How to Cope With Google Smart Shopping Campaign And Take It to Stable Results

Google Smart Shopping Campaign is a real deal for a PPC specialist. This campaign can produce brilliant results with minimum actions from an expert, it is as automated as possible. But it’s not that simple. We had to go through a tortuous path of testing, adjusting strategies, and exploring all kinds of ways to optimize the campaign to achieve its target goals. We will share our experience in this article, as well as the tools that helped us achieve the result.

Read our article on Google Shopping ads for all the latest information on this type of ad campaign.

It works just like a regular Shopping campaign. It shows products from the feed that you upload to the Merchant Center. What makes the campaign Smart is Google machine learning. Bids and ad placements are automatically selected to attract the most valuable conversions within a provided budget.

Smart Shopping ads appear on Search, Display Network, YouTube, and Gmail. So it covers more placements compared to regular Shopping campaigns.

Here’s an example of a Shopping ad in organic search:

An example of an expanded Shopping ad in Gmail:

And here’s one of the options for displaying ads on YouTube:

Machine learning in a campaign limits its management options.

Here’s a list of what you can’t do in a Smart Shopping Campaign:

• Work with search queries and exclude negative keywords;
• Exclude locations, although the location targeting can be set in the settings. Let’s say you have a feed in the United States, and you want to show ads in all states except California. In this case, you can specify all states manually in the campaign settings, excluding California. So basically you cannot exclude California, but you set all the states except this one;
• Adjust bids or exclude ad delivery on certain devices;
• Set up audience targeting;
• Manage ad scheduling.

Here is what you can manage in a Smart Shopping campaign:

• Budget (but the total budget is not supported);
• Select a New customer acquisition (NCA) conversion goal. This feature can be found in the Conversion goals section of the campaign settings. With this goal on, you can set your own customer acquisition cost;
• Set and change the value of the target ROAS in the auto strategy settings;
• Product titles and descriptions. They are handled at the feed level. So you can do it using a service or app you use for creating and managing your feed, or in the Merchant Center using feed rules.
• Products. You can filter products by name, ID number, product category or any other attributes that you can set in the labels. There are no restrictions on working with product groups in a Smart Shopping campaign. Just as you can manage the product groups, you can build the account structure according to your tasks. For example, you can create an Alpha campaign for the best-selling products and set a larger budget for it.
• Geographical targeting. As mentioned earlier, you can set specific locations within the country you specified in the feed.

Our client is an online store powered by Shopify. They sell products for fishing, hunting and outdoor activities. The client does not sell items like fishing rods, spinning rods, or shotguns, but other stuff needed for camping — ice screws for winter fishing, multi-tool shovel, hats with Led Lights and so on.

Geo-targeting: the USA.

The client was getting a good number of sales on Amazon and eBay. Google Shopping campaigns produced nice results as well, but the client was not happy with the sales volume.

We have defined 500% ROAS as the target KPI. We divided the process of achieving this goal into several stages. The first stage was to achieve 200% ROAS consistently.

The statistics were already accumulated in the account and showed that different product groups were selling with different efficiency. We decided to put the effective product groups into separate campaigns to “push” them with a budget. First, we identified the most effective groups in separate campaigns. Then, groups with average efficiency.

From time to time, we received requests from the client to promote certain products in the moment. We included those products in separate Smart Shopping campaigns.

Depending on the group’s efficiency, we set different campaign budgets – from $5 to $20.

Results:

The dynamics were unstable, the campaigns accumulated different numbers of conversions and had different ROAS, in some campaigns, it was less than 100%.

While processing the ongoing work, we found out that the beach umbrellas were the best-sellers. To enhance this effect, we have moved all of the beach umbrellas into a separate campaign. We set a daily campaign budget of $50. The target ROAS was not set.

Results:

The umbrellas sold well in a separate campaign until the end of the season.

We went further and transformed the umbrella campaign into an Alpha campaign. We added other high-demand products to it. Set the budget to $60, and 300% ROAS as a target.

