Google Shopping campaigns: feed quality, bids and product groups

Shopping ads have grown faster than text ads for US retailers for three years running. Clicks come with an image, a price and a store name, so the shopper has already filtered themselves before they click. Typical US retail accounts see Shopping cost per click somewhere between $0.30 and $1.00, well below text ad CPCs for the same products, with conversion rates that match or beat text ads. Mobile Shopping volume has also jumped this year as Google added the swipeable carousel and larger images on phones.

The catch is that you do not choose keywords. Google decides which of your products to show for which query based on your product data. That is why a mediocre feed with clever bidding still loses to a clean feed with simple bidding.

Your Merchant Center feed is the ad copy, the targeting and the quality score all at once. Work through this list before touching bids.

1. Titles. Put the words a buyer types at the front: brand, product type, key attribute such as size, color or model, in that rough order. "Nike Air Zoom Pegasus 32 Men's Running Shoe, Size 10, Black" beats "Pegasus 32." You have 150 characters; the first 70 or so show.
2. Unique product identifiers. GTIN (UPC in the US), brand and MPN. Google requires two of the three for most products from known manufacturers and will disapprove items without them. Products with correct GTINs also get matched to more queries.
3. Google product category and product type. Map every item to the most specific Google category, and fill product_type with your own taxonomy, which you will use later for product groups.
4. Price and availability must match the landing page. This is the number one disapproval reason. Turn on automatic item updates in Merchant Center so Google can correct small mismatches from your schema.org markup, and fix the underlying sync anyway.
5. Images. Clean product on white or plain background, no text, no watermarks, at least 800 pixels on the long side for apparel.
6. Shipping and tax. Set at the account level in Merchant Center. Missing shipping settings cause silent disapprovals for whole countries.
7. Submit frequently. Daily if prices or stock move. A feed older than 30 days expires.
8. Custom labels. Five free-form fields, custom_label_0 through 4. Use them for margin band, price band, seasonality, bestseller status and clearance. These are your bidding levers.

Check the Diagnostics tab in Merchant Center weekly. A feed where 15 percent of items are disapproved is a campaign where 15 percent of the catalog has no ads.

A new Shopping campaign starts with a single All Products group and one bid. Leaving it that way is the most common mistake we see. Every product gets the same bid regardless of margin, price or conversion rate.

Subdivide. The attributes available are category, brand, item ID, condition, product type, channel and the five custom labels. A structure that works for most retailers:

Always keep an "Everything else" group at each level with a lower bid so new products still show. Then set bids from the bottom up: top sellers and high-margin groups get the highest CPCs, clearance and low-margin groups get the lowest, and "Everything else" sits in between.

Use the bid simulator and the benchmark CPC and CTR columns that AdWords now shows for Shopping. If your bid sits well under the benchmark for a group that converts, raise it. If search impression share is low on your best groups, you are leaving sales to competitors.

Shopping campaigns reward retailers who know what each product is worth. A decision rule:

• Target cost per acquisition for a product group equals the average order margin for that group multiplied by the share you are willing to spend on acquisition, usually 25 to 40 percent for a first purchase.
• Divide by the group's conversion rate to get a maximum CPC.
• Check the result against benchmark CPC. If you cannot afford the benchmark, that group is not where your Shopping budget belongs; drop the bid, do not chase it.

Revisit monthly. Conversion rates move with seasonality and with how well the landing pages work on mobile, which has become a real factor this year.

Because there are no keywords, priority settings are how you control what you pay for generic versus specific searches. Shopping campaigns can be set to low, medium or high priority, and when the same product is eligible in more than one campaign, the higher priority campaign's bid wins regardless of amount.

The standard three-tier setup:

1. High priority, low bid, broad negatives. Catches generic queries such as "running shoes" cheaply.
2. Medium priority, medium bid, brand terms as negatives. Catches product-type and brand queries.
3. Low priority, high bid, no negatives. Only reached when the higher tiers have excluded the query, so it ends up catching exact product and model searches where you want to win.

Negative keyword lists are what push queries down the tiers. Mine the Search Terms report weekly and move high-intent terms down and junk terms out. Shared negative lists across the three campaigns keep this manageable.

• Search terms report. The only view of what you are paying for. Review weekly.
• Product group performance with margin. Export, join to your margin data, and look at return on ad spend by group, not just conversions.
• Impression share and click share by product group. Shows where you are losing auctions to budget or rank.
• Device segmentation. Mobile share of Shopping clicks is climbing; if mobile converts at half the rate of desktop, adjust mobile bids down rather than cutting mobile entirely.
• Item-level reports. Find products with many clicks and no sales. Usually a landing page or price problem, not a bidding one.

Set budgets to avoid "Limited by budget" status on your profitable campaigns. A Shopping campaign that runs out of money at 2 p.m. is simply handing the evening's buyers to someone else.

• One product group, one bid, for a 5,000-SKU catalog.
• Titles copied from internal product names with no brand or attribute.
• Price mismatches caused by promotions applied on the site but not in the feed.
• No negatives, so a plumbing supplier pays for "free pipe cleaning tutorial."
• Ignoring local inventory ads when you have physical stores; they are a separate feed and a real advantage for retailers with locations.
• Sending Shopping clicks to a category page rather than the exact product.

We start with discovery: a Merchant Center diagnostics review, a feed audit against the specification, and a read of the search terms report to see what the account is actually buying. We then plan a product group and priority structure tied to the client's margin data and a bidding rule the team can apply without us. Build means fixing the feed at the source, usually in the ecommerce platform or a feed tool, setting up the campaigns and negatives, and connecting reporting that shows margin rather than just revenue. Then we improve on a weekly cycle through the season. The work touches ecommerce operations, the website and the data as much as it touches AdWords, which is why we treat it as one connected engagement rather than a media buy.

A representative engagement is the ecommerce growth program we ran for an online retailer with a large catalog and a single-bid Shopping campaign. The first month went almost entirely on the feed: titles rebuilt around how buyers search, identifiers completed, disapprovals cleared, custom labels added for margin bands. Only then did we restructure product groups and introduce priority tiers with shared negatives. Revenue from Shopping rose while cost of sale fell, and the client's team ended up running the weekly routine themselves from a documented playbook.