Video retargeting: audiences from video viewers and VSL watch depth, the right ad for each segment, and how to tell if it works

Video retargeting means showing ads to people who watched part of a video: one of your Facebook or Instagram video ads, or the VSL on your own page. On Meta you build these audiences from video engagement, from 3-second views up to 95% of the video, and from pixel events on your site. Match each ad to how far the person got, exclude buyers, and measure what the ads add.

What is video retargeting?

The term covers two jobs. The first is retargeting people who watched your video ads on Facebook or Instagram. Meta knows how much of each ad they watched, and you can build audiences from it. The second is retargeting people who watched your VSL on your own page. There, Meta only knows what your page reports through the pixel, so watch depth takes extra setup.

Both work for the same reason. How far someone watched is a cheap, honest signal of interest, and where they stopped hints at what stopped them. A viewer who left during the hook wasn't interested yet; a viewer who left at the price was interested and hesitated. Showing both the same ad throws that information away. Your VSL's retention curve shows where those stops cluster, so read it before you write a single retargeting ad:

How to build a video view custom audience on Meta

Meta's Marketing API lists the video engagement events an audience can be built from (Video and Carousel Ads):

Three details change how you use them:

• "Completed" means 95% in these audience definitions. A ThruPlay is something else: a metric and an optimization goal for people who play the ad to completion or for at least 15 seconds (ad set reference).
• The percentage audiences include skippers. Someone who dragged the playhead to the end counts as a 95% viewer, so treat deep segments as "probably interested", not "heard the whole pitch".
• You choose the window. Meta's own example audience uses the past 14 days. Choose a window per segment (more on that below), and check the current maximum in Ads Manager rather than trusting a number from a blog.

Build audiences per video and name them so a teammate can read them: "Joint VSL cut, 50%, 14 days" beats "Retargeting 3".

How to retarget VSL viewers by watch depth

On your own page, Meta sees only what the pixel sends. It can build website custom audiences from rules "based on URLs visited or on custom events from Meta Pixel" (Website Custom Audiences), and Meta's conversion tracking documentation says custom events "can be used to define custom audiences for ad optimization." So the audiences you can build depend on the events your page fires: visitors, people who pressed play, people who clicked the call to action, and any custom milestone your page reports. Our VSL tracking guide covers the pixel and server setup behind them.

Here is how one week of VSL traffic might split by watch depth:

One week of VSL traffic, by watch depth

Illustrative numbers

1. Visitors8,000
2. Pressed play65% of the step before5,200
3. Watched 25%50% of the step before2,600
4. Reached the pitch45% of the step before1,170
5. Clicked the CTA30% of the step before351
6. Bought39.9% of the step before140

Two tools carry viewer behavior out of the page. The pixel feeds Meta's audiences with whatever events your page sends, and webhooks feed everything else. In RevvyScales, the Meta, Google and TikTok pixels are set once on a folder or VSL (pixels and the events they send), and webhooks (Growth plan and up) send CTA clicks, the 25%, 50% and 75% watch milestones, pitch reached, video completed and captured leads to Zapier, Make, Klaviyo, ActiveCampaign or your own backend (webhooks and CRM events). That is how "watched half the VSL but didn't buy" becomes an email flow for the viewers who gave you their address.

For supplement pages, keep health conditions out of the URLs and events your audiences are built on. Since a rollout that began September 2, 2025, Meta flags custom conversions that suggest specific health conditions, with examples such as "arthritis" and "diabetes", and stops them from being used to run campaigns (Meta). An audience built on a URL like /arthritis-vsl sends Meta exactly the kind of information that rule is about.

What should you show each segment?

Match the ad to the reason the person stopped. A rough rule: the deeper they watched, the closer the ad should be to the offer.

What to show each watch-depth segment

Starting points to test, not rules

Played, under 25%Left early

The hook missed, or the moment was wrong.

• A different opening angle on the same problem
• A short demonstration of the product in use
• Link to the VSL from the start

25% to the pitchWatched, no offer

Interested, not yet convinced.

• Real customer stories with typical results
• How it works, in under a minute
• Link back to the VSL

Pitch reached or CTA clickSaw the offer

Price, risk or timing held them back.

