adlibrary.com Logoadlibrary.com
Share
Advertising Strategy,  Platforms & Tools

Scroll Depth Retargeting Facebook: Optimize for Deep Readers

Sparse purchases starve Meta's algorithm. Fire scroll depth events at 50/75/90%, build deep-reader audiences and optimize high-ticket offers on reads.

Scroll depth retargeting facebook: article progress markers at 50, 75 and 90 percent feeding a deep-reader audience

Scroll depth retargeting facebook is the fix for offers that sell too rarely for Meta to learn from: fire a custom event when a visitor reaches 50, 75 and 90 percent of your article, build custom audiences from those events, and for high-ticket offers optimize toward the deepest readers instead of the rare purchase.

Reading is a behavior that happens hundreds of times a day, even when sales happen twice.

TL;DR: When purchases or booked calls are too sparse for Meta's delivery system to learn from, read depth becomes the proxy. Fire a trackCustom event at 50/75/90% scroll through the Pixel and the Conversions API (or GTM's scroll depth trigger), build custom audiences of deep readers, retarget the 75%+ group for ecommerce, and use deep reads as the optimization signal for high-ticket offers. Treat read-rate benchmarks as rules of thumb, and gate every event behind consent.

Why sparse purchases starve the algorithm

Meta's delivery system is a prediction engine. It looks at who converted, finds patterns in who they were and what they saw, and shows your ad to more people who look like them. That only works when there are enough conversions to find a pattern in.

Meta's own guidance puts the learning threshold at roughly 50 optimization events per week per ad set, as described in the learning phase help article. Below that, the ad set sits in learning limited and delivery stays noisy.

If you want the full mechanics, the breakdown in Meta Ads Learning Phase: 50 Events Per Week Explained goes deeper, and the learning phase calculator turns your budget and CPA into a weekly event estimate.

Run the numbers on a typical high-ticket funnel. You spend 500 euros a day on traffic to a long-form article, and the offer is a 3,000 euro coaching program sold through a booked call.

Say you land one to two booked calls a day. That is 7 to 14 optimization events a week against a 50-event requirement. The algorithm is guessing from a sample the size of a dinner party.

That event is reading. A visitor who reaches 75 percent of a 2,000-word article did something costly in attention terms. A visitor who bounced at 8 percent did not.

The gap between those two is the signal you are missing, and it is available hundreds of times a day instead of twice. Scroll depth retargeting Facebook setups exist to capture it.

Scroll depth retargeting facebook: reading as a signal

The idea is simple. Measure how far each visitor gets in your article, report the milestones to Meta as custom events, and use those events in two places: audiences and optimization.

In a scroll depth retargeting facebook setup, audiences are the safe use. A custom audience of people who reached 75 percent is a warm pool that Meta did not have to guess at, because they proved their interest on your own page.

Optimization is the aggressive use, and it only makes sense when conversions are too thin to optimize on directly.

Why does reading work as a proxy at all? Think about who finishes a long article that arrived from a cold ad. They either have the problem or they are curious about it.

Both groups convert at a higher rate than someone who closed the tab. That ordering holds across most niches, and it is why I treat scroll depth as a ranking of intent rather than a measurement of it.

The proxy has limits, and it is worth being blunt about them. Reading does not equal buying. A well-written article attracts readers who enjoy reading and never intend to spend.

If you optimize hard toward depth without checking downstream outcomes, Meta will find you more people who like long content. The fix is a feedback loop: compare deep-reader audiences against actual sales or calls on a regular cadence, and drop the proxy when real volume arrives.

One boundary before we go on. How the article itself is built, from the pain hook to the 1-2 structure, is a separate topic and lives in Advertorial Landing Pages: When Education Beats the Product Page.

General retargeting strategy is covered in Retargeting in 2026 and Facebook retargeting ads. This post is only about the measurement layer: the event, the audience, and the optimization choice.

