Off-Site Extension and DSP Integration

The fastest-growing piece of retail media. How retailer first-party shopper data activates against open-web inventory.

Why off-site extension is the fastest-growing piece

On-site retail media (sponsored search, sponsored display on the retailer's property) is mature and well-understood. The fastest-growing piece is off-site — using the retailer's first-party shopper data to target advertising on the open web, social platforms, CTV, and audio. Off-site retail media revenue is projected to triple between 2023 and 2027, outpacing on-site growth meaningfully.

The reason is structural. As third-party cookies disappear and walled-garden platforms tighten their data, retailer first-party data becomes one of the most valuable targeting signals in advertising. Off-site extension lets advertisers attach that signal to inventory they could not previously target with retailer-grade precision.

How off-site extension works

1. Retailer audience build. Inside the retailer's DSP, advertiser builds audiences from shopper behavior — bought brand X in 90 days, viewed category Y, lapsed buyer of SKU Z.
2. Bidstream activation. The retailer's DSP exposes those audiences as biddable segments across open-web inventory through its programmatic supply — either direct integrations or via a primary SSP partner.
3. Closed-loop measurement. When the user later returns to the retailer (in-store loyalty card, on-site purchase, app purchase), the retailer matches the conversion back to the off-site impression for measurement.

The closed-loop piece is what makes off-site retail media structurally different from open-web programmatic. A normal DSP buy on, say, The Trade Desk targeting a CPG audience gives you impressions on inventory but no first-party tie-back to actual purchase. The retailer DSP gives you both.

The major retailer DSPs

DSP	What it does best	Notes
Amazon DSP	The most mature retailer DSP. Display, video, audio, OTT. Strong reach including Twitch and IMDb inventory.	Self-service requires minimum spend commit ($35k+ for managed service, lower for self-service via partners).
Walmart Connect DSP	Powered by The Trade Desk under the hood. Display, video, CTV (Vizio post-acquisition).	Strong for omnichannel campaigns combining in-store and online Walmart shoppers.
Roundel (Target)	Roundel self-service platform plus managed service. Premium-positioned shopper audience.	Strong for brands wanting Target's slightly higher-HHI shopper.
Kroger Precision Marketing (KPM)	84.51°-powered audience science. Strongest in grocery CPG.	Higher CPMs reflect data quality. Loyalty-card-tied audiences are exceptional.
Instacart Ads (Display/CTV)	Cross-retailer grocery shopper data; CTV launched 2023–24.	Useful for CPG advertisers without a primary grocer relationship.
Criteo Commerce Media	Network of 200+ retailers. Off-site shopper graph that aggregates across retailers without exposing identity.	The closest thing to a cross-retailer DSP for off-site.
Albertsons Media Collective	Loyalty-card-tied audiences across Safeway, Vons, Albertsons banners. Display and video.	Strong for west-coast and pacific-northwest grocery CPG.
CVS Media Exchange & Walgreens Advertising Group	Pharmacy retail; OTC health, beauty, wellness audiences.	Required for any major pharmacy CPG brand.

Use cases: when off-site beats alternatives

Off-site beats on-site when…

• You need awareness reach, not just bottom-funnel intent capture. On-site retail media reaches shoppers already on the retailer; off-site reaches them everywhere else they live.
• You need to influence shoppers before they reach the retailer — pre-shop research, category education, brand-building.
• You want video or CTV inventory; on-site video options are growing but still limited.
• You're targeting lapsed shoppers who aren't returning to the retailer naturally.

Off-site beats open-web programmatic when…

• You can target by actual purchase history rather than third-party-data approximations.
• You need closed-loop measurement that attributes off-site impression to retailer sale.
• Privacy regulation or signal loss has degraded your open-web audience targeting.
• You're running co-op campaigns with the retailer where shared measurement matters.

Planning off-site campaigns

Audience construction

• Buyer audiences: Bought brand X in 30/60/90/180/365 days. Lapsed buyers (bought 6–12 months ago, none since). Heavy buyers (top quintile of category spend).
• Category audiences: Bought anything in category in window. Useful for cross-sell or new-product launches.
• Browser audiences: Viewed PDP but didn't buy. Added to cart but didn't purchase. Search queries indicating intent.
• Lifestyle audiences: Loyalty card data overlays — pet owners, parents of toddlers, premium-tier loyalty members.
• Competitor audiences: Bought competing brand. The conquest play.
• Lookalikes / predictive audiences: Retailer-built lookalikes from your seed list.

Creative strategy

Off-site creative differs from on-site creative because the user isn't already on a shopping property. The creative must do the work of bringing them back — or, if the goal is awareness, build memory for when they next shop.

• Lead with the product, not the retailer. The retailer is the audience source, not the campaign theme.
• Include retailer logo and CTA if the goal is store visit. "Now at Target," "Shop at Kroger," etc. drives lift in store visits.
• Match creative to audience stage. Lapsed buyer creative differs from new-to-category creative. Both differ from heavy-buyer cross-sell creative.
• Frequency cap aggressively. Retailer audiences are smaller than open-web audiences. Without caps, you'll show the same person 30+ ads/month.

Measurement: closed-loop attribution

The premier feature of off-site retail media is closed-loop measurement against retailer sales. Walk through what that looks like in practice.

