Plenty of guides will tell you what a retail media network is and why the category is booming. We've written one ourselves. What almost none of them tell you is the part that actually determines whether you make money: how to build and run the campaign once you've decided to show up. This is that guide — the mechanics under the hype, aimed at the person who now has to log in and make something work.
If you're still deciding whether retail media is for you at all, start with why brands are paying attention to retail media, which covers the what and the why. Assume that's settled. From here on, we're inside the account.
How the machine actually works
At its core, a retail media network runs an auction on a retailer's own property. When a shopper searches or browses, advertisers bid to place their product in a sponsored slot, and the retailer's first-party purchase data decides who's most likely to convert. The reason this converts so well is position in the journey: you're reaching someone at the digital shelf, wallet already out, rather than interrupting them mid-scroll somewhere else — the same intent advantage that makes social commerce convert where it does, applied to the retailer's own storefront.
That's the whole engine — an auction, priced per click, settled with data nobody else has. Everything that follows is about feeding that auction the right inputs so it works in your favour instead of quietly draining budget. And the first input has nothing to do with ads at all.
Before you spend a penny: the listing is the ad
The most expensive mistake in retail media is running traffic to a listing that can't convert. Unlike search ads that point to a landing page you control, a sponsored product ad sends the shopper to your retail listing — so that listing is your conversion page, and if it's weak, every click is a partial refund to the retailer.
Before launching anything, the product page needs strong images, a clear title, complete details, and credible reviews, because the ad only buys the visit; the page has to close the sale. The same discipline we lay out in product page optimization applies doubly here, because now you're paying for every visitor who lands on a page that isn't ready. Fix the shelf before you pay to drive people to it.
Tidio
Popular Shopify live chat and AI chatbot tool that converts visitors into buyers
Best for: Live Chat & Chatbot
The rule that saves budgets A great campaign on a bad listing loses money faster than no campaign at all — you're paying, at auction prices, to prove your page doesn't convert.
The account structure that makes optimization possible
Here's the single most important build decision, and the one beginners most often get wrong. The temptation is to throw every keyword into one campaign and let it run. Doing that quietly destroys your ability to optimize, because you can no longer control bids or budget by how different keywords actually perform — the winners and the wasters share one pot.
The structure that works separates campaigns by match type, each with a different job:
| Campaign | Match type | Its job | Rough budget share |
|---|---|---|---|
| Discovery | Automatic / broad | Find search terms you didn't know converted | ~15–20% |
| Expansion | Phrase | Scale terms that show promise | ~30% |
| Proven | Manual exact | Your winners, tighter bids, best return | ~50% |
The magic isn't in any one campaign — it's in the flow between them. Discovery finds raw terms, the promising ones graduate to expansion, and the consistent winners get promoted to the proven campaign where you can bid confidently because you know they pay back. This is the architecture the rest of the account depends on. Get it right and everything else becomes possible; skip it and you're flying blind.
Keyword harvesting: the loop that compounds
That flow between campaigns has a name — keyword harvesting — and it's the closest thing retail media has to a growth engine. It's a weekly habit, not a one-time setup.
→ Pull the search-term report — roughly every seven days. Weekly is the sweet spot: enough data to spot real trends, not so frequent you overreact to noise.
→ Promote the winners. A term converting consistently in discovery gets added as a manual exact keyword in the proven campaign, where it earns a controlled, confident bid.
→ Block the overlap. Add that same term as a negative exact in the discovery campaign — otherwise your own campaigns bid against each other and inflate your costs.
→ Cut the wasters. Terms burning spend without converting become negative keywords. This is the single most underused profit lever in retail media — most advertisers add negatives far too late.
Repeat weekly. Winners concentrate, waste gets pruned, and the account gets more efficient with every cycle.
Negative keywords deserve the emphasis. Most accounts leak money not through bad bids but through paying for irrelevant searches nobody bothered to exclude. A disciplined negative-keyword habit does more for profitability than almost any bid adjustment, and it costs nothing but attention.
What the AI changed, and what it didn't
By 2026, the major networks give you bid automation free and native, which has genuinely changed the job — but not in the way the marketing suggests. It hasn't removed the work; it's moved it.
