How to Optimize Google Shopping Ads Results
What Do You Mean “Optimize?” I mean these four things: Fixing errors, 150-character titles, optional attributes, and testing data are simple ways to increase shopping ad revenue. No tech magic.…
What Do You Mean “Optimize?”
I mean these four things:
- Fix Errors: missing data, wrong prices, invalid GTINs, etcetera.
- Write 150-Character Titles: about 30 characters are visible when a shopping ad appears. Up to 70 characters are visible on mouse over. Note that 150 characters is the limit. Use them all.
- Add Optional Attributes*: add watts to electronics items, material to apparel items, and MPNs to any product that has a manufacturer-assigned MPN.
- Test Changes: test different keywords and images. I don’t see a need to test anything beyond these two product attributes in most cases.
- A keyword test could be adding keywords that aren’t necessarily part of the product title in your shop, but might be used by Google search users nonetheless. For example, add “anniversary” to jewelry product titles, even if “anniversary” is not part of the product title in your shop. Or, move the OEM part number from the end to the beginning of the title.
- An image test could be between your first and your second product image on your website. These tests nearly always lead to substantial changes in CTR, which is the most immediate and telling indicator of whether or not you’ve selected the best image for shopping ads. These images formats always win: pure white background, apparel items pictured on a human model, product photos (not lifestyle photos).
Fixing errors, 150-character titles, optional attributes, and testing data are simple ways to increase shopping ad revenue. No tech magic. Just ensure that data is clean, filled out completely, and then test titles and images.
*OEM part numbers are an oft-forgotten title addition to Google shopping ads. Search users sometimes only search the OEM part number when using Google, however, so it’s essential to include.
How to Run a Test
Testing CTR will yield immediate results, so start with that as your metric.
Why not test other metrics?
You could test number of impressions (more impressions good, less impressions bad) but the problem with this is that impressions might rise or fall because of factors other than your test variable: increased ad budget, seasonal shopping activity. These would lead to changes in impressions regardless of your test variable.
CTR is a direct indicator of how enticing your shopping ads are, regardless of budget and seasonal changes.
Concurrent or Consecutive A/B Test?
You can run A/B tests concurrently (like DFW’s title A/B test feature which measures conversion rate changes on your website). However, it’s technically to run an A/B test’s impact on CTR because it’s not possible to know exactly when version A or B is live when viewing CTR data.
A consecutive A/B test is preferable for this reason. The control period will be the 30-day period before you change a variable. The test period will be the 30-day period when your variable change is live.
Test All or Some Products?
Test changes to all products at once for fast results at the risk of lost ad revenue. Or, test a selection of products to allow non-tested products to keep generating sales at the status quo pace, uninterrupted by your test.
Summary and Tips
While tests can confirm what you already know (e.g., pure white background images always perform best), they can also help you uncover new ideas (e.g., adding “Valentine” to a product title during Valentine’s Day, or “Mother” during Mother’s Day boosts ad revenue).
Tips
Don’t bother testing different background images on your products. A pure white background always wins. I don’t know why Google offers AI-generated background changes. This will certainly reduce ad revenue because they muddle up the product image, which is already quite small and hard to see.
Make substantial changes, not minor ones. Moving a word in your title from 1st to 2nd word position likely won’t make any noticeable change. For example, “Nike Men’s Running Shoes” vs. “Men’s Nike Running Shoes” is probably a wash, there won’t be a difference. However, testing “Nike Men’s Road and Track Race Running Shoe” vs. “Nike Men’s Running Shoe” will likely yield a difference.
Automate data collection using software like DataFeedWatch or Supermetrics. Otherwise, gathering reports to compare the outcome of your tests will be so laborious that you never get around to it. Everything should be automated. Here’s an example of how I automate test results tracking on my ad accounts.