The challenge
The brief had 2 halves. Grow ecommerce sales internationally, across markets where the currency, the shipping terms and the brand’s own recognition all change at the border. And pull more people through the doors of the US boutiques without taking a single sale off the website. Those goals compete for the same searches, and the next step was a structure that could tell them apart.
We saw room to grow in 4 areas.
- Campaigns with different jobs sharing one budget, so the account was competing with itself for the same searches
- Brand searches mixed in with everything else, which hid what it actually cost to win a new customer
- Searches for a specific product type landing on the homepage, when a category page for exactly that already existed
- A shopping feed that had not kept pace with a catalog turning over 4 times a year
Promotions had more to give, too: the biggest revenue weeks of the year ran with the same ads as any other week.
Underneath it all, the sales data behind the bidding was not fully reliable, so $160K to $180K a month was being optimized on incomplete data. The account had simply outgrown its original setup.
The solution
We started with the audit, then took Google Ads and Microsoft Ads live together, both on the same structure. The audit findings set the order of work, and conversion tracking went first.
Pay Per ClickTaking the search side apart and rebuilding it
ServicesGoogle Ads
We worked the audit findings through with the client’s team and fixed the conversion setup before touching a single campaign, so every change after that could be judged on a signal worth trusting.
Then we pulled the structure apart. Match types came out of the shared campaigns into their own, so exact, phrase and broad each carried their own budget and their own bid logic. We separated trademark searches from everything else, and that one split is what makes a $180K month legible: brand demand and acquisition demand stop being a single total and become 2 numbers we can act on.
Searches for a product type were reaching the homepage. We routed those searches to the category page written for that exact thing.
Ad groups built from what already sold
Once the account was producing clean data, we automated the part that never ends. A script generated new ad groups straight out of converted search queries, so every search that had already produced a sale earned a place in the structure. We reviewed each one by hand for the landing page, because a script can find the query but not the page that answers it.
Dynamic Search Ads and display remarketing filled in around that.
The sale calendar
Spring/summer, fall/winter and the evergreen promotions between them each arrived with their own copy across search and display, written and scheduled weeks ahead of the sale going live. We planned the display creative together with the Meta Ads agency, so the same promotion read the same way on both channels.
Keeping the click data intact
After the site migration, the data went quiet, and the cause sat outside the campaigns. A redirect had started stripping URL parameters, and the gclid and wbraid parameters went with them. wbraid was brand new at the time, so very few people were watching for it. We spotted the drop, traced it to the redirect, and the developers fixed it.
Pay Per ClickA catalog that reinvents itself 4 times a year
ServicesFeed Management
Several thousand products, updated continuously, with new lines landing each season and older ones sliding into sale. The feed we inherited was built for a smaller account, and Shopping at this budget needed more from it.
We rebuilt it in DataFeedWatch, with custom rules doing the assignment work.
- Attributes filled in automatically as products arrive
- Categorization held steady across the whole catalog
- Pricing and promotional tagging driven by rules, not by someone remembering
That last one is what makes a limited-time sale workable at this size. The promotional tag goes on and comes off across thousands of products on a schedule, and Shopping follows it.
The same feed carried the international expansion into 3 additional countries, each with its own currency, shipping terms and level of brand recognition, all running off one rule set with the local details layered on top.
Sorting the catalog by what each product earns
In the last year we ran the account, the labels did more than describe the catalog. A labelizer sorted products by how they were actually performing, into bleeders, zombies and heroes, and each group got its own treatment.
- Bleeders, the products spending without returning, came under tighter control
- Zombies, the ones nothing was reaching, got deliberate visibility to clear warehouse stock
- Heroes took the weight of the budget
New collections needed somewhere else to go. A product arriving with almost no history reads like a zombie to any algorithm, and would have been buried in that segment before it had a chance to sell. So every new collection went into a seasonal campaign of its own, held out of the labelizer until it had enough data to be judged on.
Pay Per ClickSending the search to the door as well as the cart
ServicesGoogle Ads
Next to the online store, the brand runs physical boutiques in the US, and the brief was to fill them without costing the website anything.
We launched local campaigns with driving directions set as the conversion, so they optimized toward someone walking in. Geo-targeting held them to the areas that could realistically produce footfall, and the ecommerce campaigns carried on against their own objectives alongside.
They settle at roughly 3,000 interactions a month, and the online channel held its performance throughout.
Pay Per ClickThe same structure, a very different auction
ServicesMicrosoft Ads
Microsoft Ads went live alongside Google, built on the same restructure. The campaigns came across through the automated import, which gets the shape right and the settings generic, so the real work started the moment the import finished.
Bids went manual. We tested automated bidding on Microsoft and it did not hold up against manual max CPC at this account’s volumes, so we set and maintain max CPC bids by hand.
Microsoft will start triggering queries with no relationship to the catalog from one day to the next, so we review the query report regularly and negative out whatever has wandered in.
The results
Weekly calls and weekly reporting kept the account honest. We also supported the client’s team in building a custom reporting dashboard and checked that what it showed matched what the platforms reported, so everyone was working from the same numbers.
With the structure clean, the budget had somewhere to go. Return on ad spend settled at 600%, and impression share on the searches that matter runs above 95%, so demand the brand has already built is not leaking away.
Daily spend runs around $5K on an ordinary day. Through Black Friday it has passed $40K in a single day, and the campaigns take it because each one has a single job and a clean signal to bid against.
600%
ROAS · across the account
>95%
Impression share · on the searches that matter
<$3
Cost per acquisition · on a $160K–$180K monthly budget
3,000
Monthly local interactions · through to store directions