Fire the scroll event with the Meta Pixel

Start with the browser side. The Meta Pixel supports custom events through fbq('trackCustom', ...). According to Meta's custom event documentation, custom event names must be strings of no more than 50 characters, and you can pass an optional parameter object. Keep names plain and consistent. Read50, Read75, Read90 will do.

Here is a working snippet. It measures progress through the article element rather than the whole page, so footers and comment sections do not inflate the number.

It fires each milestone once per page view, ignores fast flick-scrolls by requiring a minimum time on page, and attaches an eventID so the browser event can be deduplicated against the server copy later.

html
<script>
(function () {
  var article = document.querySelector('article');
  if (!article) return;

  var milestones = [50, 75, 90];
  var fired = {};
  var start = Date.now();
  var pageViewId = (window.crypto && crypto.randomUUID)
    ? crypto.randomUUID()
    : String(start) + Math.random().toString(16).slice(2);
  var params = new URLSearchParams(location.search);

  function readPercent() {
    var rect = article.getBoundingClientRect();
    var seen = window.innerHeight - rect.top;
    return Math.max(0, Math.min(100, Math.round((seen / rect.height) * 100)));
  }

  function report(m) {
    var eventId = pageViewId + '_' + m;
    var data = {
      depth: m,
      article: location.pathname,
      seconds: Math.round((Date.now() - start) / 1000),
      ad_set: params.get('utm_content') || 'unknown'
    };
    // Browser side: only call this after consent is granted.
    fbq('trackCustom', 'Read' + m, data, { eventID: eventId });
    // Server side: first-party endpoint that forwards to the Conversions API.
    navigator.sendBeacon('/api/read-event', JSON.stringify({
      event_name: 'Read' + m,
      event_id: eventId,
      event_source_url: location.href,
      custom_data: data
    }));
  }

  function check() {
    var p = readPercent();
    var elapsed = (Date.now() - start) / 1000;
    milestones.forEach(function (m) {
      // 10 seconds per 25% of depth filters out flick-scrolls
      if (!fired[m] && p >= m && elapsed >= (m / 25) * 10) {
        fired[m] = true;
        report(m);
      }
    });
  }

  window.addEventListener('scroll', check, { passive: true });
  setInterval(check, 2000);
})();
</script>

A few choices in that snippet are deliberate. The time gate (10 seconds per 25 percent of depth) is a crude bot and flick filter, and you should tune it to your article length.

The ad_set parameter pulls from utm_content, so you can later see read rate per ad set without joining tables. And the setInterval fallback catches visitors whose page loads already scrolled, which happens on mobile restores.

Before you ship, open Events Manager and confirm the three events show up under your pixel. Our guide to Meta Events Manager shows where custom events live and how to read the test events tab. If you are still wiring the base setup, How to Set Up Meta Pixel + CAPI in 2026 covers the stack from scratch.

One more constraint matters for the optimization step later. Meta's documentation notes that custom events can define custom audiences, but the page points to custom conversions, created in Events Manager, as the route for campaign optimization. Custom conversions are limited to 100 per ad account. Plan your three milestones accordingly and do not create one per article.

The GTM version and the Conversions API twin

If your site runs Google Tag Manager, you can skip the custom script. GTM has a built-in Scroll Depth trigger. You enter thresholds as percentages or pixels, for example 50, 75, 90, and the trigger fires once per threshold.

A built-in variable, Scroll Depth Threshold, tells your tag which milestone fired, so one Meta tag can send Read plus the threshold value.

The GTM route has one catch. By default the trigger measures the whole page, not your article. On a page with a long footer or related-posts block, a 90 percent reading means something different from what you intended.

If the article is most of the page, GTM is fine. If not, use the custom script above, which anchors to the article element.

Now the server side. The Conversions API lets you send events from your server to Meta rather than relying only on the browser. Meta's documentation says to send an event_id with every event, and that when Meta receives the same event_id and event_name more than once, it keeps only the first copy.