Results:

The campaign did not spend its limit. The sales volume was small and the sales were irregular.

We did not receive positive results with product segmentation, so we put forward several hypotheses:

• The problem is in the feed. The product feed can be generated by one of the many apps for Shopify stores. We used a standard free app with minimum control over the options. We assumed that the problem could be in the application itself;
• The problem is the segmentation itself. A unified Smart campaign was driving more conversions than individual campaigns we created by category or by product performance.

We switched to the Feed For Google Shopping app. It is not expensive, but quite functional. In the new application, it became easier to:

• set a product category in Google,
• edit Titles and Product Descriptions right in the app (no more feed conversion rules!). Titles and descriptions can be edited directly in the application in the product card. You can always roll back to the original value from Shopify with one click.

Using the new feed, we set up a campaign with all products. All other campaigns were disabled. We set a budget of $80 and a target ROAS of 200% for the new campaign.

At the same time, we changed the attribution model in conversion from Last Click to Position based. Now, if several ad campaigns were involved in a chain of interactions, the conversions were distributed among these campaigns (in addition to the Smart Shopping campaign, we also used other types of campaigns, we’ll talk about it later).

Check out our article to learn more about Google attribution models.

Results:

3 conversions in 1.5 weeks. 258% ROAS.

On the one hand, we achieved our target ROAS, but on the other hand, we were unhappy with the number of conversions. With this volume of conversion data, further growth and scaling were impossible.

Therefore, once we got the results, we decided to remove the target ROAS in the strategy settings. The vertical line on the chart below marks the day we made the change. After that, the traffic and ad spend increased dramatically. The conversions have become more regular. The number of conversions per day has increased. ROAS dropped to 127%.

The campaign was giving stable conversions, but the results were still insufficient. There was a feeling that we were missing something fundamental.

We assumed that the campaigns based on machine learning may not receive enough data to train and learn successfully. So we came up with an idea to test the larger budget. Based on the recommendations of our personal Google manager, we have calculated the optimal budget: CPOx10. In our case, it turned out to be $330. When implementing an online sales growth strategy, it’s important to choose the correct budgets for your advertising channels.

While we’ve been waiting for the new e-commerce PPC strategy approval, we felt inner resistance. In our field, you normally increase the budget once you achieved the goal and are ready to scale the obtained result. But we had nothing to scale yet.

The other part of the problem was that we were afraid the campaign will spend the entire budget available and we might fail in terms of profit.

To minimize the risks, we decided that we will:

• “not think for the client”. We won’t be thinking if the client will agree on budget increases or not. Instead, we organized a meeting, proposed the strategy, and stated the reason for the increase. We clearly showed that we need this much for a daily budget;
• set 140% target ROAS. Based on our experience, specifying target ROAS has always limited campaign costs;
• agree with the client to test the new settings for 2 weeks. At the same time, it was important to make no changes to the campaign, give it time to get data and learn from it.

The time frame of two weeks was based on the recommendations of our personal  Google manager. She suggested that based on their data, this time should be enough. In this case, in the first week the campaign collects data for training, and starting from the second week, it optimizes itself based on this data. Positive results can be considered as increasing the number of sales per day starting from the second week.

The results were summed up after 2 weeks of the test: we got 28.26 conversions with a 135% ROAS. With a daily budget of $330, the campaign spent $934 in 2 weeks.

Since the first week of the campaign was a week of data collection and training, we did not count it for a final conclusions.

Result for the second week of the test:

11.85 conversions, 192% ROAS. $376.71 ad spend, revenue $721.94. We got positive dynamics in terms of ROAS and sales.

Based on the test results, we decided to keep the settings and give the campaign time to increase the number of conversions per day.

We set the following event dynamic parameters: Page Type, Product ID and Total value at different stages of work. The campaign was running well until we got a warning about the Google Ads tag audience source error in the Google Ads account. We got the same warning message in several eCommerce projects we run at the same time. We solved this problem by renewing the script for dynamic remarketing (borrowed the latest release from digitaldarts.com.au).

