Cutting Shopping CPA by 62% With Feed Strategy Alone
No new budget, no new creatives — just a rebuilt product feed and smarter bidding tiers. The title rewrite, the margin tiers, and the negative-keyword categories that recur in almost every account.
By Bharathidasan Moorthi
A client came to us convinced they needed a bigger budget and fresh creative to fix a Shopping CPA that had crept up for months. They needed neither. We cut their Shopping cost per acquisition by 62% without adding budget or making a single new ad.
62%
lower Shopping cost per acquisition, with no budget increase and no new creative
One GrowthPartners e-commerce client account, comparing Shopping CPA before and after the feed rebuild and bidding-tier split described in this post.
The entire fix lived in the product feed and the bidding structure. This post is the method, in the order we run it. The examples are illustrative rather than the client's own catalogue, but the moves are the ones that did the work.
The feed is the campaign
In Shopping you do not write ads. The feed is your ad. Titles, product types and attributes decide which searches you are eligible for, what the listing looks like, and how confidently Google matches you.
This account's titles were manufacturer defaults: brand name, model number, and nothing a customer would ever type. Google did what it always does with a vague title, which is match it broadly and let the budget find out. Spend leaked across queries nobody had chosen to bid on.
What a rebuilt title actually looks like
The fix is not clever copywriting. It is putting the words a buyer would search at the front, in the order they would search them.
| Title | What happens | |
|---|---|---|
| Before | ACME WB-750 SS | Matches almost nothing intentionally. Google guesses, and guesses broadly |
| After | Insulated Stainless Steel Water Bottle 750ml — Leak Proof, Keeps Hot or Cold 24 Hours | Matches category, capacity and the two attributes people actually filter on |
The order that works
- 1Category first. What the thing is, in the words a stranger would use — not your internal product line name
- 2The attribute people filter on. Size, capacity, material, colour, compatibility. Whichever one a buyer would narrow by first
- 3The variant that distinguishes it from the other twelve near-identical SKUs in your own catalogue
- 4The differentiator last, if there is room. Titles get truncated, so nothing load-bearing goes at the end
Brand goes first only when people search the brand. For most catalogues they do not, and a brand-led title spends its most valuable characters on a word nobody typed.
Segment by margin and intent, not one big campaign
Every product had been dumped into a single Shopping campaign with one target. That means the high-margin best-seller and the low-margin long-tail item compete for the same budget at the same bid, and the algorithm has no way to know one sale is worth four times the other.
We split the catalogue into tiers and set targets per tier, so the account could spend aggressively where a sale was worth it and stay cautious where it was not.
| Tier | What goes in it | How it is bid |
|---|---|---|
| Proven | Best-sellers with healthy margin and real conversion history | The most aggressive target. This is where budget should concentrate |
| Thin margin | Sells well, but there is little room between price and cost | A conservative target that each sale has to clear on its own |
| Long tail | Low volume, or no conversion history yet | Capped spend. It either graduates on evidence or comes out |
| Excluded | Products priced below the point where any realistic CPA works | Out of the feed entirely, not merely bid down |
It is the same account structure thinking applied to a catalogue rather than a keyword list: separate things that behave differently so you can control them differently.
Add the negatives Shopping cannot set for you
Shopping matches on the feed, not on keywords you chose. Negative keywords are almost the only relevance control you have, and most accounts have barely any.
We mine the search terms report and cut by category rather than one term at a time. These are the buckets that recur across almost every account we take over:
- Informational — 'how to', 'what is', 'DIY', 'review', 'vs'. Someone reading, not buying
- Condition mismatch — 'used', 'second hand', 'refurbished', 'wholesale'. They will not pay full price for new
- Service intent — 'repair', 'service centre', 'spare part', 'replacement'. They own one already and want it fixed
- Audience mismatch — searches for a segment your catalogue does not serve. If you sell only one gender, age group or vehicle type, the other one is quietly eating budget every day
- Marketplace and competitor — rival brand names and 'amazon', 'flipkart' style queries, where you are paying to lose a comparison
- Price shoppers — 'cheap', 'free', 'discount code'. Whether this belongs here depends on your positioning, so decide it deliberately
In this account those categories were absorbing 20-30 per cent of spend. Audience mismatch is the one that surprises people most: a catalogue serving exactly one segment will still collect a steady stream of searches for the segment it does not serve, and nothing in Shopping stops it automatically.
The order we actually run it in
- 1
Read the search terms report before touching anything
It tells you what Google currently thinks you sell. The gap between that and what you actually sell is the whole diagnosis.
- 2
Fix titles on the products that carry the revenue
Not the whole catalogue. The top sellers first, because that is where a matching improvement pays for itself fastest.
- 3
Cut the negative categories
Work in buckets rather than individual terms, and expect to repeat this monthly rather than once.
- 4
Split the catalogue into tiers and set targets per tier
Including the exclusion tier. Products that can never clear a viable CPA should leave the feed.
- 5
Let it re-learn before you judge it
Feed changes need a re-crawl and the bidding needs fresh data. Reading results too early is how a correct fix gets reverted.
Check the feed first
Before you ask for more budget or new creative, check whether your feed is actually telling Google what you sell and who should see it. Most Shopping accounts leak spend on queries they never meant to bid on, and fixing the feed costs nothing but attention.
Budget and creative are the expensive levers. The feed is the free one, and it is almost always the one that is broken.
It is the first thing we look at on any e-commerce account we take over, and the cheapest lever available in Google Ads Shopping.
Common questions
Do I need a developer to change my product titles?
Usually not. Most platforms let you set a separate feed title independent of the title shown on your website, either natively or through the app that pushes products to Merchant Center, so you can optimise for search without changing what shoppers see on the product page. That separation is worth setting up before you start, because the best feed title and the best on-site title are rarely the same sentence.
Does any of this apply to Performance Max?
The feed work does, and arguably matters more. Performance Max draws on the same product feed, so titles, attributes and product types still decide what you are eligible for. What changes is the control surface: you have far less direct say over negatives and structure in Performance Max, which makes getting the feed right the main lever you still hold.
Want this applied to your account?
Skip the reading. Get a strategist to map the exact plays your account needs right now.
Get a free growth audit
