Costco Retail Media, Explained: How Costco Velocity's Ad Stack Actually Works

What Costco Retail Media Actually Is

Costco Retail Media is Costco's own advertising network, now marketed under the name Costco Velocity. It runs on a stack of four layers, and Costco's AVP of retail media, Mark Williamson, laid all four out publicly in early 2026 — an unusual move in an industry that normally treats vendor choices as trade secrets.

The Unified Data Foundation sits underneath everything: Costco's 100%-member-identified transaction data on a private Google Cloud, with MetaRouter and LiveRamp handling identity resolution and Transcend managing privacy compliance. On top of that sit three Activation Channels — onsite ads on Costco.com and the app through Criteo and, since the Moloco Commerce Media integration, a new Reserved Display format; offsite programmatic pushed to The Trade Desk, Google, Yahoo DSP, Epsilon and StackAdapt, plus Google Shopping via a Symbiosys/DoorDash integration; and in-warehouse formats — Costco TV Network screens, a gas-pump pilot, and print in Costco Connection. A Measurement layer built on Habu clean rooms ties exposure back to both online and in-warehouse purchases, reporting new-to-brand and new-to-category rates. Operations runs self-serve and direct-sold side by side.

That's the whole architecture. It's built on vendor partnerships rather than proprietary in-house tech the way Amazon's or Walmart's stacks are — Costco decided speed to market mattered more than owning every layer.

Costco's Ad Stack, Layer by Layer

The table below is the stack itself, not a vendor comparison — it's what each layer does and who Costco has partnered with to build it.

Who It's For and How to Get In

Costco Retail Media sells to suppliers and the agencies that represent them, not to shoppers, and access depends on the format. Sponsored Products-style placements run through Criteo and can be reasonably self-serve. The Multi-Vendor Mailer, the 'Savings Book,' is a bigger commitment — brands already running an MVM offer can layer on 'MVM Amplification,' a coordinated onsite/offsite media package timed to the mailer dates. Anything bigger — Reserved Display, in-warehouse TV, audience-building against your own DSP seat — goes through Costco's Grocery Growth team.

One detail brands miss: the offsite audience activation tool pushes Costco's first-party segments into your DSP seat, not a Costco-owned one. If you don't already have an active seat at The Trade Desk, Yahoo DSP, Epsilon or similar, there's nowhere for that audience to land. Smaller brands without an existing programmatic setup, or without the volume to get a Grocery Growth manager on the phone, are exactly the gap aggregators like RMIQ are built to fill — they sit between smaller advertisers and networks like Costco's, handling onboarding and API integration so a brand too small for direct engagement can still buy in. That's a legitimate lane; it's just a different one from going direct.

Pricing Structure: What Costco Publishes and What It Doesn't

Costco has not published a rate card for Costco Velocity. Nothing in its own materials states a CPM, a percentage of ad spend, or a minimum budget — third-party sites sometimes print numbers, but those aren't sourced to Costco and we won't repeat them here. What Costco has published is the pricing model type, not the rate: onsite Sponsored Products bid CPM/CPC through Criteo's auction, and larger placements — Reserved Display, MVM Amplification, in-warehouse TV — read as negotiated media packages rather than self-serve buys.

If you need a number to plan a budget, the honest path is the one Costco itself offers: email retailmedia@costco.com and get a quote against your category and volume. Any source quoting a fixed Costco CPM without citing Costco directly is guessing, and the guess is usually stale within a quarter — retail media rate cards move often and rarely get corrected in the places that quoted them first.

How to Tell If Costco Retail Media Is Actually Working

Costco's Measurement layer is a real strength worth taking seriously: Habu clean rooms tying ad exposure to actual transactions, split by new-to-brand and new-to-category, is closed-loop measurement done properly, and it's harder to build than most retail media pitches let on. The catch is what it can't tell you on its own — a lift in attributed sales inside a clean room still doesn't separate people who saw your ad and bought anyway from people who bought because they saw it. That gap is why 'sales velocity' and 'incremental sales' get used almost interchangeably in retail media decks, including Costco's, when they're not the same claim.

The fix, wherever you're running paid media, is a holdout or matched-control test, not a bigger attribution window. As one concrete example of what that discipline produces: across 30 Amazon DSP advertisers in July 2026, our own book delivered a 6.04x return on ad spend on 78.4 million impressions at a $4.00 CPM and a blended $1.42 cost-per-click — but the figure that mattered more was the 20.1% of purchases that came from shoppers new to the brand, because that's the segment where display is doing work a search ad wouldn't have done anyway. Ask any retail media network, Costco included, whether their headline ROAS is attributed or incremental. It's rarely the same number.

Common Mistakes — and What to Do When the Numbers Look Wrong

The most common mistake is assuming Costco Retail Media works like a self-serve console you can log into and launch a campaign from the same afternoon. Costco's own hierarchy is explicit: Williamson has said sales velocity is a prerequisite for merchant support, meaning placements get filtered through whether they help a merchant turn product, not just whether a brand wants to buy them. Expect a Grocery Growth conversation, not a self-checkout.

• Treating the LinkedIn page as the program. Costco Retail Media's LinkedIn presence is a marketing feed, not a portal — there's no login there, and the real point of contact is the Grocery Growth team or retailmedia@costco.com.
• Assuming in-warehouse formats close the loop the same way onsite does. TV network impressions and print get folded into the same Habu measurement, but they're modeled exposure, not click-level data — worth asking directly how attribution works for those specific formats.
• Trusting attributed ROAS as incremental ROAS. This is a mistake we've made ourselves on the Amazon DSP side before we started reconciling in Amazon Marketing Cloud — a last-click number that looks great can still be crediting sales that would have happened anyway.

If Costco's reported new-to-brand rate doesn't match your own data, check definitions first — 'new-to-brand' and 'new-to-category' aren't interchangeable, and clean-room lookback windows vary by report. Check also whether you're comparing an onsite-only figure to a blended one; a blended number moves differently even with no real change in sales. And if the lift genuinely isn't there — MVM Amplification didn't move volume against the mailer date — the next step isn't a bigger budget on the same tactic. It's a holdout: comparable stores or members, same offer, a real comparison. If Costco's team can't set that up at your current spend level, that's a real limit of the program worth knowing before committing more of next quarter's budget.