Interactive like a short video.
  • Looping MP4 (5 to 15 seconds): auto-loops on the Interactive surface and also works as input to the Interactive Builder. A smooth element-fade or a blank-color bookend both work.
  • Image (GIF, JPEG, or PNG): the simplest to produce, with the least builder friction. GIFs can animate, up to 60 seconds. AppLovin discourages plain statics (they drive platform-leave), but statics beat video on sales for some accounts, so test per account.
  • AI-generated: fast to produce at scale. It runs about half the CTR of hand-built (smaller tap zones), but conversion and CPA can match, so judge it on conversion, not CTR. Watch for false-green CTR on auto-generated units.

Default to video, and test statics against it per account. Neither wins everywhere, so produce both from a winning concept and let the data decide.

Sound

The Interactive plays with sound on, which the video step usually doesn't. Videos get watched muted about 90% of the time, so audio in the video is a bonus at best. On the Interactive, audio actually reaches the shopper, so use it where it helps the interaction (a voiceover cue, a product sound) instead of treating every Interactive as silent.

Interactive Builder workflow

The Interactive Builder is an in-platform AI tool that generates Interactives from input images or video.

  • Run one job per account at a time, and keep the queue feeding so there's always one generating in the background.
  • Run time is highly variable, from about 10 minutes to several hours. Plan for it to be asynchronous.
  • Access needs admin permission on the account. Our seat on a client account often lacks it, so the client may need to grant admin.
  • The quiz Interactive builder needs 4 images, and they have to be in the media library before the builder will accept them.
  • "Create with our studio" is a separate option at the bottom of the Create flow.

Three production patterns that work:

  1. Static-stripped: take a client's existing static image creative, upload it to the Interactive Builder, strip the auto-generated overlays, and use the result as the Interactive.
  2. Native AI generation: trigger the AI Create button and let it run. Judge on conversion, not CTR. The main Interactive Builder is more reliable than rep-built one-offs.
  3. Manual HTML upload: the highest-effort option, with full design control. It needs the per-account whitelist toggle from AppLovin. Advertorial and presell-page-as-Interactive are both viable HTML formats.

Authoring interactives programmatically

When you build the HTML yourself instead of using the Interactive Builder, the CTA text, its color, the logo, and any overlay are baked into the HTML. None of them are exposed as separate API fields, so you set them in the markup and rebuild the asset to change them. Everything the interactive references also has to live inside that one file: non-self-contained HTML (external references or lazyloaded assets) gets reclassified as a plain image, so embed all media directly in the HTML. The API-side mechanics of that reclassification live in dev, so don't restate them here.

A pipeline that turns a live landing, advertorial, or VSL page into an interactive end card:

  1. Fetch the page.
  2. Inline the images and fonts so nothing loads over the network.
  3. Strip every tracker: tag manager, pixel, session replay.
  4. Rewire each CTA to open the destination.
  5. Confirm it renders offline with zero external requests before you ship it.

Call-to-action placement

Put the call-to-action at the bottom or the middle by default, and always include one. It's fine if it overlaps other text or imagery; that trade-off is worth it. Top placement generally isn't recommended.

Relaunch strategy

Once you have a selection of assets live, launch your top 3 to 5 spending videos against your top 4 to 8 CTR Interactives as fresh creative sets. Run it about once a month, twice at most, selecting on the last 7 days of data.

A pre-build buffer helps:

  • Build 20 creative sets ahead of time (10 videos by 10 Interactives), splitting the Interactives across the video pairings.
  • Un-pause 3 a day for a 7-day deployment.
  • Prefix relaunch sets with ECO_R so reporting can separate relaunch performance from net-new performance.

Interactive longevity

Interactives often outlast videos. On one account, a hero-format family launched in December 2025 was still top-spending in May 2026, roughly a five-month run, and Interactive winners can hold $10 to 35k a week for months.

Don't churn a winning Interactive just because it's old. Refresh on a monthly-to-bi-monthly cadence at scale, but keep the winners running in parallel. High-CTR Interactives are durable inventory. They do stale at scale eventually, so the refresh cadence still applies even when winners are durable.