That is how pixel deduplication works, and it is why the snippet above uses the same ID for both paths.

The flow looks like this:

  1. The browser detects the milestone and fires the Pixel event with an eventID.
  2. The same script sends the same payload to your own endpoint.
  3. Your server forwards it to Meta with the matching event_name and event_id, plus whatever hashed identifiers you hold.
  4. Meta keeps one copy and discards the duplicate.

Two details from the docs matter in practice. event_time can be up to 7 days old when you send it, and if any event in a batch is older, the whole request errors. And action_source must be accurate, so use website for browser-originated reads.

Scroll depth is observed in the browser, so the server cannot detect it by itself. What the server path adds is resilience. Browser events can be lost to ad blockers, strict tracking prevention, page unloads and consent banners that load late.

A first-party beacon is harder to block and survives more of those failures. For the full build, see Meta Conversions API: The Complete 2026 Implementation Guide, and check your match quality with the EMQ scorer.

Scroll depth retargeting facebook audiences and windows

Once the events flow, build audiences. In Ads Manager, create a custom audience from your website, pick the pixel, and choose the custom event as the rule. You can then set a retention window.

Meta's website custom audience documentation lists retention as 1 to 180 days. Some third-party guides cite longer limits for certain event types, but I could not confirm those against Meta's own docs, so I plan around 180.

Do not build one audience. Build a ladder:

  • Read50, last 30 days. Broad warm pool. Large enough to run, weak enough that you should treat it as a prospecting seed.
  • Read75, last 30 days. The core retargeting audience for most offers.
  • Read90, last 14 days. Your tightest, hottest group. Often small, so check audience size before relying on it.
  • Read75 minus purchasers or booked calls. Always exclude people who already converted.

Window choice follows the sales cycle. A 49 euro impulse product has a short decision window, so 7 to 14 days is plenty.

A 3,000 euro program has a long consideration period, so extend the deeper audiences to 60 to 90 days and run them against a slower creative cadence. Frequency matters, because a small Read90 pool burns out in days if you push it hard. The frequency cap calculator helps size the daily reach.

Exclusions deserve as much thought as inclusions. Layer Read90 over Read50 so the same person is not in two ad sets competing in the same auction. If you do, you pay against yourself. The structure ideas in Advanced Retargeting Segmentation by Market Awareness pair naturally with a depth ladder, because reading depth is itself a rough proxy for awareness stage.

People who read to the bottom may have come to you anyway, the mid-funnel trap described in Warm Audience. Test with a holdout, as laid out in Meta incremental attribution, before you credit the audience with sales. Once the audience proves itself, how much to increase Facebook ad budget sizes the step up.

Sparse purchase dots compared with a dense stream of reader events used as a proxy signal

Ecommerce: retarget the 75% readers

Ecommerce accounts have a different problem with scroll depth retargeting facebook: purchases are frequent enough to optimize on, so the proxy is not rescuing delivery. The proxy is there to sharpen your retargeting.

Take a supplement or skincare brand sending cold traffic to an article about a specific problem. Most visitors never buy on the first visit.

The ones who read 75 percent of a piece about, say, afternoon energy crashes have told you something no interest target can: they recognize the problem. That group is worth separating from the ones who skimmed the headline and left.

The playbook is short:

  1. Keep the main campaign optimizing on purchases, as usual.
  2. Build a Read75 audience with a 14 to 30 day window and exclude purchasers.
  3. Run a retargeting ad set with different creative from the cold ads: a testimonial, a before-and-after, a comparison against the alternative the reader was probably considering.
  4. Send them to the product page rather than back to the article.
  5. Compare the Read75 ad set's cost per purchase against your general website-visitor retargeting. If the depth audience is not meaningfully cheaper, the signal is not earning its keep.

If your funnel has a cart stage, depth audiences complement it rather than replace it. A visitor who read 75 percent and added to cart sits in a better bucket than either alone.

