$449,769 in Cosmetics Store Sales From Google Ads at $0.17 per Click

An online cosmetics retailer with roughly four thousand SKUs: face and body care, colour cosmetics, fragrance and haircare. Average order value just under $84, in a category where the auction is crowded with large chains, marketplaces and dozens of comparable independent stores.

Advertising was already running in the account. The problem was not that it did not exist, but how it was built.

What we found in the account

• One campaign for the entire catalogue. Four thousand products, one ad group, one budget. The system had no way of understanding that a $60 serum and a $4 sample carry completely different economics.
• Broad match with no negatives. The search terms report contained "how to make a face mask at home", "DIY cosmetics", "cosmetics shop assistant jobs". The store was paying for all of it.
• No Shopping campaigns at all. The product feed was never connected, even though Shopping is the most natural format for cosmetics: people search for a specific product and want to see the photo and the price immediately.
• No remarketing. People who added items to the cart and did not complete checkout simply disappeared.
• Conversion tracking pointed at the wrong action. The recorded goal was reaching the cart page, not a paid order. The algorithm was optimising towards something that generated no revenue.

The consequences: clicks cost around 42 cents, conversion to order sat at 0.35%, and return on ad spend never exceeded 1.4 - which, once cost of goods was accounted for, meant the channel was effectively breaking even. When we first met, the owner was seriously considering switching Google Ads off entirely and moving the budget into social.

So the first job was not to optimise the campaign but to take it apart into pieces, each with its own economics and its own rules.

Everything was built around a single principle: the account has to be flexible. That means every type of demand lives in its own campaign with its own budget, its own return target and its own negative keyword list. Budget can then be moved surgically instead of shifting one bid for the entire store.

1. Search campaigns: from brand down to a specific shade

Instead of one campaign we built separate campaigns for every major manufacturer brand in stock, with ad groups inside them for individual product lines and items. The logic is simple: someone searching for a specific serum from a specific brand and someone searching "face cream" are at different stages and carry different value.

• Manufacturer brand campaigns. The highest conversion rate in the account. We deliberately held higher bids here because every one of those clicks paid for itself.
• Product-level ad groups. Queries with SKU, volume and shade. The ad repeated the query word for word, which kept quality scores high and pulled the cost per click down.
• Own brand defence. A cheap standalone campaign on the store name. Competitors bid on it regularly, and without this campaign a share of already-decided buyers went to them instead.
• Category queries. The broadest and riskiest tier. It received less budget but carried the most aggressive negative list.

2. Shopping campaigns and feed work

Shopping is the primary format for a cosmetics retailer, and the biggest gain came not from the campaign itself but from the feed. We rewrote product titles to a single formula, "brand + product type + volume + shade", because that is how people phrase their searches. We then labelled products by margin and stock level so budget went first to the items where the store actually earns.

3. Negative keywords: the cheapest optimisation in the project

For the first two months the search terms report was reviewed daily, then weekly. Over the full engagement the negative list grew past three thousand entries. The main groups we filtered out:

• Do-it-yourself intent: "homemade", "recipe", "how to make".
• Job seekers: "vacancy", "consultant", "beauty store jobs".
• Informational queries with no purchase intent: "difference between", "ingredients", "dermatologist review".
• Products the store did not carry but whose names were close to ones it did.
• Wholesale queries, since the store was retail only.

4. Competitor campaigns

A separate block bidding on rival store names. This traffic is always more expensive and converts worse, so it ran on a capped budget with its own ad copy built around concrete advantages: in stock now, next-day delivery, prices below the category average. We used competitor names only as keywords, never in ad text, as policy requires.

This block was never the most profitable part of the account, but it consistently brought in people who were already ready to buy and were simply choosing where.

5. Behaviour-based remarketing

Instead of a single "all site visitors" audience we split it into segments, each with its own message:

• Abandoned cart. The cheapest conversions in the account, at three times the project's average return.
• Viewed a specific product. Dynamic remarketing served the exact item the person had been looking at.
• Previous buyers. Segmented by consumption cycle: a face cream runs out in roughly two months, shampoo in about six weeks. Ads reached people exactly when their product was due to run out.
• Blog readers. The coldest audience, served collection-style ads rather than a single product.

6. Bidding strategy in three stages

Automated bidding was deliberately not switched on at the start. Clean data had to come first.

1. Month one, manual bidding. While the account is still full of junk queries, an automated strategy simply spends the budget on them faster.
2. Months two to four, maximise clicks with a cost cap. Accumulating statistics on the cleaned-up query set.
3. From there, target ROAS per block. Brand campaigns and remarketing got one target; category queries and competitor terms got a far stricter one.

That last step was only possible because of the structure. Setting different return targets for different types of demand inside one shared campaign is not possible at all.

7. Conversion rate work on the site

Half the result sat outside the ad account. Conversion rose from 0.35% to 1.03%, driven by:

• Price and stock status made visible without opening the product page.
• Delivery terms moved onto the product page instead of hiding in a separate section.
• A sticky buy button on mobile scroll.
• Checkout reduced from four steps to two.
• Customer reviews added to the highest-traffic product pages.
• Conversion tracking corrected so the goal became a paid order, not a cart visit.

What did not work

Honestly, the failed hypotheses, because a real project always has them:

• Display to cold audiences. Cheap clicks, almost no orders. Switched off after three weeks and never revisited.
• A fully automated Shopping campaign across the whole catalogue. Early on it consumed budget on cheap, low-margin lines. It only started working once the feed was labelled.
• Bidding on very broad terms like "cosmetics". High volume, almost no purchase intent. Kept only with heavy qualifiers.

The result across the full period

The number that matters most here is not the revenue but the share: by the end of the period Google Ads was producing 62% of the store's total revenue. The channel the owner had almost switched off became the primary source of sales.