Ten categories — ten campaigns? Each of them needs 15 conversions a month

In short: what you’ll take away

• Who it’s for: owners of stores with several product lines running Shopping or Performance Max.
• What you’ll learn: what technically decides which product gets shown for a query, and why it isn’t campaign structure.
• The outcome: three checks after which you’ll see for yourself how many campaigns your account should have — and whether you have the data for more.

1. The question that derails half our conversations

A few months ago a store owner — let’s call him Yaroslav — agreed to a free catalogue review. Nine days later he pulled out and revoked access. Not because he disliked anything in the report. He simply went and googled, and here’s what he found:

Verbatim, from our call

“Everyone everywhere writes that dumping all the products from different lines into one campaign is somehow a bit wrong.”

“I’m still looking for information, but everywhere I look… everyone says unanimously that dumping kitchen sinks, knives and camping gear into one campaign is absurd. I have 10 product lines that have nothing to do with each other.”

Underneath this objection lies a specific technical question, and it’s a perfectly fair one: how will Google decide who to show a frying pan to and who to show a sleeping bag, if all the products sit in one campaign?

It’s an honest question. And for a long time we answered it badly — we said “our algorithm labels the products”, which to someone who checks information themselves sounds like a dodge. That’s why this article exists: here is the full answer, with references to Google’s documentation rather than to our opinion.

2. Why “split by category” sounds logical

The logic is simple and at first glance flawless: different products mean a different audience, a different margin, a different return. So let’s split them across campaigns, give each its own budget and its own target. Sinks separately, knives separately, tents separately.

The advice isn’t nonsense. It just answers a different question from the one that actually worries the owner. People hear “split the campaigns” and think that’s how they control impressions — who gets shown what. But a campaign controls something else entirely.

3. What technically decides who sees your frying pan

Here’s how Google itself puts it in the Shopping ads help:

Google Ads Help, “About Shopping ads”

“Shopping ads use your existing Merchant Center product data (not keywords) to decide how and where to show your ads.”

In other words: Shopping ads use product data from Merchant Center — not keywords — to decide how and where to show the ad.

Read that again, because it’s the whole point. The match between the query “3-season sleeping bag” and your product is established from product data: title, description, category, product type, brand, feed attributes. Not from which campaign the product sits in.

Someone searching for a sink will never be shown your tent — regardless of whether they sit in one campaign or ten different ones. Not because it’s configured that way, but because a tent doesn’t match a query about a sink. Campaign structure physically takes no part in that decision.

What happens if the same product ends up in several campaigns at once? Google describes this too: in Standard Shopping campaigns the conflict is resolved by campaign priority — “the campaign with the higher priority will bid” — and if the priority is the same, the highest bid is used. If the higher-priority campaign has run out of budget, the next one bids. And when Performance Max and a Standard Shopping campaign run on the same products in one account, the impression goes not to the “more important” campaign type but to “the campaign with the highest Ad Rank”.

Again, not a word about categories. Everywhere it’s about budget and bid.

4. What a campaign actually controls: budget and learning

A campaign isn’t a way of telling Google who to show a product to. It’s a container that determines two things: how much money is allocated and how much data the automated bidding learns from.

And this is where splitting by category stops being neutral. Automated strategies have a formal threshold. Google’s help on Target ROAS for Search and Shopping campaigns:

Minimum data for Target ROAS

“At least 15 conversions in the past 30 days at the conversion tracking level.”

At least 15 conversions in the past 30 days. This isn’t a “works better this way” recommendation — it’s a condition of access to the strategy.

Now just multiply. Ten product lines spread across ten campaigns means 150 conversions a month purely so that each of them qualifies for sensible bidding. Not “preferably” — minimum.

A store asking this question usually has 30–60 conversions a month across the whole account. Divide that by ten and you get three to six per campaign. Each one lands below the threshold, each one learns from noise, and none accumulates enough statistics to understand anything about your demand.

That’s the real price of splitting. Not “Google will get confused about who to show the tent to” — it won’t. It’s ten blind campaigns instead of one that can see.

In one sentence

Category governs query matching — and lives in the feed. Campaign governs budget and learning — and lives in the account. By splitting campaigns along categories you’re trying to control the first using the second. The first already works, and the second is what you break.

5. So what axis should you cut along, if not category

Here we move from “why not” to “how instead”. And the data gave us the answer.

We analysed 1,404,808 products across 134 stores — around 17 million clicks over 13 months. Advertising budget went to 45% of products, and conversions to only 17.6% of those that received budget. In the three reports we examined most closely, the same picture appears from another angle: 4–8% of products generate 80% of revenue. In a store with 827 products, six SKUs — six — produced 28% of all revenue.

The key isn’t the concentration itself but where those products sit. They do not cluster in one category. They’re scattered across every line: a couple of items among the sinks, one among the knives, three among the camping gear. And right next to them, in the very same category, sit products that eat budget for months with zero sales.

So within a single category the difference between products is enormous, while across categories strong products behave similarly. Category is a weak predictor. Performance is a strong one.

That’s why we cut not by category but by how a product behaves: sells steadily, sells weakly, eats budget without sales, has no history yet. And those groups genuinely need different budgets and different targets — unlike sinks and knives, which need exactly the same thing.