• The guarantee, explained plainly
• What each bundle includes, per unit
• Answers to the top three objections
• Link to the offer, not the start of the video

Keep the follow-up honest. Testimonials in a retargeting ad must reflect real customers' honest experience, and if they imply typical results, the ad should disclose the results people can generally expect (16 CFR Part 255). Urgency has to be true: a deadline that resets for every visitor is deception, and buyers who feel tricked ask for refunds or dispute the charge. A guarantee you advertise has to work the way the ad says; our post on the money-back guarantee covers the terms. For viewers who left before the pitch, a mini VSL, a short cut that reaches the offer sooner, can carry them to the price without replaying what they have already seen.

Sending deep viewers back to the VSL shouldn't make them sit through the hook again. In RevvyScales, a returning viewer on the same browser resumes from the deepest point they watched (player conversion features).

How to set windows, frequency and exclusions

Exclusions first. Exclusion targeting "prevents a particular audience from seeing your ad," in Meta's words (Website Custom Audiences), and Meta's Advantage+ documentation uses a custom audience exclusion "to exclude people who are your existing customers" (Advantage+ documentation). Exclude purchasers from every retargeting ad set unless the ad sells them something new. Then exclude deeper segments from shallower ones, so a person who reached the price doesn't also get the "you left early" ad.

Windows second. Interest fades, so give deep segments short windows and shallow ones longer. Illustrative starting points, not Meta guidance:

Frequency third. Meta defines frequency as the average number of times each person saw your ad, and notes that it is estimated (Ad Insights reference). Small audiences with big budgets mean high frequency, and you can estimate it before launch. Daily budget ÷ CPM × 1,000 gives daily impressions; daily impressions × 7 ÷ audience size gives impressions per person per week, if delivery spreads across the whole audience. With $40 a day, a $25 CPM and a 2,000-person audience, that is 1,600 impressions a day and 5.6 per person per week. Halve the budget and it drops to 2.8.

A hard cap won't do this job for you on a sales campaign. Meta's ad set reference says frequency caps can only be set on ad sets whose performance goal is reach or ThruPlay (ad set reference). On a purchase-optimized retargeting ad set, your controls are the budget, the audience size, the window and fresh creative.

Do you still need retargeting with Advantage+ sales campaigns?

Advantage+ sales campaigns automate budget, audience and placements, and Meta presents them as a way to win new customers. Its developer documentation shows how to keep existing customers in a separate ad set, with a custom audience exclusion keeping them out of the new-customer one (Advantage+ documentation). Meta's own video walks through the settings you still control:

If you run both, give the retargeting campaign a job the prospecting campaign can't do: messages for people who already know the offer, segment by segment. If your retargeting ads are just your prospecting ads shown again, test whether you need the campaign at all. The Facebook ads for a VSL guide covers the prospecting side.

How to measure whether video retargeting adds sales

Retargeting audiences are people who were already close to buying, so some of them would have bought anyway. Attribution credits the ads for those sales too, which makes retargeting look like your cheapest campaign. The honest question is incremental: how many sales happened because of the ads?

Meta has a tool built for that question. In a lift study, a randomized test group sees your ads, a control group doesn't, and the conversions of the two groups are compared. Meta's developer documentation says "Conversion Lift Measurement is currently limited" and sends advertisers to their Meta representative for access (Lift Study).

Without it, run a holdout yourself. Split your market into two similar groups of regions, run retargeting in one group only, and compare later purchases from each group in your own checkout. A worked example with illustrative numbers, over two weeks:

Say Meta credits the retargeting campaign with 150 purchases: a $20 cost per purchase. But only 60 purchases (240 − 180) happened because of it, so the real cost per extra sale is $50 ($3,000 ÷ 60). If your break-even cost per purchase is $40, the campaign that looked cheapest is losing money.

Holdouts are blunt instruments. The regions must be alike, two weeks of a small account is noisy, and regions far apart can have different seasons. Repeat the test before you trust it, and run it again when the offer or the creative changes.

A video retargeting setup for this week

Start small, and make it measurable from the first day:

A video retargeting setup in five steps

1. 1Build segmentsVideo viewers by depth, page visitors by event, buyers
2. 2ExcludeBuyers everywhere; deeper segments from shallower ones
3. 3Match the messageOne ad per segment, closer to the offer as depth grows
4. 4Set windows and budgetShort windows for deep segments; estimate frequency first
5. 5Test the liftA holdout, judged on your own orders

Then hold yourself to three numbers each week: the cost per kept order from retargeting in your own records, the frequency of each segment, and the gap between the two halves of your holdout. If retargeting can't beat the holdout, move its budget back to prospecting and to new creative; our post on ad creative testing covers that side.