We were spending a third of the budget on 5% of the demand

I was recently asked to look at a food brand's advertising on one of the quick-commerce apps. Healthy business, healthy budget, and a blended return comfortably above 3×. Nobody thought there was a problem — the blended number looked fine, so nobody had looked underneath it.

Underneath it, a third of the entire budget was going to one product that accounted for about 5% of what people were actually searching for.

The gap you can only see in three columns

The audit was not sophisticated. It was one table, built in an afternoon, with three columns per product:

1. Share of spend — what percentage of the ad budget this product consumed
2. Share of sales — what percentage of ad-driven revenue it returned
3. Share of search demand — what percentage of category searches were for it

Individually, those three numbers tell you nothing surprising. Side by side, they tell you where the money is in the wrong place. You're looking for products where the first column is much bigger than the third — you're funding demand that doesn't exist — and products where the third is much bigger than the first, which is demand you're leaving to competitors.

Two rows, and the entire media plan was wrong. A third of the budget was defending a product almost nobody was searching for, while the product a fifth of the category wanted was getting two percent of the money.

Nobody chose this. That's the thing worth understanding — it accreted. That product had been the hero of a festive push a year and a half earlier, the campaign was never wound down, and budgets have a way of staying wherever they were last put.

The second finding was worse

On a smaller brand in the same account, roughly 90% of keyword spend was going to keywords that more than one of its own campaigns was bidding on.

If that sentence doesn't immediately alarm you: when two of your campaigns bid on the same keyword, they enter the same auction. The platform doesn't know they're related. They compete, and the winner is whichever of your own campaigns bid higher — which means you set the price you pay, against yourself, and you pay it.

You do not see this in any standard report. Campaign-level reports show each campaign performing acceptably. The waste only appears when you pull every keyword across every campaign into one sheet and look for duplicates. It took twenty minutes and it was the single most expensive problem in the account.

What the fix was worth

Reallocating spend toward demand, and de-duplicating the keywords so campaigns stopped bidding against each other, was worth somewhere between 15% and 30% of the budget depending on the brand — recovered without asking for a single additional rupee.

I want to be precise about what "recovered" means, because this is where these claims usually get slippery. It doesn't mean we found free money. It means that much spend was moved from places returning below the account average to places returning above it. The budget didn't change. The output did.

Almost every underperforming ad account I have opened had more money available inside it than the client was about to ask for. Reallocation is unglamorous and it is nearly always the first move.

Do this yourself, this week

You don't need an agency or a tool. You need a spreadsheet and about two hours.

1. Export the last 90 days of spend and sales by product or campaign. Not 30 — you'll be reading noise. Not 365 — you'll be reading history.
2. Add a share-of-total column for each. Spend as a percentage of total spend, sales as a percentage of total sales.
3. Get a demand number. On marketplaces, search-term reports give it to you directly. Elsewhere, Google Keyword Planner or even the platform's own search suggestions will get you close enough. You need the ranking to be roughly right, not the absolute numbers.
4. Sort by the gap between spend share and demand share. Look at the top three and bottom three rows. That's your reallocation.
5. Separately, pull every keyword you're bidding on into one column and check for duplicates. Any keyword appearing in two campaigns is you bidding against yourself. Pick one campaign to own it and add it as a negative in the other.

Two things I'd warn you about

Don't cut a product to zero because its gap looks bad. Some spend is defensive — you're holding a position against a competitor, and the moment you leave, the cost of coming back is higher than what you saved. In this account we protected the top products entirely and only redistributed from the middle. The rule I use: never cut something that is both profitable and category-leading, however inefficient it looks.

Demand share is directional, not gospel. Search volume tells you what people type, not what they'd buy if they saw it. A product with no search demand and excellent conversion is a discovery product, and it needs a different kind of campaign rather than a smaller budget.

The underlying point

Blended numbers hide their own composition. A 3× account average can be one campaign at 6× subsidising four at 1.5×, and the average tells you nothing about which. Every account has a distribution inside it, and the money is almost always in fixing the distribution before adding to the total.

Check the gap before you ask for more budget. In my experience you'll find somewhere between 10% and 30% of your spend is in the wrong place, and it costs you nothing but an afternoon to move it.