You're mid-campaign. A shopper taps the paid social ad, watches half the video, reads a page about the problem, and leaves. Age and city don't tell you whether that person is ready to buy. Those actions do.
Behavioral targeting examples earn a place in a pre-sell plan for that reason. You're not buying a persona slide. You're buying a trigger, then sending that person to a page written for how far they already got. A cold scroller needs a problem-led listicle. A shopper who finished most of the video needs a shorter advertorial that assumes the pain is already felt.
This is a working list for paid social, email, onsite, and programmatic. Eight triggers. Each one is a test you can ship this week, with an audience rule, a page frame, one duplicate, one metric, and a consent check before spend.
Behavior-based buying is not a side tactic. In 2025, behavioral targeting represented $1.55 billion in revenue and 28.5% market penetration in the global selective advertising market, according to DataIntelo. That money is wasted when every segment hits the same product page.
Pick the trigger with the clearest intent in your catalog. Build the page to match it. Then hold out a slice of the audience so you can see whether the page, not the bias in the audience, did the work.
Table of Contents
- 1. Website Visitor Retargeting
- 2. Dynamic Product Ads Based on Browsing Behavior
- 3. Cart Abandonment Sequences
- 4. Purchase-History Recommendations
- 5. Engagement-Based Custom Audiences on Social Platforms
- 6. Lookalike Audiences from Purchasers
- 7. Exit-Intent and On-Site Behavioral Triggers
- 8. Interest and Topic-Based Behavioral Segments
- How the eight behavioral targeting examples compare for DTC testing
- Turn Each Trigger Into a Tested Page This Week
1. Website Visitor Retargeting
A pixel or first-party cookie records which pages someone viewed, then an ad finds that browser later. Cookies became the default identifier that can connect visits, and often other sites. A 2017 literature review in the Journal of Advertising reported that the 100 most popular websites collected more than 6,000 cookies in total, and that 83% were third-party cookies. An "all visitors" pool built on that kind of tracking mixes buyers, bounces, and real readers.
Split by depth before you write the next ad.
- Advertorial readers: They know the problem, not the SKU. Keep a proof-led advertorial.
- Product viewers: They saw the offer and stalled. Use a testimonial ad plus a short comparison page.
- Recent buyers: Suppress them, and cap frequency so the same face doesn't follow them all week.
Say a supplement brand's cold ad opens on afternoon energy. Readers who leave should see that claim again, then hit a duplicate of the same advertorial. In Landra, duplicate the page, change only the opening line so it acknowledges the return, and point the ad set at the new URL. Hold out part of the audience. Clicks will lie to you here.
Wojo Media's website retargeting guide walks through the pixel and the audience window.
Practical rule: The second click should continue the first story. A product grid makes them start over.
Honor browser opt-outs before launch, and don't infer a health condition from an energy or skin article.
2. Dynamic Product Ads Based on Browsing Behavior
A fashion shopper views a black jacket on Tuesday. On Thursday the ad is that jacket, that price, that photo. The catalog did the creative. That's browsing behavior, and it only works if the feed is clean.
The lift is smaller than most decks claim. A House Energy and Commerce filing reviewed millions of ad transactions and found publisher revenue only about 4% higher when a user's cookie was available, after the value of the impression itself was accounted for. The same filing notes that temporal signals, such as time on the advertiser's site and time since the last visit, increased effectiveness by more than 20%. Recency is the lever. One product view is not.
Clean the Feed Before You Scale
Bad titles and stale prices make the ad look broken, and people learn to ignore it. Fix the feed, then test one narrative hook against the plain product template for cart viewers only. Cold traffic should not see a raw catalog unit. Send those clicks to an advertorial. Save the dynamic ad for people who already touched a SKU.
You can build Meta ad creatives with ProdSnap when catalog shots are too flat for a warm placement. I still wouldn't send that click to the product page if the first ad was a story. Duplicate a short "you left this" presell, keep the product module, and change only the objection on the first screen.
Watch the template once before you scale it.
Catalog matching still needs a valid basis for the pixel. Suppress anyone who already bought that SKU in the current use cycle.
3. Cart Abandonment Sequences
A cart add with no order is the cleanest intent signal in the account. Treat it as a three-step test, not a discount machine.
Hour one is a plain reminder with the item image and a return link. No code. Day one addresses the objection that kills the sale, whether that's shipping, ingredients, or fit, and it can point to a short proof listicle. Day three is the only step that may carry an incentive, and only if recovered revenue on the first two steps is thin. The sequence stops the moment they buy.
A discount on the first email trains the list to abandon.
Landra's cart recovery guide lays out the message logic. The page test is separate. Build one objection-led listicle for abandoners, duplicate it, and change only the headline from a generic "still thinking" line to the specific doubt, such as texture or sizing. If the ad that started the session was educational, send the first email to that listicle, not straight back to the cart. Send the incentive step to checkout.
