The Performance Max search terms report sits inside the campaign, under the Insights tab, listed as search terms or search categories depending on which rollout your account is on. Click into the campaign, open Insights, and look for the search terms card. It shows the queries Google matched to the Search portion of your PMax spend, grouped into categories, with impressions, clicks and conversions against each. What it does not show is cost per query. That single omission is why most people give up on the report, and it's also why the workaround below matters more than the report itself.
We run PMax for D2C brands on Shopify across India, the UAE and the US, and the pattern is consistent: somewhere between a tenth and a third of PMax clicks are queries you would never have bid on deliberately. Gold rate today. Amazon. Careers. Repair near me. The report will tell you they exist. It won't tell you what they cost. Here's how we close that gap.
Estimating the waste when Google won't give you cost
Export the search categories table. You get clicks per category. Then take the campaign's average CPC from the main campaign view. Multiply. It's crude, because branded queries cost far less than generic ones, but it's directionally right and it's enough to decide whether a category is worth acting on.
A jewellery client's report showed 1,340 clicks in a month against queries around live metal rates. Campaign average CPC was ₹9. That's roughly ₹12,000 a month going to people checking the day's gold price. Zero conversions attached. Small enough that some founders would shrug at it, large enough that over a year it's ₹1.4 lakh, and the real cost is worse than the cash because those clicks also train the campaign's audience signals toward window shoppers.
The categories with no conversions and meaningful click volume are your list. Everything else can wait.
Isolate brand queries before you touch anything else
PMax will happily serve on your own brand name. Those clicks are cheap, they convert at a high rate, and they make the campaign's blended ROAS look excellent while telling you almost nothing about whether PMax can find new customers.
Do the arithmetic on your own account. Say the campaign spends ₹6,00,000 a month and reports 6.2 ROAS, so ₹37.2 lakh in revenue. The performance max search terms report shows brand queries account for a chunk of conversions; pair that with a rough ₹60,000 of brand-query spend and ₹14 lakh of brand-driven revenue. Strip both out: ₹23.2 lakh on ₹5,40,000, which is 4.3 ROAS. That's the number that tells you whether to scale. The 6.2 was a blend of a campaign doing an ordinary job and a brand term doing what brand terms do.
Now the unwelcome part. If you exclude brand from PMax and don't run a separate brand Search campaign, revenue drops and you'll blame the exclusion. It didn't vanish. It moved, or it went to organic, or a competitor picked it up at ₹6 a click. Set up the brand Search campaign first, let it accumulate a week of data, then apply the brand exclusion to PMax. In that order.
Google's brand exclusion controls work off brand lists rather than plain keywords, and they're stricter than a negative keyword: they cover misspellings and variants of the brand you nominate. Useful for your own name. Also useful for the competitor whose name your product titles keep matching against.
The negatives we add to almost every Indian account
Not a template to paste in blind, but this is where we start looking:
- Marketplace names. Amazon, Flipkart, Myntra, Nykaa, Meesho, Ajio. Someone typing your product name plus "amazon" has already chosen where they're buying. Let them.
- Rate and price-check queries. "Gold rate today", "price list", "wholesale rate", "MRP". High volume, near-zero purchase intent on a D2C site.
- Job seekers. Careers, vacancy, internship, recruitment, salary. Free ad spend for your HR team, otherwise useless.
- Second-hand and repair. Used, second hand, OLX, refurbished, repair, service centre.
- Informational modifiers where you don't sell into them: how to, recipe, meaning, in hindi, PDF, download, app.
- Free. Obvious, still missing from half the accounts we audit.
Two cautions. "Near me" converts perfectly well for brands with retail stores or fast metro delivery, so don't reflexively kill it. And be careful with category words that overlap between your catalogue and something cheaper — a home fragrance brand blocking "agarbatti" may be blocking half its intent.
Campaign negatives versus account-level lists
PMax accepts negative keywords at campaign level, capped at 100. That runs out fast. The way past it is account-level negative keyword lists, built in the shared library and applied across the account, which PMax respects. Keep the campaign-level 100 for exclusions specific to that campaign's asset groups, and push the evergreen junk — jobs, marketplaces, free, second-hand — into the account list where it protects your Search and Shopping campaigns too.
One thing we get wrong on the first pass about a third of the time: adding a broad negative that quietly strangles a converting long-tail. Add negatives in small batches, note the date in a shared sheet, and check the campaign's conversion volume a week later. If it dipped, you know which batch to look at.
Search themes steer, they don't fence
Search themes are an input. You're telling the algorithm what you think you should be showing up for. They add reach; they remove nothing. Plenty of founders add ten themes expecting the campaign to stay inside them, then find the performance max search terms report full of queries nowhere near the themes. That's the system working as designed. Exclusion only happens through negatives and brand lists.
Your product titles filter more traffic than your negatives do
PMax matches heavily off the feed. A Shopify product titled "Aria Kurta" gives Google nothing; the same product titled "Aria Cotton Straight Kurta – Indigo, Women's" gives it colour, fabric, cut and gender to match against. Fix the titles and a chunk of the bad queries stop appearing without a single negative keyword.
Do this in the feed rules or through a feed app rather than in the Shopify product title, so your storefront copy stays human. Add product type, GTIN where you have it, and Google product category. On stores with large catalogues we've seen title work shift more spend from irrelevant to relevant queries than three months of negative keyword pruning.
COD and RTO make the ROAS you're optimising against wrong
This one is specific to India and it undoes everything above if you ignore it. If a meaningful share of your orders are cash on delivery and a quarter to a third of those come back as RTO, the conversion value Google receives is gross order value, not money in the bank. The algorithm then finds more of whatever produced those orders. Tier-3 pin codes, price-led queries, first-time buyers with no payment friction.
Two fixes, both worth the effort. Use conversion value rules to discount COD orders by your historical RTO rate, or upload adjusted values once delivery is confirmed. And where your margins allow, push prepaid with a small discount at checkout through Razorpay so the mix shifts on its own. Once the campaign is optimising against delivered revenue, the search terms report starts telling you a different story about which query categories are actually profitable.
Where the money leaks after the click
A relevant query landing on a product page that takes 4 seconds to render is still wasted spend, and Google's landing page signals will quietly raise what you pay for the next click. Before you spend a weekend on negatives, check what your top PMax landing pages score on mobile. Our Shopify speed work usually starts with the same three culprits: unoptimised hero images, a review app loading on every template, and three tracking scripts doing the job of one. If you'd rather see the numbers yourself first, SwiftStore will scan the store and track the score as you fix things.
A twenty-minute weekly routine
- Open the campaign, Insights, search terms. Sort by clicks descending.
- Mark every category with 30+ clicks and no conversions in the last 30 days.
- Multiply those clicks by campaign average CPC. Anything over about ₹5,000 a month goes on the negatives list.
- Decide campaign-level or account-level for each. Date the change.
- Check last week's batch: did conversions hold?
Weekly for the first month while the obvious waste is still there, then every fortnight. During the festive run-up, weekly again, because the query mix shifts as gifting intent floods in and categories that were dead in July suddenly convert.
Brand defence deserves a mention here too. If a competitor is bidding on your name, the cheapest counter is usually not a bigger ad budget but owning the branded SERP properly with content and structured data so the paid click is defending ground you already hold.
If you want a second pair of eyes on where your PMax spend is going, send us read access and the last 90 days of Shopify order data and we'll do a free audit of the query mix against delivered revenue. Bring the RTO numbers. They change the answer.