How to Build a 3-Stage Retargeting Video Sequence

Stage 1 is video viewers at 50% or more of a prospecting video. Stage 2 is site visitors who did not purchase. Stage 3 is add-to-cart or initiate-checkout. Exclude purchasers from all three. If you cannot fill Stage 3 yet, still build Stages 1 and 2 — do not invent a cart audience from page views. Draw the three audiences on one page with arrows for who is excluded from whom before you touch Ads Manager.

Heat is about what they already know. A 50%+ video viewer has seen the product and the hook; they do not need another cold problem-agitation open. A site visitor has seen the offer on the page and bounced on a specific objection — price, shipping, proof, or fit. A cart abandoner has chosen the variant and stopped at friction. Serving all three the same 'wait till you see this' UGC ad wastes the information you already paid to collect. If your pixel is thin, start with video viewers and site visitors only; a cart audience of a few dozen people a week cannot exit learning and will only inflate frequency.

Build the audiences in Events Manager and Ads Manager as separate custom audiences, not as one stacked 'engaged' pile. Use the prospecting campaign's video views for Stage 1 so the sequence is actually fed by the ads you are running, not by old organic views of a different product. If you sell several SKUs, split Stage 2 and Stage 3 by product or collection once volume allows; until then, one product-agnostic sequence with dynamic catalog fallback is cleaner than six empty ad sets. Empty custom audiences are a structure problem, not a reason to widen Stage 3 to 180-day page viewers.

Use a short window for Stage 1 (about 7 days of video viewers), 7–14 days for site visitors, and a tighter 3–7 days for cart abandoners. Exclude each hotter stage from the colder ones so a cart abandoner is not also paying for the Stage 1 testimonial. Refresh the purchaser exclusion on a schedule. If your product has a longer consideration cycle, extend Stage 2 first; leave Stage 3 short so it stays close to intent.

Windows that are too long are how retargeting earns a reputation for annoying people. A visitor from three weeks ago has usually bought elsewhere or lost the itch, and paying to re-reach them drags efficiency down. Seven to fourteen days is the default for most ecommerce products; high-consideration items can stretch Stage 2, but Stage 1 should stay short so it stays adjacent to the prospecting video they actually watched. Cart windows should be short because intent decays fast and because a long cart audience overlaps last week's purchasers if your exclusion list is stale.

Nesting is the structure: Stage 1 excludes purchasers, site visitors, and add-to-cart. Stage 2 excludes purchasers and add-to-cart. Stage 3 excludes purchasers only. That waterfall is what makes it a sequence rather than three overlapping blasts. If you skip nesting, CBO or even ABO will over-serve the cheapest, warmest people and you will think Stage 1 'doesn't work' when it never got a clean look at video viewers who had not already come to the site. Re-download the purchaser list on a weekly cadence if your store and pixel disagree about who already bought.

Stage 1 answers 'does this actually work?' with proof. Stage 2 answers the objection that stops a browser — size, ingredients, shipping, returns, or social proof. Stage 3 removes friction: what's in the box, restock, easy returns, or a time-bound offer. Do not reuse the cold hook on any of the three. Three scripts, three jobs; if two scripts could swap stages without anyone noticing, they are not specific enough yet.

Match warmth, not just format. Video viewers who never clicked are missing information, so a testimonial-style UGC video or a tighter demo is the right next asset. Site visitors who browsed and left need reassurance or specificity the landing page did not land — a 'still deciding?' angle, a fit or texture close-up, a review paraphrase. Cart abandoners respond to friction removal, not to another origin story. An offer-led static can outperform another cold-style video at this stage because the shopper already knows the product; they need a reason to finish.

Write the three scripts before you produce anything, and lock claims to the live product page. A Stage 3 urgency line that names a discount the site is not running will convert and then refund, or it will fail policy. If you do not have a real offer, use operational friction removal instead of inventing a sale: shipping timeline, returns, how to pick a variant. Honesty at Stage 3 is cheaper than a coupon you cannot honour. Warm traffic can fact-check you in real time, which is why stale offers hurt more here than in cold.

Generate a proof-led UGC cut for Stage 1, an objection-handling cut for Stage 2, and a short friction-removal cut for Stage 3, plus one offer-led static for the warmest traffic. Keep the product, avatar family and captions consistent so the sequence feels like a continuation, not three unrelated brands. If you only have time for two cuts, skip a new cold hook and make Stage 1 proof plus Stage 3 friction removal.