Measure recovered revenue per step, plus unsubscribe rate. If the last step prints revenue and also spikes unsubscribes, the offer is too loud. One retargeting ad can repeat the cart items with a review line. Don't put a second discount in that ad if the email already has one. You'll pay twice to teach a lower price.
Cart email and SMS go only to people who opted in for that channel. A pixel match is not permission to text.
4. Purchase-History Recommendations
Past orders beat interest labels for the next offer. You already know the SKU, the date, and whether they paid full price.
Time the Offer to the Cycle
Map the use window for a consumable. A 30-day pouch should not get a reorder note on day three, and it should not wait until they've already switched brands. A cleanser buyer can get a serum advertorial framed as a second step, not a random cross-sell. Build bundles from pairs that show up in the same orders.
The sharper cut ranks buyers instead of blasting all of them. Rokt's onX case study describes a model trained on purchase, trial, and upgrade decisions. Only customers in the 95th percentile and above were targeted, a control group stayed out, and membership-tier upgrade conversions rose 44% versus control. The result reached significance within one month and did not increase cancellations or uninstalls. I treat that shape as the standard even without a model. Score repeat sessions, time since last order, and email clicks. High scores get a tight comparison page. Lower scores get a longer use-case advertorial.
Generate the use-case page from the product URL, duplicate it, and open by naming the product they already own. Suppress anyone still inside the refund window. Read refund rate next to conversion. A reorder push that lifts orders and refunds is a loss.
5. Engagement-Based Custom Audiences on Social Platforms
Match View Depth to Page Length
On Meta and TikTok, a 25% video view and a 75% video view are different buyers. The first group sampled the hook. The second sat through the argument. One URL for both is how a warm audience gets wasted.
Build two custom audiences off the same ad.
- 75% viewers: Short advertorial. Lead with proof and the offer. They've heard the problem.
- 25% viewers: Problem-led listicle. Restate the pain on the first screen. Don't assume they finished the video.
- Lead-form openers who didn't submit: Give them a third page only if volume is real. Otherwise fold them into the shallow-view group so spend doesn't fragment.
A fitness brand can retarget people who watched most of a Reel with a program advertorial, and send hook-only viewers to a listicle that restates the struggle. Use engagers as warm retargeting before you seed anything from them. Engagers are cheaper to reach and worse as a seed than buyers. I'll take purchasers as the seed whenever that file is large enough.
Refresh the definition when the winning creative changes. Last quarter's 75% audience holds a different belief. Duplicate the page per tier and change only the opening hook. Same body, same offer, so the test isolates the match between view depth and the first screen. Stay inside the platform audience. Don't upload a scraped list of people who watched.
6. Lookalike Audiences from Purchasers
A lookalike extends a seed. It doesn't invent intent. The platform studies purchasers and finds new people with similar behavior. If the seed is every buyer, including one-time discount hunters, the model learns the wrong customer.
Use high-value buyers when the list is big enough to model. Test a tight lookalike from that seed against a lookalike from all email subscribers. Keep each in its own ad set with a matched budget. Pair each with an angle-specific advertorial so you can see which story new people finish.
Identity is thinner than the old playbooks assume. Experian's guidance on cookie deprecation describes a shift toward first-party data, modeling, and exclusions, while Chrome has moved toward user controls rather than a universal third-party cookie phaseout. A page view still doesn't prove purchase intent. Build the lookalike from consented purchase events, set a retention window, and exclude customers from the acquisition campaign.
If you can't explain the seed in one sentence, don't spend against the lookalike.
Generate one advertorial per angle, duplicate it, and change only the hook that matches the seed, such as what repeat buyers reorder versus what first-timers compare. Judge cost per acquisition against a product-page control, not CPM. A cheap lookalike that doesn't buy is just a broad audience with extra steps.
7. Exit-Intent and On-Site Behavioral Triggers
The visitor is still on your URL. You don't need another ad platform to react. Scroll depth, time on page, a second visit, or a move to leave can change what they see. Three popups in one session will wreck that same page.
Repeat visits are the stronger onsite signal. A published test changed the experience only for companies that returned within 60 days of a first visit. WebEo's Asset-Map write-up says the test reached significance after two weeks, with conversion uplift of 221% and more than three times the previous number of demo requests. Those visitors already intend more. They need a different page, not a louder popup.
One Trigger Per Session
- Past the midpoint of an advertorial: One slide-in with a checklist.
- Exit on a product page: A line on shipping or the guarantee, not a new discount if email already has one.
- A second visit in the same week: Swap the hero to proof. That returning page is the one worth duplicating.
Test mobile on its own. Desktop exit intent is a mouse movement. On a phone it's a back tap or a stall, and the same script misfires. Read scroll depth to see which section people finish, and analyze visitor behavior on landing pages before you add another module.
