How to Plan a Retail Media Campaign in 2026: An 8-Step Playbook for Amazon, Walmart and Instacart

To plan a retail media campaign in 2026, fix your product pages first, then split budget roughly 70/20/10 across your proven network, a second network, and tests. US retail media spend reaches $71 billion this year, and Amazon and Walmart absorb about 89% of the incremental growth — so network choice sets your ceiling before a single bid is placed.

Most retail media plans fail at the setup stage, not the optimization stage. Teams pick a network because a rep called, pour budget into branded keywords, watch a 6x ROAS come back, and declare victory — while the campaign quietly cannibalizes sales that would have happened anyway.

This is a planning playbook, not a platform tutorial. Eight steps, in order, that a brand team can work through before launch. Each one closes a specific hole that shows up in post-mortems.

Step 1 — Decide what the campaign is actually for

Retail media does three different jobs, and they need different structures, different budgets and different scorecards. Pick one as primary before you open a platform.

• Defend — hold share on your own branded terms and category pages against competitor conquesting. Cheap, high ROAS, low incrementality.
• Grow — win unbranded category searches from shoppers who have not decided on a brand. Lower ROAS, much higher incrementality.
• Launch — buy trial for a new SKU that has no organic rank yet. Worst short-term ROAS, and the only way to get a new item moving.

The mistake is scoring all three against one blended ROAS target. A defensive campaign at 6x and a launch campaign at 1.8x can both be correct. Blend them and you will kill the launch to protect an average.

Write the primary job on the brief. Everything downstream — network, structure, bid, KPI — follows from it.

Step 2 — Audit the digital shelf before you spend a dollar

This is the step teams skip, and it is the most expensive one to skip. Roughly 44% of retail media campaigns send shoppers to product pages that fail basic content standards — incomplete titles, weak imagery, missing bullets — according to digital shelf analysis published by CommerceIQ. You are paying auction prices to deliver traffic to a page that cannot convert it.

Run every SKU you intend to advertise through a five-point check:

1. In stock — and forecast to stay in stock through the flight. Advertising into a stockout burns budget and damages organic rank.
2. Title and bullets — primary keyword in the title, benefits in the first two bullets, no truncation on mobile.
3. Imagery — six or more images, at least one infographic, at least one scale or in-use shot.
4. Ratings — 3.8 stars minimum with a meaningful review count. Below that, paid traffic converts poorly no matter how good the targeting is.
5. Price — competitive within the category at the moment of the flight, not at the moment of planning.

Any SKU that fails two or more of these gets pulled from the media plan until it is fixed. That single rule typically recovers more efficiency than any bid change you will make all quarter.

Step 3 — Choose networks with a 70/20/10 split

The consolidation in retail media is real. EMARKETER's H1 2026 forecast puts US retail media at roughly $71 billion, up about 18% year over year, with worldwide spend near $197 billion — but Amazon and Walmart together take about 89% of the incremental dollars. Amazon alone holds an analyst-estimated 75–77% of the US market.

That argues for concentration, not spray. A workable default:

• 70% to your proven network — for most brands, Amazon. This is where you already have data, rank and a working structure.
• 20% to your second network — usually Walmart Connect. Skai's 2026 research found leading advertisers put 21% of retail media budget into Walmart Connect versus 17% for laggards, which is a small gap that compounds.
• 10% to tests — Instacart, Target Roundel, Kroger, or a retailer-specific network where your category over-indexes.

Adjust for category. A shelf-stable grocery brand should weight Instacart far above 10%; a consumer electronics brand probably should not be on Instacart at all. If you are still deciding which networks belong on the plan, our breakdown of how the major retail media networks actually work covers the structural differences.

Step 4 — Structure campaigns by intent, not by product

The most common structure error is one campaign per product line. It feels tidy and it makes every report useless, because branded and unbranded traffic get averaged together.

Structure by intent instead. Four campaign types, each with its own budget and its own target:

• Branded defense — your brand name and product names. Target: high ROAS, capped budget. This should never be your growth engine.
• Category / unbranded — generic category searches. Target: incremental new-to-brand customers. Expect ROAS 40–60% below branded.
• Competitive conquesting — competitor brand terms. Target: new-to-brand rate. Tolerate weak ROAS if the new-to-brand percentage is strong.
• Product targeting / retargeting — competitor detail pages and complementary products. Target: efficiency, small budget, tight bids.

Give each type a separate budget line so nothing is silently borrowing from anything else. When branded defense and category growth share a budget, branded always wins the auction race and growth quietly starves.

Step 5 — Set bids against real network benchmarks

Set targets from published benchmarks, not from what last quarter's blended number happened to be. Here is where the three major networks currently sit.

Retail Media Network Planning Benchmarks — 2026

Two things to take from this table. First, the branded-versus-unbranded gap is far wider than the network-versus-network gap — which means your campaign structure matters more than your network choice. Second, Instacart's CPC range makes it the cheapest place to buy trial for a CPG launch, even though it will never carry the volume Amazon does.

For a fuller picture of where these numbers sit against the rest of your media, compare them to our 2026 retail media benchmarks across Amazon, Walmart and Instacart and to average ROAS by industry and channel. Retail media ROAS is not comparable to prospecting ROAS on paid social, and treating them as one number is how budgets get misallocated.

