Hero SKU concentration: when one product carries the account.

What happens to your business if your best product disappears for a month?

Not the brand. Not the store. One product. The one that wins the Shopping auctions, carries the best Meta ads, and brings in most of the people who have never bought from you before.

Most owners can answer that question in revenue terms. Fewer have asked it in acquisition terms, and that is where the real exposure sits. A hero product is not only your best seller. It is usually the front door. If the front door closes, the rest of the catalog loses the visitors it was quietly relying on.

None of this is an argument against heroes. Many accounts scale on the back of one product the market wants more than the others. The argument is narrower. Concentration is a number, it can be measured in an afternoon, and past a certain level it is a risk your ROAS report is structurally unable to show you.

Three shares, not one

A hero SKU is usually measured by revenue share: the top product is 55% of sales, so the brand is concentrated. That number is useful and incomplete. Paid media concentration has three parts, and they tell different stories.

• →Share of ad spend. How much of the budget sits behind the top product. In Shopping and Performance Max this is often the platform’s decision more than yours, because bidding goes where the conversions are.
• →Share of ad revenue. How much of the revenue your ads are credited with comes from that product. If this runs well above the spend share, the hero is subsidizing the rest of the catalog.
• →Share of new customers. How many first-time buyers made their first purchase on that product. This is the one that matters most, and the one almost nobody computes.

The new-customer share needs your order data, not your ad platform. Take every first order in the last 90 days, matched against your customer list so returning buyers are excluded, the same discipline behind new-customer CAC. Then look at which product was on it. If an order holds several products, count the one with the highest line value. It is a simplification, and it is good enough to see the shape.

Measure at the level the customer experiences. Five sizes of the same shoe are one product for this exercise. Two different shoes that share a name are not.

A worked example

Here is an illustrative brand, the arithmetic rather than a client result. Five products, $150,000 a month in ad spend, 90-day figures expressed as monthly averages.

Read the ROAS column and this account is a success story with a clear winner. Read the last column and it is a business with one door. Three in four new customers arrive through a single product. The hero’s share of ad revenue, about 70%, sits between the other two. That ordering follows from how people shop: returning buyers explore the catalog, while first-time buyers mostly arrive on the product the ads show them.

If you want one number for the whole catalog, borrow the Herfindahl-Hirschman Index, which US antitrust authorities use to measure market concentration. Square each product’s percentage share of spend and add them up. Here that is 64² + 16² + 10² + 6² + 4² = 4,504. Five products splitting the budget evenly would score 2,000. One product alone scores 10,000. Track it monthly. The direction matters more than the level.

The five risks the ROAS column does not price

A high ROAS on the hero tells you it is efficient today. It says nothing about how fragile that efficiency is. Five things can take it away, and each hits harder the more the account depends on one product.

A stockout

When a product is marked out of stock in the feed, it stops being advertised. Meta’s catalog documentation is explicit that out-of-stock items do not display in ads at all. That is the right behavior. For a concentrated account it means acquisition drops by roughly the hero’s share of new customers on the day inventory runs out. A reorder that lands three weeks late is a three-week acquisition gap.

Creative fatigue

When one product carries the account, one product’s creative carries it too. The same audience sees variations of the same offer, and fatigue arrives sooner than it would across a spread of products. More angles on the hero works, until it does not.

A competitor copy

A winning product is visible. Its ads are in public ad libraries and its listings rank in Shopping results. A cheaper imitation does not need to be better to raise your costs. It only needs to enter the same auctions.

A margin change

A supplier price rise, a tariff, or a freight change on the hero moves the breakeven ROAS of most of your spend at once. If the hero’s contribution margin falls from 50% to 40%, its breakeven ROAS rises from 2.0x to 2.5x, and in this example 64% of the budget now has to clear the higher bar. The breakeven ROAS calculator does that arithmetic per product in a minute.

Learning tied to one product

Automated bidding learns from conversions. Google says the length of a learning period depends on the number of conversions your campaigns, ad groups, keywords, or products obtain, and that a bid strategy typically needs one to two conversion cycles to calibrate. If most conversions come from one product and that product goes dark, the bid strategy loses most of its signal at once, and the rest of the catalog is bid on thin data while it relearns.

What losing the hero costs, as arithmetic

Model the stockout first, because it is the risk with a date attached. Take the example above and assume the hero is unavailable for a full month.

The second row keeps the other products’ efficiency and loses 70% of revenue. The third keeps spend and loses a third of revenue, because the $96,000 has to go into products that were already getting what the platforms judged they could use. I assumed the moved money returns 70% of their current rate, to stand in for diminishing returns. Your own marginal ROAS on those products would replace that guess with a measured figure.