The automation optimizes toward conversions within the parameters you hand it, which means your leverage is entirely in the quality of those inputs: a clean campaign structure, strong negative lists, retail-ready listings, and sensible budget thresholds. Give it good inputs and it's a real force multiplier. Give it a messy single-campaign account with no negatives and a weak listing, and it will optimize your way into confidently spending money on the wrong things. The lesson mirrors what we found watching AI move into ads-manager tools elsewhere: automation rewards the operator who sets it up well and punishes the one who uses it to avoid setting it up at all. Your job shifted from tweaking bids to engineering the conditions the AI optimizes inside.
It isn't only Amazon — and that's an opportunity
Amazon takes the lion's share of retail media spend, which leads a lot of brands to assume it's the only game. The more interesting money is often elsewhere. Grocery chains, big-box retailers and category specialists all run their own networks, and because they're less saturated, clicks are frequently cheaper and a dominant share of voice is actually achievable — something close to impossible on Amazon.
They do differ in the mechanics, and it's worth knowing before you commit. Some are fully self-serve; others are managed-service with real minimum spends that can run to tens of thousands a month, putting them out of reach for smaller brands. And on networks where far fewer people search, manual keyword-loading matters less and the retailer's audience and contextual targeting matter more — you're buying access to their shopper data as much as their search box. The right network isn't the biggest one; it's the one where the people who buy your category actually shop, a point the shift in where retail media budgets are moving makes at the portfolio level.
A launch sequence that works
Pulling the mechanics into an order you can actually follow:
- Fix the listing first — images, title, details, reviews. The ad buys the click; the page makes the sale.
- Pick one network where your customers genuinely shop, and go deep rather than spreading thin.
- Build the three-campaign structure — discovery, expansion, proven — separated by match type from day one.
- Start with sponsored products, the lower-funnel format that captures existing demand and proves return fastest.
- Harvest weekly — promote winners, negate wasters, block overlap. This is the whole job, repeated.
- Set clean automation parameters and let the AI optimize inside them, rather than in place of them.
- Only then expand — into display, other formats, or a second network — once the first is provably profitable.
If that's more account management than you want to run in-house, having it built and harvested weekly is exactly what e-commerce marketing support is for.
The part that actually matters
Retail media works the way most performance channels work once you get under the branding: an auction that rewards whoever feeds it the cleanest inputs. The winning inputs here are unglamorous and completely learnable — a listing ready to convert before you spend, a campaign structure separated by match type so you can actually steer, a weekly harvesting habit that concentrates winners and prunes waste, automation set up deliberately rather than trusted blindly, and the sense to fish where your customers swim instead of defaulting to the biggest pond. None of that is secret and none of it is hard; it's just the work the explainers skip because it isn't exciting. Do it, and retail media stops being a line item you hope is working and becomes a channel you can actually run.
Retail media spend you're not sure is working?
We build the structure and run the weekly harvesting that makes it profitable.
Explore E-commerce Marketing →Frequently asked questions
How do you structure a retail media campaign?
Separate campaigns by match type rather than mixing them. The standard architecture uses an automatic or broad-match campaign for discovery, a phrase-match campaign for expansion, and a manual exact-match campaign for your proven, highest-converting keywords with tighter bids. Mixing all match types in one campaign is the most common structural mistake, because it removes your ability to control bids and budget by how each keyword actually performs. Separation is what makes optimization possible at all.
What is keyword harvesting in retail media?
Keyword harvesting is the core optimization loop. You let a discovery campaign surface which real search terms convert, then promote the consistent winners into a manual exact-match campaign as permanent targets, while adding them as negatives in the discovery campaign so the two don't bid against each other. Reviewing the search-term report roughly weekly and harvesting winners is what turns a static campaign into one that compounds — the single habit that most separates profitable accounts from stagnant ones.
Do you still need to manage bids now that retail media has AI automation?
Yes, but the job has shifted. Basic bid automation is now a free native feature on the major networks, so manual bid-tweaking matters less. What matters more is giving the automation the right inputs: a clean campaign structure, strong negative-keyword lists, retail-ready listings and sensible budget thresholds. The AI optimizes toward conversions within the parameters you set, so your leverage is in setting good parameters, not babysitting individual bids. Bad inputs produce confidently optimized waste.
Is retail media only worth it on Amazon?
No, though Amazon dominates spend. Other networks — from large grocery and big-box retailers to category specialists — often have lower competition and cheaper clicks, which can mean strong share of voice that's impossible on Amazon. They differ in mechanics: some are self-serve, others managed-service with meaningful minimum spends, and on networks with sparse keyword data, contextual and audience targeting matter more than manual keyword-loading. The right network is the one where your actual customers shop, not the biggest one.