That was the last day of the test when we solved the issue with dynamic parameters. This day is shown as the rightmost red vertical line on the chart. Once we fixed the issue, the average daily number of conversions increased, the campaign started spending more and getting more conversions. We’ve got many more conversions in the sequence of interactions after multiple visits from a Smart Shopping campaign. That made us think that the Smart Shopping campaign with the dynamic remarketing finally started working.

At the end of the test, the campaign was spending 18% of its daily limit per day and getting 2-3 conversions per day. We decided to keep the settings and give the campaign a chance to increase the average daily sales.

We made the changes 3 weeks after the end of the test.

Results in 3 weeks:

106 conversions, ROAS was 198%. The campaign was already spending $100 per day. The BFCM sale played an important role here. These days we have a significant increase in daily sales and campaign costs.

Once we received such results on ROAS, we decided to increase the target value in the campaign settings and raised it from 140% to 150%. We gave the campaign a week to work with these settings. During this week the campaign got 153% ROAS, we consistently received at least 10 sales per day, and the daily ad spend has almost reached the limit.

We decided to increase the target ROAS again with a radical step — from 150% to 215%. Our KPI — to consistently receive at least 200% ROAS.

Results:

210% ROAS. We have reached our ROAS target. But the dynamics changed 2 weeks after the last ROAS increase, and the number went down, as well as the number of conversions. This result was influenced by the seasonal change of demand as the Christmas shopping season ended.

After the Christmas holidays, the situation got stable, and we reached 324% ROAS in January. Today, we continue to work with adjustments to the target ROAS value in the Smart Shopping campaign and do our regular optimization.

Throughout the entire period of work with the project, we tested other types of advertising campaigns — Smart Display, Search by keywords, Brand campaign, Dynamic Remarketing in GDN, DSA, YouTube video campaign, Cent shopping campaign and Smart campaign. We had the conversions and interaction sequences in the Brand campaign (4.60 conversions, 3441% ROAS) and a Cent shopping campaign (4.55 conversions, 488% ROAS). The Cent campaign is a regular Google Shopping, and it displays all the same products as the Smart campaign. It runs on minimum bids — a few cents. The goal is to get residual cheap traffic.

The rest of the campaign types either did not bring the results or it’s still too early to make conclusions because the campaigns are still being tested.

We are regularly optimizing the titles and product descriptions in the feed. Here’s what we are testing:

• The first five words should contain the key information about the product;
• Brand name included / not included;
• Include important characteristics in the heading, in our case, these are dimensions, important qualities of the product;

We constantly monitor the status and errors of the products in the Merchant Center:

• Mistakenly rejected products;
• Images with overlays (captions) that are automatically pulled from the Shopify store. We turned on the auto image change in the Merchant Center, but it needs to be checked on from time to time;
• Products for dynamic remarketing. Sometimes the products are mistakenly banned, for example, assigned 18+, they cannot be advertised in remarketing for ethical reasons. We fix this by sending the rejected products for re-moderation.

Working with the Smart Shopping Campaign, we went from one unified campaign and back to it, by segmenting products by category and performance. The key decision was to give the Smart campaign time to accumulate data on a larger budget. Before the test, the campaigns did not spend even small limits, so it seemed that the increased budget would not help. We also had concerns that a large budget means high risk. It was also mentally difficult to keep the campaign working for 2 weeks without any changes. Silently monitoring the campaign with a large budget feels too luxurious when the time is limited.

Now we know that a Smart Shopping campaign powered by machine learning can successfully learn and start performing well on a large budget when it can get at least 10 conversions per day (budget = CPOx10). It is important to specify the target ROAS in the auto strategy settings. Don’t overestimate it: do not highlight something that is very different from your actual ROAS, especially at the initial stages of work.

We hope this article helps you get the results you want with Smart Shopping campaigns. You can do it yourself or order eCcommerce PPC management services from our team.

We have extensive experience in running PPC for eCommerce businesses. If you have any questions, email us or use the special form to send a question securely and confidentially.