Either way, new customers fall by 44% to 75% in a month. That is the number to put in front of whoever manages inventory, because it turns a purchasing decision into a growth decision.

A decision rule for concentration

Here is the rule I use. The thresholds are a rule of thumb from running accounts, not a statistic. Apply them to the new-customer share first, because that is where the dependency lives.

• →Under 40% of new customers on one product: healthy. Keep feeding the hero. Review the shares quarterly.
• →40% to 60%: concentrated. Choose a second hero and write its budget line now, launch the program in January so peak weeks do not distort it, and check the hero’s stock cover every week.
• →Over 60%: exposed. Make a second hero a priority, agree a stock floor for the hero with operations, and write down what the account does on the day the hero goes out of stock.

Two conditions move you up a band. If the hero’s stock cover is shorter than its reorder lead time, or its cost price is under negotiation, act as if you were one band higher. Both make the bad month more likely, not only more expensive.

Spend share and revenue share are supporting evidence. If the hero’s spend share runs far above its new-customer share, the platforms are concentrating budget on it to reach people who already buy. That is a retargeting question, not a portfolio one.

What common advice gets wrong

The usual advice is to feed the winner and cut the losers. It is half right. Cutting products that lose money after margin is sound, and a product-level profit review is how you find them. The mistake is treating the ROAS ranking as the whole portfolio decision.

A ROAS table ranks products by how efficient they were last month. It carries no information about how tied your outcomes are to one supply chain, one supplier price, one creative theme, and one set of auctions. Nobody in finance would call a portfolio with 75% in one stock well managed because that stock had the best quarter. The return is real. So is the exposure. Paid media reporting shows the first and ignores the second.

Some of the products at the bottom of the ROAS table are losers. Some are future heroes being judged on a budget too small to show what they can do. The concentration number is how you tell a review from a purge.

How to build a second hero without starving the first

The failure is familiar. A brand decides to diversify, moves budget away from the hero, watches blended efficiency fall, panics, and moves it back. The hero never lost its place. The challenger never got a fair test. Four rules prevent that.

1. Fund the challenger from the tail, not from the hero. Take budget from products already below breakeven, or use growth money you would have added anyway. The hero keeps its budgets and targets, so its learning is not disturbed.
2. Give it its own line. In Shopping and Performance Max, tag the candidate with a custom label in Merchant Center and give it its own campaign or listing group, so its spend is set by you rather than absorbed by the hero. On Meta, give it its own product set and ad set. As a rule of thumb, 10% to 20% of spend is enough to learn something without putting the month at risk.
3. Judge it on new customers, not blended ROAS. The challenger will not match the hero’s ROAS, because the hero has reviews, creative learning, and search demand behind it. The fair comparison is cost per new customer against the hero’s, and whether the challenger’s buyers come back.
4. Time-box it. Eight to twelve weeks, with the graduation rule written before you start. Mine: the challenger graduates when it brings at least 15% of new customers at a new-customer cost within 25% of the hero’s. If it has not moved by the deadline, pick the next candidate rather than adding budget.

Choosing the candidate is usually easier than it sounds. Look for a product with a decent conversion rate on low traffic, a margin at least as good as the hero’s, and stock you can reorder quickly. A higher-margin second hero is also one of the few ways to lift blended margin without a price rise, which is the constraint the $2M to $3M a month rung is really about.

When this advice is wrong

Some businesses are one product on purpose. A brand built around a single flagship, where the rest of the catalog is accessories, should not force a second hero into existence. Its concentration risk is managed through supply, creative depth, and margin protection. For that brand the decision rule becomes a stock floor, not a diversification program.

It is also wrong for early accounts. Until the hero has saturated its own demand, splitting budget spreads thin data across more products and slows everything down. If the next dollar on the hero still clears breakeven comfortably, that dollar belongs to the hero.

And it is wrong when the concentration is an artifact of counting. If the report splits one product into six SKUs, it may look broader than it is. Group before you worry, in either direction.

What to do this month

1. Pull 90 days of first orders, matched against your customer list, and compute the three shares for your top product: spend, ad revenue, and new customers.
2. Check the hero’s stock cover against its reorder lead time. If cover is shorter, raise it with operations now, before peak demand draws it down.
3. Place yourself in a band and write down the trigger that would move you to the next one.
4. If you are above 40%, choose one challenger, tag it, and write the graduation rule and the date you will judge it. Fund it from the tail and launch it in January, because a test that runs through Black Friday and December reads the peak, not the product.
5. Model the stockout month for your own account, using the second table as the template.

November is the month concentration risk is most likely to arrive, because demand spikes while reorders slow down. If you would rather work through the product-level numbers with someone who reads them daily, that is where our Google Shopping ads work starts.