The benchmarks in add to cart rate benchmark help you decide whether the page after the article is leaking. If it is, landing page testing for Facebook ads shows how to compare replacements.

High ticket: optimize on deep reads

High-ticket and lead-gen offers are where this approach earns its place. A call-booking funnel on a 500 euro daily budget might deliver one or two qualified calls a day, far under what Meta needs. The ad set cannot learn from that, so it wanders.

The workaround is to raise the volume of the optimization signal while keeping it correlated with intent. In practice that means optimizing toward Read75 or Read90 events. If 1,000 euros of daily traffic produces 500 clicks and one in five reaches 75 percent of the article, you now have around 100 events a day, or 700 a week.

That clears the 50-event bar by an order of magnitude, and the algorithm has something to learn from. (These numbers are illustrative, so substitute your own click and read rates.)

Here the documentation limit bites. Per Meta's docs, custom events are not directly selectable as an optimization target, and the route described is custom conversions built in Events Manager. Those are defined by rules, and the documented example matches on URLs recorded with page views.

Which rule types your account offers can differ, so check the Events Manager interface before you design around any one approach. Many media buyers also send a standard event, with the depth in a parameter, so that a standard optimization goal applies. Pick whichever your account supports and test it on a small budget first.

Protect yourself against three failure modes:

  • Proxy drift. After a few weeks, check whether Read90 volume tracks booked calls. If reads climb while calls flatten, you are buying readers, not buyers.
  • Audience capture. Meta may find a cheap pocket of people who scroll deeply on mobile and never respond to a sales conversation. Review lead quality by ad set, and look past the raw event count.
  • Over-trust. Read depth is an early signal. Once you have 30 to 50 real conversions per week, switch the optimization back to the actual outcome. Feeding Meta real conversions beats any proxy.

This pairs with a broader structure for expensive offers. The math in High Ticket Facebook Ads explains why only a small fraction of leads ever buy, and Value-Based Optimization in Meta Ads covers sending value signals when you have them.

If your sale happens off-site, Meta Offline Conversions shows how to feed closed deals back in. And for the front-end choice between a lead magnet and a call, see lead magnet vs book a call.

Read-rate heuristics: 10% versus 50–60%

Once your events flow, you can compute a read rate per ad set: the share of landing-page viewers who reach a given depth. Choose one consistent definition, for instance the share of landing page views that fire Read50, and stick to it.

A common practitioner rule of thumb holds that a read rate near 10 percent signals the wrong audience, while 50 to 60 percent signals a good one. Treat those as rough bands, not laws.

They depend on article length, device mix, traffic source, and how you define a view. A 600-word piece will beat a 4,000-word piece on completion without being a better audience.

How I use the bands:

Read rate (Read50 / landing views)ReadingAction
Under 15%Likely a mismatch between the ad promise and the article, or the wrong audienceCheck the headline-to-hook match, then audience
15% to 40%Normal for cold traffic on long contentCompare ad sets against each other, not against a benchmark
40% to 60%+Strong early signal of intentBuild depth audiences, consider scaling

The comparison between ad sets matters more than any absolute number. Same article, different audiences or different image ads, and one has a read rate twice the other's. That gap is a result you can act on in days, weeks before conversions accumulate.

Two cautions. First, a low read rate can be a tracking bug. Check for the event firing on slow connections and in in-app browsers, because most Meta clicks open there. Second, the heuristic is a screening tool.

A 55 percent read rate with zero calls still means something is broken downstream. The wider signal hierarchy is laid out in Data-Driven Ad Decision Making, and the conversion rate calculator and CPA calculator help you connect read rate to outcomes.

If you want a drop-off view, compare Read50, Read75 and Read90 counts per article. A cliff between 50 and 75 shows where readers abandon. That feeds article edits directly, which is the cheap iteration loop most teams skip. On the ad side, creative testing on a small budget uses the same idea: read a cheap signal before the expensive one.

