Where the money actually went.

Ask a seller where their wasted ad spend is and most will say they are not sure, but probably a bit everywhere. That is almost never true. Waste concentrates. On a typical account a small number of search terms account for most of it, which is good news, because a concentrated problem is a solvable one.

The reason it feels diffuse is that campaign-level reporting cannot see it. A campaign at a 32% ACOS looks acceptable. Inside it, forty terms convert well and eleven have spent steadily for three months without a single order. The average absorbs them. Nothing looks wrong, so nothing gets done, and the same eleven terms spend again next month.

What counts as wasted, precisely

It is worth being strict, because a loose definition produces a list too long to act on. Wasted spend is clicks you paid for that produced no order, on terms with enough clicks that the absence of orders is a signal rather than bad luck.

That second clause does the work. A search term with three clicks and no order tells you nothing at all — plenty of converting terms have three-click dry spells. The threshold worth using is roughly the reciprocal of your conversion rate: the number of clicks at which you would normally expect one order.

Below your threshold, leave it alone and let it gather more data. Above it, the term has had a fair run and declined to convert. That is a result, not an accident.

The layer underneath: one word, forty terms

Sorting the search term report by spend and negating the worst offenders is the obvious move, and it will find real money. It will also miss the larger, quieter leak, because that leak does not live in any single row.

Take a word like "free", or "used", or "for kids", or a competitor's brand name. It might appear in forty different search terms, each spending too little to draw attention. Individually every one of those rows looks like noise. Together they can be a meaningful share of the month.

N-gram analysis finds them. It splits every customer search term into its component words and word pairs, then totals spend, clicks and orders per fragment rather than per whole term. It is the same export read a different way, and it surfaces something scanning the term list never will — because no individual row looks wrong.

The check before you act: look at the orders column on the fragment, not just the spend. A word can carry both waste and revenue. "Organic" might be draining money across thirty terms and also sitting inside your best converting phrase. Negate it at campaign level and you stop the winner along with the losers.

What the wasted spend is worth in real money

Abstract percentages make this feel smaller than it is, so here are actual numbers — the same ones our calculators use throughout.

That is a good month. Now suppose a fifth of the spend went to terms past their click threshold with no orders — which is not an unusual finding on a first audit. That is ₹3,680 that produced nothing.

Nothing about that requires selling more, raising a bid, or improving a listing. It is the same revenue with a cost removed, which is why it is the first thing worth doing on any account and the thing most often postponed.

The other half of the same report

It would be a shame to open the search term report only to find bad news, because it holds both halves of the opportunity and people tend to stop after the first.

The terms converting well inside automatic and broad campaigns are candidates for promotion to their own exact-match targeting, where you set the bid deliberately instead of paying whatever the loose match decided. A harvest candidate is a term with enough clicks to be credible, a conversion rate at or above your account average, and an ACOS under break-even.

Two traps here. Harvesting on order count alone finds expensive coincidences — three orders from ninety clicks is not a winner. And promoting a term without negating it in the campaign that discovered it leaves two of your campaigns bidding for the same query, which costs more than doing nothing would have.

A workable cadence

Search terms and negatives deserve a fortnightly pass, because new queries enter automatic campaigns continuously and the waste rebuilds. Structure, budget allocation and placement bids move more slowly and monthly is usually enough.

Pull at least thirty days, and do not compare a fresh export against a settled one. Attribution lags — sales keep arriving for up to fourteen days after the click — so a recent period always looks worse than it turns out to be, and a term that looks dead in a seven-day-old pull may simply not have been credited yet.