The full numbers from this sample — ROAS benchmarks by category, how many clicks a product needs before its first sale, and how long a bestseller lives — we published separately: a study of 1,404,808 products across 134 stores.

This is not “everything in one campaign”

We split accounts too. Just along a different axis. A product that sells steadily and a product that has only spent for six months will go into different campaigns with different targets — even if they’re two tents from the same shelf. Meanwhile a tent and a knife that both sell steadily live together quite happily: they need the same regime, and together they produce statistics neither would have alone.

6. Three checks: how many campaigns you specifically should have

This isn’t a matter of taste. There’s arithmetic for it, and you can run it yourself — all the data is in your own Google Ads reports.

1. Count your monthly conversions across the account. That’s the ceiling on your structure. The rule is simple: a campaign earns its right to exist if it gets at least 15 conversions a month — Google’s official threshold. In practice we keep a margin and aim for 50, because at 15 the strategy works but still reacts very nervously to a random week.

Up to 50 conversions a month across the account means 1–2 campaigns, and no category splitting at all. 50–150 — two or three. 150–300 — three or four. Over 300 — four or five is fine.

2. Check whether your axis is even populated. Take the attribute you want to cut along (category, brand, product type, price range) and look at what share of the catalogue has it filled in the feed. If fewer than half the products carry that value, it isn’t an axis. Most of the catalogue will fall into “(empty)”, and you’ll be cutting along an attribute you don’t actually have.

3. Check whether there’s a real difference between the groups. Calculate cost per conversion separately for each group over 12 months. We only care about groups with at least 12 conversions in the year — the rest don’t statistically exist. If the spread in cost per conversion between meaningful groups is less than 1.3×, splitting is pointless: you’ll get several campaigns that behave identically, each with a third of the statistics.

A tip from practice: always test price range as an axis, even if you never considered it. Every product has a price, and it regularly divides a catalogue more sharply than category does.

If your category axis passes all three checks — split by category, it’s justified. In our practice it passes rarely, and almost always in large accounts. But “rarely” isn’t “never”, and that’s exactly why we calculate it rather than deciding in advance.

7. When splitting by category is genuinely needed

So as not to leave the impression that we defend a single campaign at any cost, here are the situations where you should split:

• Each line produces 50+ conversions a month on its own. Then splitting breaks nothing — there are enough statistics for everyone.
• The lines really do have different return targets. Not “different margins in the owner’s head”, but a different target ROAS you can name as a number and justify.
• Pronounced seasonality in one line. A product that sells three months a year drags the averages of a shared campaign down out of season.
• An external budget constraint. For example, a supplier co-funds advertising for a specific brand — then you need a separate campaign purely for accounting.

What all four have in common: the reason is budget or target, not a wish to “keep products from mixing”. Mixing isn’t the problem. The problem is a smeared budget and fragmented statistics.

8. “Yes, but…” — honest answers to the counterarguments

“Every specialist online advises otherwise.” Some of them work with accounts doing 500+ conversions a month — there, splitting really is safe. The advice is correct, but it comes from a different scale. Ask the author how many conversions a campaign needs for Target ROAS to work. If there’s no answer, the advice isn’t about your account.

“My products have different margins, they can’t be mixed.” They can and should be — but at the data level, not the campaign level. Margin is passed into the feed and taken into account in conversion value. Then automated bidding sees the difference for each product individually, instead of you approximating it with ten category averages.

“What if Performance Max gives the whole budget to one line?” That’s a real risk, not an invented one. But the cure isn’t splitting by category — it’s controlling which products are admitted to the campaign at all, and a separate budget for those with no history yet. In other words, the same performance axis.

“I’m still uncomfortable with sinks and tents in one campaign.” That’s normal, and it’s the most honest answer of them all. So we don’t ask you to take it on faith: run the three checks from section 6 on your own data. If the axis passes — we split by category, that’s your case.

9. Summary

• Who gets shown your product is decided by product data in the feed, not by campaign structure — it’s stated outright in Google’s help.
• A campaign controls budget and bid learning. The minimum for Target ROAS is 15 conversions in 30 days. Ten campaigns require 150.
• Revenue concentrates in 4–8% of products, and they’re scattered across every category — which makes performance a stronger axis than category.
• Splitting by category is justified when each line reaches 50+ conversions a month on its own or has a separate return target.

The worst scenario is neither “one campaign” nor “ten”. The worst is splitting blind, without calculating whether you have enough data. So start with the three checks: how many conversions, is the axis populated, is there more than a 1.3× difference between the groups.

🔍 Look at your catalogue through the performance lens

Shopping X-Ray connects to Google Ads in read-only mode and shows how revenue and budget are distributed across your products: which sell steadily, which eat budget with zero sales, and which could sell but get no impressions. Free, no commitments.

X-ray your catalogue →

Sources

• About Shopping ads — Shopping ads use product data, not keywords.
• About Target ROAS bidding — a minimum of 15 conversions in 30 days for Search and Shopping campaigns.
• Use campaign priority for Standard Shopping campaigns — how the conflict is resolved when a product sits in several campaigns.
• How Performance Max interacts with other campaigns — the impression goes to the campaign with the highest Ad Rank.