Scroll events look harmless, but they are behavioral data tied to a browser, and in Europe that has legal weight. The ePrivacy Directive (2002/58/EC) requires consent before storing or reading information on a user's device, with narrow exceptions.

The GDPR governs the personal data you send onward, and it sets the bar for what counts as valid consent. See the GDPR glossary entry for the short version.

Meta's Business Tools Terms also put responsibility on you. As the advertiser, you are the one who must obtain any consent that applicable law requires before the tools collect data. Read the current wording yourself rather than trusting a summary, because it changes.

Practical rules:

  • Fire fbq and the server forward only after consent for advertising is recorded.
  • Do not send scroll events for visitors who declined, even through the server path. A server event does not bypass the consent decision.
  • Keep event names and parameters neutral. Meta's custom conversion documentation notes that conversions suggesting certain health or financial information may be flagged and blocked from running. An event called ReadDiabetesSection is a bad idea.
  • Expect fewer matched events in consent-strict regions, and lean on first-party data where you have it.

This is a note, not legal advice. Have your own counsel confirm the setup for your markets.

Context on why browser tracking got fragile in the first place is in iOS 14 ATT: Five Years On, and the aggregated event measurement entry covers Meta's response for iOS traffic. Scroll depth retargeting Facebook audiences inherit every consent rule that applies to the Pixel itself.

Which long-form ads deserve instrumenting

Before you build your own article and tracking, learn from the ads already running to long-form pages. Advertorial-style ads that stay in market for months are, almost by definition, being paid for by somebody who sees a return.

An ad that ran for ten days tells you little. One that has run for six months tells you the funnel behind it works.

Pull the longest-running ads from competitors in your niche using unified ad search, then check how long each has been live with ad timeline analysis. Open the landing pages those ads point to.

Note the length, the hooks, where the offer first appears, and how many times it is mentioned. Those are the articles worth modelling, and the ones where your scroll milestones should sit where the offer shows up.

That workflow slots into the daily routine described in competitor ad research, and the broader teardown approach is shown in Native Ads Funnel: Direct Response Teardown. For a wider view of the image formats that feed these articles, the native image ads guide covers the creative side.

Choose milestone placement against the article's own structure, not a default 25/50/75/100 split. The ad that sends the click is often a news-style static, and Meta static ads covers those formats.

Frequently Asked Questions

How does scroll depth retargeting facebook work without a purchase event?

You fire a custom event at fixed scroll milestones, such as 50, 75 and 90 percent, and Meta stores them against the visitor. You then build custom audiences from those events in Ads Manager. Purchases are not needed, because the audience is defined by reading behavior on your own page.

Can I optimize a campaign directly on a scroll depth event?

Not directly, according to Meta's documentation. Custom events can define audiences, and campaign optimization runs through custom conversions built in Events Manager. Check which rule types your account supports, and test with a small budget before moving real spend.

How long can I keep deep readers in an audience?

Meta's website custom audience documentation lists a retention window of 1 to 180 days. Choose the window by sales cycle: one to two weeks for low-ticket products, 60 to 90 days for expensive, slow decisions.

Do I need the Conversions API for scroll events?

You can start with the Pixel alone, but a server path adds resilience against blocked or lost browser events. Send the same event name and event ID from both paths so Meta deduplicates them. Gate both behind consent.

What is a good read rate for an advertorial?

As a rule of thumb, around 10 percent suggests the wrong audience, and 50 to 60 percent suggests a good one. Article length and traffic source shift those bands, so compare ad sets against each other first.

Reading is the one conversion-adjacent behavior that is plentiful when sales are scarce. Run scroll depth retargeting facebook as a bridge, and let real purchases take over the day they are numerous enough to carry the account.

Which of your articles would you measure first? Starter at €29 a month covers focused manual research, and Pro at €179 gives power users and small teams the volume. Compare them on pricing.

Related Articles

Scroll Depth Retargeting Facebook: Optimize for Deep Readers