1. User sees ad on open-web inventory (e.g., NYTimes.com display).
2. Retailer DSP logs impression with hashed user identifier.
3. User shops at retailer (online with login, or in-store with loyalty card).
4. Retailer matches purchase event back to impression log.
5. DSP reports attributed sales lift, new-to-brand %, category share.

Watch out for these measurement issues:

• Attribution windows. 7-day click + 14-day view is typical. Adjust based on purchase cycle — daily-purchase categories need shorter windows; quarterly-purchase categories need longer.
• Causal vs correlational lift. Closed-loop attribution shows correlation. Incrementality testing shows causation. Both required for full picture.
• Cross-retailer cannibalization. Off-site campaign on Retailer A might shift sales from Retailer B without growing total category — bad for the brand even if Retailer A reports lift.
• In-store match rates. In-store loyalty match rates run 30–60% in most categories. Means measurement under-counts true off-site impact for in-store-heavy categories.

Inventory types

Inventory type	Strengths	Use case
Open-web display	Reach, low CPMs, frequency build	Always-on category presence; retargeting cart abandoners
Open-web video (pre/mid/post-roll)	Storytelling, brand-building	New product launches, category education
CTV (connected TV)	Premium attention, household-level addressability	Brand-building with retailer audience overlay; the fastest-growing off-site piece
Audio (podcast, streaming radio)	Sustained attention, intimate format	Brand awareness for categories with strong audio listenership (auto, finance, health)
Native	Looks like content; lower banner blindness	Mid-funnel education; sponsored content
Social platforms (Meta, TikTok)	Retailer-audience activation on social inventory	Available through some retailer DSPs as social inventory partnerships expand
DOOH (digital out of home)	Geo-located inventory near retailer locations	Drive store visits for offline-heavy retailers

Advanced playbook

• Sequenced cross-retailer activation. Use Criteo's cross-retailer graph or Instacart's aggregated network for category brands that aren't exclusive to one retailer. Avoid wasted impressions on shoppers loyal to a single retailer not in your distribution.
• Lapsed buyer reactivation. The highest-ROAS use case in off-site. Buyers who lapsed 6–18 months ago convert 2–3× better than cold prospects when shown reactivation creative.
• Trade event amplification. When a retailer features your brand in flyer, app, or end-cap, mirror that with off-site retailer-audience media to drive incremental visits during the promo window.
• Audience suppression. Always suppress recent purchasers from off-site spend. Always suppress employees and known fraud signatures. Always suppress audiences you're reaching elsewhere if frequency control matters.
• CTV-on-Amazon (Streaming TV / Freevee). Combines Amazon DSP audience precision with premium CTV inventory. The strongest off-site format Amazon offers for awareness-stage brands.
• Loyalty tier targeting. Kroger and Albertsons expose loyalty tier as audience attribute. Premium tier targeting (top 20% spend) is gold for premium brands.
• Recipe and meal-occasion targeting. Grocery retail media DSPs let you target by recipe interaction, meal-occasion search, basket pattern (e.g., taco-night basket). Brand creative aligned to occasion lifts CTR meaningfully.
• Co-op funding negotiation. Retailers often co-fund off-site campaigns at 50–100% when tied to JBP commitments or new product launches. Always ask.
• Frequency-capped sequencing. 4–6 impressions in week 1 of awareness flight, taper to 2–3/week in maintenance. Sustained at 0 outside flight windows.
• Open-web display + retailer audience as cookieless replacement. As third-party cookies fail, retailer audiences become one of the strongest substitutes for legacy programmatic targeting. Build the bench now.

Common mistakes

• Treating off-site retailer DSP as just another programmatic buy — you ignore the audience data, which is the whole value.
• Running off-site without closed-loop measurement turned on — you optimize against open-web KPIs that don't match the brand's actual goal.
• Choosing one retailer DSP as a monopoly because their sales rep is convincing — ignore alternative networks where your audience overlap differs.
• No frequency caps — small retailer audiences burn through impressions to the same households many times over.
• Creative built for on-site reused as-is for off-site — doesn't translate.
• Off-site measurement compared apples-to-apples with on-site ROAS — off-site is upper-funnel and will look worse on last-click; that's fine, plan for it.
• Confusing closed-loop attribution with incrementality; running no holdouts.
• Forgetting to suppress current customers from prospecting off-site flights — you pay to talk to people already buying.
• No alignment with retailer co-op or JBP commitments — missing free or shared-funded inventory.
• Underinvesting in CTV through retailer DSPs because the team's open-web CTV team thinks they have it covered — closed-loop CTV is structurally better.

Operating checklist

• Active off-site programs on each of your top 3 retailers (or Criteo/Instacart equivalents if not direct)
• Audience strategy documented per retailer: buyer, lapsed, browser, lifestyle, competitor, lookalike
• Closed-loop measurement turned on; attribution windows aligned with category purchase cycle
• Creative variants for each audience stage (cold/category/lapsed/heavy-buyer)
• Frequency caps set; documented per audience size
• Suppression lists active: current customers, recent purchasers, employees
• Inventory mix beyond display: video, CTV, audio where supported
• Quarterly incrementality testing on top retailer DSP
• Co-op funding inquiries during JBP cycles
• Open-web cookieless transition plan that incorporates retailer DSPs as primary substitute

---

Part of the Retail Media series. Continue to the next module or take the series exam.