Duplicate the presell, shorten the hero for the exit audience, and leave the rest alone. If the trigger slows the page, kill it. A slow page loses more buyers than a missed popup saves.
8. Interest and Topic-Based Behavioral Segments
Interest targeting is a hypothesis about topics, not a receipt. The platform infers dog training, sensitive skin, or anti-aging from what people engage with. Two segments can share a product and still need opposite openings.
A pet brand shouldn't run one "best dog food" page at both training and nutrition. Training gets a listicle that starts with behavior problems. Nutrition gets an advertorial that starts with ingredients. Split sensitive skin and anti-aging into separate ad sets, with the same budget and bid type and a different first screen. Stack them and the winning angle disappears into the average.
Keep the body of the page stable so the test is the hook. That's how you find a message before you scale spend. Interest pools also overlap people you can already reach. Exclude purchasers and your video retargeting audiences so you don't pay interest prices for warmer traffic.
To map personas to pre-sell funnels, write the interest as a belief, not a demographic. "Thinks their skin reacts to everything" is a page. "Women 25 to 34 interested in beauty" is not.
Generate a listicle for one interest and an advertorial for the other from the same product URL. Duplicate each and change only the opening hook. After learning, read cost per acquired customer by ad set. Kill the segment that spends without finishes on the first screen, even if click rate looks fine.
Interest is a door, not proof they want your SKU.
A health-adjacent interest is not a license to state the inference in the ad. Talk about the problem in general language. Don't announce that you know why they were reading.
How the eight behavioral targeting examples compare for DTC testing
| Item | Trigger Behavior | Best Audience Temperature | Primary Channel | Primary KPI |
|---|---|---|---|---|
| Website Visitor Retargeting | Ads follow a site visit or specific page activity | Warm | Paid ads across social and search | Conversions from returning visitors |
| Dynamic Product Ads Based on Browsing Behavior | Ads show products viewed or added to cart | Warm | Paid social and shopping ads | Product purchases |
| Cart Abandonment Sequences | Messages follow a cart creation without purchase | Hot | Email, SMS, and retargeting ads | Recovered revenue |
| Purchase-History Recommendations | Offers reflect past purchases and replenishment patterns | Existing customers | Email and on-site recommendations | Repeat purchases |
| Engagement-Based Custom Audiences on Social Platforms | Audience forms from video views, post interactions, or lead-form activity | Warm | Social platforms | Conversions from engaged users |
| Lookalike Audiences from Purchasers | Platform finds new users similar to purchaser seed audiences | Cold | Paid social and digital ads | New-customer conversions |
| Exit-Intent and On-Site Behavioral Triggers | On-site messages respond to exit intent, time, or scroll activity | Warm | Website pop-ups, slide-ins, and banners | Lead capture or recovered visits |
| Interest and Topic-Based Behavioral Segments | Audience targeting reflects inferred interests or topic activity | Cold | Paid social and search ads | Prospecting conversions |
Turn Each Trigger Into a Tested Page This Week
Pick two triggers with the strongest intent for your catalog. For most stores that's a cart abandoner and someone who watched at least three quarters of the video. Give each its own pre-sell page. The product page is the control, not the destination for both.
- Warmer trigger: An advertorial with the offer visible early.
- Colder trigger: A problem-led listicle that teaches before it sells.
- Control: The same ads pointed at the product page, plus a holdout.
Generate those two pages in Landra from the product URL. Duplicate each and change only the opening hook. Keep the offer, the proof, and the close identical. If the discount and the headline move in the same week, you can't tell which one changed cost per acquisition.
Log the trigger, the URL, the offer, and the consent basis on every launch. Email and SMS need an opt-in for that channel. Onsite events need a plain notice and a working opt-out. A retargeting pixel is not a yes. Cap frequency, and suppress recent buyers before the campaign spends.
Comfort is part of the result. A 2025 YouGov report on personalized advertising found that 56% of Americans are uncomfortable with companies using online behavior to personalize advertising, while IAB research cited there says more than 70% of U.S. consumers are willing to share data to support ad-funded internet access. Value and control change the answer. Location usually doesn't. In 2024, 55% of consumers in France, the United Kingdom, and the United States described ads based on location data as creepy, according to Statista's summary of an August 2024 survey. Leave location out unless the product is local.
Run the pair long enough to clear learning, with a holdout in place. Read incremental purchases, refunds, and unsubscribes together. A lower acquisition cost that raises refunds is not a scale candidate. When one hook beats the product page, duplicate that page for the next trigger. Keep the losing hooks in the log so next month's test doesn't rebuild them.
Landra turns a product URL into an editable advertorial or listicle you can publish to Shopify, a hosted URL, or HTML, so each behavioral trigger gets its own pre-sell page instead of the same product grid. Duplicate the page, change the opening hook, and put the warmer audience on the shorter story. Start with the trigger you're already spending on this week at Landra.