You do not need three production days. From one product brief, change the beat mix: Stage 1 leans on demo and a review line, Stage 2 leans on the specific objection, Stage 3 is often 15 seconds and closer to a packshot-plus-CTA. Export 9:16 and 1:1 for each. Klip Kanvas can generate those three script shapes from the same product URL so claims and price stay aligned with the listing, which matters more in retargeting than in cold — these people can check the page in another tab while the ad plays.

Hold creative constant for at least a week per stage once live; this is not the campaign where you test five hooks at once. The sequence is already a test of message-by-heat. If you also rotate hooks weekly inside each stage, you will not know whether Stage 2 is weak or whether you keep resetting it. Refresh when frequency on that ad set climbs past roughly 2.5–3, using a new proof angle, not a new product. A sequence that changes product mid-stream trains people to ignore the next ad, not to buy.

Put the three stages in a dedicated retargeting campaign, not inside the prospecting test or the CBO scaler. Set a frequency cap or, if the UI is limited, control frequency with window length, audience nesting, and a modest budget. Watch frequency weekly; it is the fatigue signal that shows up before CPA does. Prospecting CBO must not be allowed to 'find' this list; if it can, your exclusions are wrong.

Retargeting audiences are small. The same 2,000 site visitors will absorb a prospecting-sized budget in a few days and frequency will climb into the range that burns the list. That is why retargeting should not sit under a prospecting CBO campaign — the algorithm will over-serve cheap warm traffic until everyone hates the ad. A separate campaign with three ad sets (ABO is fine here, because you want each stage to actually spend) keeps the cap honest. Frequency is the retargeting dashboard's first column, not a vanity metric you check when CPA is already ugly.

There is no single perfect cap for every brand, but treating frequency above about 2.5–3 as a warning is consistent with how you already read prospecting fatigue. If Stage 3 is at a high frequency and still converting at target CPA, leave it; if CPA is rising with frequency, shorten the window or rotate the friction-removal creative. Do not 'fix' high frequency by adding a cold hook — that is how sequences turn back into one noisy ad set. If you cannot set a hard cap in the UI, treat window plus budget as the cap and write the numbers down.

Budget the retargeting campaign as a small, deliberate share of total media — about 10–15% is plenty, because the audience is only as big as the traffic your cold campaign generates. Size Stage 3 smallest, Stage 1 largest among the three, unless cart volume is genuinely high. If cold spend doubles, the retargeting share can rise in absolute terms without rising as a percentage.

Retargeting cannot scale past the prospecting that fills it. If you raise the retargeting budget because its CPA looks pretty, you will either hit frequency walls or start leaking into colder lookalikes you did not intend. Keep 10–15% as the default ceiling and only raise it when the audiences are growing because cold traffic grew. Judge the retargeting campaign on its own CPA and on blended return, not on whether it 'beats' prospecting — of course warm traffic converts cheaper; that is not a reason to starve the top of the funnel.

Use ABO so each stage gets a floor of spend, then check weekly whether Stage 1 is wasting money on people who never click through. If Stage 1 CPA is worse than cold, the proof video is not doing its job — fix the message, do not just cut the stage without a readout. If Stage 3 cannot spend its tiny budget, the cart audience is too small; merge a 7-day initiate-checkout with add-to-cart rather than inventing fake scale. Pretty warm CPA is not a licence to starve prospecting; without cold traffic the sequence empties in a week.

Launch all three stages together once prospecting is already producing views and visits. Do not edit budgets or creative for 72 hours. Then read CPA, frequency, and hook rate per stage — not as one retargeting blended number that hides a broken Stage 1. Launching retargeting before prospecting has views is how Stage 1 spends on leftover organic viewers from another product.

A blended retargeting CPA is how weak Stage 1 creative survives. If Stage 3 is doing all the conversions at a high frequency, the sequence is not working; you have a cart closer and an expensive screensaver on video viewers. Keep the same funnel metrics you use in cold: hook rate still tells you whether the first three seconds of a proof video are even being watched, even though the bar can sit a little different on warm traffic. Kill a stage video that spends past 1.5× that stage's target with no conversions, the same written rule as prospecting.

Reconcile retargeting purchases against the store. Warm audiences are where duplicate pixels and unexcluded purchasers show up as miracle ROAS. If platform-reported conversions on Stage 3 dwarf Shopify orders, fix tracking and the exclusion list before you scale that ad set. Retargeting should make blended return look better because it recovers people you already paid to reach — not because it is double-counting them. A tracking gap on warm traffic will make you scale the wrong stage and train the pixel on duplicates.