Hero Interactive template

A strong default template, seen across many accounts (adapt it, it's not a hard rule):

  • 4 benefits listed (not 3, not 6).
  • Clean copy, with no explanatory parentheticals. Write "Long-Lasting Energy," not "Long-Lasting Energy (so you stay alert)."
  • Large benefit font. Small benefit text doesn't read at glance distance.
  • A single product image (a clean shot, not lifestyle).
  • Call-to-action at the bottom or the middle.
  • Branding follows the funnel: a product-blind video sales letter (VSL) gets no logo on the Interactive; a branded direct funnel can carry the logo.

Offer alignment

Make sure the Interactive offer matches the landing-page offer:

  • Landing page runs % off, Interactive runs % off.
  • Landing page runs "buy 4 get 4 free," Interactive runs "buy 4 get 4 free."
  • Landing page runs "$99 with code," Interactive runs "$99 with code."

Historically, % off outperforms buy-N-get-N on AppLovin. When the offer is flexible on the client side, push for % off framing.

When the landing-page offer changes mid-flight, either rebuild the Interactive to match or revert the landing page. So avoid seasonal or promo banners on the site that lock the Interactive into a specific timeframe. Align Interactive imagery with the longest-lived on-site element, not a seasonal banner.

Content quality

Custom call-to-action button copy isn't available. The labels are pre-populated dropdown options only, so plan the creative around what's available.

Trick-click Interactives (built to look like a popup or close button so users tap to dismiss) violate AppLovin's ad policy, and we never run them. They spike clicks with junk taps and poison the account's targeting signal.

Don't split-test headlines mid-launch on a single Interactive family. Consolidate to the winning variation or pick a winner before launch. Split-tests at launch waste spend.

Rendering and QA

Advertorial and HTML interactives render inside AppLovin's in-app webview, and that surface has its own failure modes a desktop preview won't show. QA every HTML build in a phone-sized webview before it ships.

  • Page weight is a hook-killer in the webview. Winning advertorial and HTML pages tend to sit around 550KB. Past about 1MB, first paint is delayed enough that the headline never lands before the shopper bounces. Target under 800KB and keep hero images small.
  • The ad webview renders dark, so force a light color scheme or the page can paint unreadable. Check it in a phone-sized dark webview before shipping.
  • Rep-supplied templates sometimes ship with another advertiser's headline copy baked into the animation timeline, which a static check misses. Catch it by taking timed render screenshots, around 1, 3, 6, 10, and 15 seconds in.
  • Date cards that render day/month/year misread in the US: "2/7" reads as February 7, not the 2nd of July. Hold any dated interactive until it's reformatted to month/day/year.

Set composition

The standard creative-set shape, from an AppLovin-sponsored reference:

  • Around 4 videos as AppLovin's baseline, though our rosters often run more in practice.
  • 10 Interactives or fewer per set, typically 5 to 8.
  • Roughly 25 to 30 sets per campaign at typical budget scale.
  • A 10x10 set is 100 nodes, too many for the model to learn from. Stay smaller.

Keep one thing constant across the set and let the rest vary, so you can read what wins. A concept-pure set uses one concept across all its videos; a hook-pure set uses one hook. A set that mixes concepts gives you spend you can't attribute to anything. Encode the constant in the set name.

Pair tests across sets:

  • Top spenders by new Interactives: recycle proven video winners with fresh Interactives.
  • Top CTR Interactives by new videos: recycle proven Interactive winners with fresh videos.
  • Cross-pain-point pairing: for example an arm-focused video with a neck-focused Interactive. One account beat its average with this kind of mixing.
  • Rep-built versus agency-built end cards: point the top spending video concepts at two end-card pools in parallel, rep-supplied product-page interactives against agency-built advertorial interactives, and read D0 ROAS per arm. Recommended as a test, not a settled winner.

Naming conventions

Encode the angle, concept, format, source, and hook in the set name. On one account, baking format, source, hook, and angle into every set name made per-set performance trackable.

Examples from the field:

  • lowest_price, Val_price_discount (angle-named)
  • ECO_top_spend_vids_new_end_cards (relaunch composition, legacy set name)
  • ECO_R_top_spend_vids_top_spend_end_cards (relaunch plus relaunch composition, legacy)
  • ARM_top_spend_3_use_vids_mix_end_cards (vertical plus composition, legacy)

Tokens like end_cards in those names are historical labels baked into live set names. Leave existing set names intact for reporting continuity, and use "Interactive" in new names.