Not sure whether your retail media ROAS is real growth or just repurchases you were going to get anyway? Our free paid media audit separates branded defense from incremental growth across your current retail media buys and shows what each is actually worth.

Step 6 — Decide how you will prove incrementality before launch

Incrementality is now the number one measurement problem in retail media. Skai's 2026 measurement research found 75% of advertisers name incrementality as their biggest measurement challenge, cross-channel measurement second at 59%, and only 15% describe themselves as very or extremely effective at measuring retail media performance. Separately, while 67% of CMOs plan to increase retail media investment, only 53% believe their networks provide adequate attribution.

You cannot fix that after the fact. Pick a method during planning:

1. Geo holdout — suppress ads in matched markets and compare total category sales. The cleanest read available to most brands, and increasingly supported natively by retailer networks.
2. Audience holdout — where the network supports it, withhold a percentage of the addressable audience and compare conversion rates.
3. Branded-term pause test — turn off branded defense for two to four weeks and measure total branded sales, not just ad-attributed sales. Uncomfortable, cheap, and frequently revealing.
4. Pre/post with a control category — the weakest option, but better than reading platform-attributed ROAS as truth.

Whichever you choose, write the test window into the media plan before launch and protect it. The most common failure is agreeing to run a holdout and then cancelling it in week two because someone gets nervous about the suppressed market.

Step 7 — Set a review cadence with pre-agreed triggers

Annual retail media plans do not survive contact with a market where CPCs move monthly. Build three review layers and define in advance what each one is allowed to change.

• Weekly — bids, negative keywords, stockout pauses, budget pacing. Tactical only.
• Monthly — reallocation between campaign types and between SKUs. Requires two weeks of stable data before acting.
• Quarterly — network mix, the 70/20/10 split itself, and the incrementality read.

Write the triggers down. Pause any SKU whose in-stock rate drops below 90% for three consecutive days. Shift budget away from any campaign type running more than 30% below target for two consecutive weeks. Pre-agreed triggers prevent the two failure modes of retail media management: changing nothing for a quarter, and changing everything every Monday.

Step 8 — Connect retail media to the rest of the media plan

Retail media is a closing channel. It converts demand that upper-funnel media created, which means its measured ROAS is partly borrowed from campaigns that never get credit for it.

Two practical moves. First, time your retail media flights to overlap with your awareness activity rather than running them independently — CTV and paid social drive the search volume that retail media then harvests. Second, model retail media inside your broader attribution setup rather than reading platform numbers in isolation; our overview of attribution models from last-click to MMM covers how to weight a closing channel without over-crediting it.

If you want that modelling built rather than theorized, that is what our media strategy practice and programmatic team do — and cross-channel benchmarks like cost per lead by industry across channels give you the comparison points to plan against.

The eight-step retail media planning checklist

1. Name the primary job: defend, grow, or launch. One, not three.
2. Audit every SKU on stock, title, imagery, ratings and price. Pull any that fail two checks.
3. Split budget 70/20/10 across proven network, second network, and tests.
4. Structure by intent — branded, category, conquesting, product targeting — with separate budget lines.
5. Set ROAS targets per campaign type from published benchmarks, not blended history.
6. Choose and schedule an incrementality test before launch.
7. Define weekly, monthly and quarterly review layers with written triggers.
8. Align flights with upper-funnel activity and model retail media inside full-funnel attribution.

Frequently asked questions

How much budget do I need to start a retail media campaign?

Enough to reach statistical significance on at least one campaign type, which in most categories means a minimum of $5,000–$10,000 per month per network. Below that, you will spend the flight reading noise. It is better to run one network properly than three networks at a level where none of them produce a readable result.

Should I start on Amazon or Walmart?

Start where your sales already are. If Amazon is your largest ecommerce channel, start there — you have rank, reviews and history to build on. If you have strong Walmart in-store distribution, Walmart Connect gives you an omnichannel read that Amazon cannot. Do not start on both simultaneously unless your budget supports significance in each.

What is a good ROAS for retail media in 2026?

There is no single number, which is exactly the problem with how most brands set the target. Branded keyword campaigns commonly return 5x–8x and prove very little. Unbranded category campaigns typically land at 2x–4x and represent actual growth. Set separate targets per campaign type and judge the portfolio on incremental sales, not on the blended average.

How long before a retail media campaign is worth judging?

Four to six weeks for a launch or unbranded campaign — long enough for the algorithm to gather conversion data and for organic rank to respond. Branded defense reads faster, often within two weeks. Judging any campaign inside the first fortnight almost always produces the wrong decision.

Does retail media cannibalize my organic sales?

Branded defense often does, at least partly. That is not automatically an argument to stop — holding position against competitor conquesting has value — but it is an argument to measure it. A branded-term pause test over two to four weeks, measuring total branded sales rather than ad-attributed sales, will tell you how much of that ROAS is real.

Plan it properly

Retail media rewards planning discipline more than it rewards spend. The brands pulling ahead in 2026 are not bidding harder — they are structuring by intent, protecting their incrementality tests, and refusing to advertise SKUs that are not shelf-ready.

If you want a second set of eyes on your plan, book an intro call and we will walk through your network mix and campaign structure. Or start with a free paid media audit — we will show you where your current retail media spend is buying growth and where it is buying sales you already had. You can also see the full range of what we do on our services page.