HOW TO INCREASE SALES IN US ECOMMERCE USING GOOGLE ADS AND MICROSOFT ADS

Working with large budgets is always a challenge—quick results and constant control. See in our case study how the right strategy, even though it cannot avoid difficulties, ultimately always wins.

Client

A large American online store specializing in the sale of power tools and equipment for woodworking and metalworking reached out to us. The company has been successfully operating in the US market for over 30 years, offering a wide range of products from leading global brands. Its main philosophy is not just to sell products but to provide solutions that best meet the needs of the customer, regardless of their budget.

The client has its own warehouse and a physical store in Missouri and actively promotes its products online through Google Ads, Microsoft Ads, SEO, Amazon, and partner sites.

The client approached us at the end of June 2024 with a request to improve sales results in Google Ads, as the company had experienced a significant drop in revenue and profitability over the past two months.

Marketing Audit

Previously, the client worked with an agency, but in recent months, changes to the account were sometimes made only by their in-house marketer. Therefore, we first conducted a detailed, free marketing audit of the Google Ads account.

The client had 7 campaigns set up:

• 4 Performance Max campaigns, segmented by margin, each promoting multiple brands.
• A Performance Max campaign focused on the priority brand Festool.
• A Performance Max campaign for the local GEO.
• A Demand Gen campaign.

We identified several significant errors in these campaigns and in the account overall that hindered its performance. The main issues were:

• Automatic application of some recommendations was enabled.
• Budget redistribution was not carried out.
• Inappropriate bidding strategies were set.
• Performance analysis was not conducted at the level of individual products/categories/brands.
• Problems in asset groups: poorly written texts, incorrect final URLs, and low-quality images.
• For some brands, the display of products from the Merchant Center feed was disabled, causing Google to generate only text, display, and video ads (without shopping ads) using the available assets in asset groups. This led to budget wastage on non-targeted platforms such as YouTube, GDN, and mobile apps.
• Important supporting campaign types were missing: search, shopping, and display remarketing.

Additionally, just before the start of the cooperation, the client’s in-house marketer caused significant damage to the advertising account through the following actions:

• Sharp budget increases.
• Disabling effective asset groups.
• Unthoughtful changes to bidding strategies.

This significantly worsened the results of the advertising campaigns and complicated our work. Therefore, the primary task was at least to return to the level of sales and profitability that existed in June 2024.

To achieve this goal in the first month, we carried out the following optimizations:

• Reallocated budgets based on campaign performance.
• Applied the Target ROAS (Target Return on Ad Spend) strategy in the Performance Max campaigns.
• Disabled unnecessary automatic recommendations to ensure full control over the account.
• Reviewed and improved texts, images, and URLs in asset groups and ad extensions.
• Verified whether the Merchant Center feed was connected for each brand.
• Disabled ads in poorly converting states.
• Created a standard shopping campaign.
• Created a search campaign for the top priority brand Festool.
• Set up display remarketing.
• Provided recommendations to the client on site changes to pass moderation for several hundred products in the Merchant Center for further promotion.
• Disabled product groups with the worst ROAS.
• Launched a new Performance Max campaign for the brand Triton, which had the highest ROAS.
• Optimized the local Performance Max and Demand Gen campaigns.

Analytics

We also faced issues with proper conversion tracking. There were two purchase conversions through the cart set up in the advertising account. The first one transmitted data directly to Google Ads through GTM. The second one first transmitted data to Google Analytics 4 through GTM, and only then were these data imported into the advertising account.

We compared the number of these conversions and found that we get 57% more data for the first conversion than for the similar one. But why is there such a difference?

The point is that Google Analytics 4, as an additional link in the conversion tracking process, loses some data during transmission to the advertising account. Additionally, by sending conversions directly to the advertising account from GTM, some privacy restrictions in Google Analytics 4 can be bypassed. Therefore, we left only direct purchases as the main conversions. The more conversion data we have, the better Google’s machine learning algorithms work, and the more effectively bids are optimized.

We also set up tracking for phone number clicks on the website and defined this conversion as the main one in the advertising account, as 40% of the client’s sales came from phone calls. Dynamic call tracking was not suitable in this case because the client had already developed a strong brand over 30 years, and a significant portion of buyers were repeat customers, so it was important not to confuse them with constantly changing numbers.

The client set clear KPIs for us to ensure business profitability, according to which marketing expenses should not exceed 7% of the company’s total monthly turnover. A budget of $75,000 was allocated to achieve these goals, with a target to secure at least $1,000,000 in sales.

Considering all the nuances and inaccuracies in the conversion tracking process, we calculated that to reach the break-even point (Break Even ROAS), the minimum ROAS in the advertising account had to be 4.17. This was the benchmark we relied on and actively worked on improving, which we eventually succeeded in achieving.

The client also provided us with a file of actual sales and profitability every week for additional comparison and monitoring to assess how well things were going and whether the set goals were being met.

Account Optimization

Within the first three months, we achieved the desired ROAS level while maintaining the same budget. The client noticed a significant increase in sales volume and, accordingly, revenue—without additional expenses. Subsequently, both sales and ROAS began to grow steadily each month. This became possible thanks to a series of effective optimization measures, which will be discussed further.

Optimization of Performance Max Campaigns

Since the most costly and lead-generating campaigns in the account were Performance Max, we focused the most attention and time of our marketers on them. We regularly analyzed their performance and implemented the following optimization measures to further improve the results:

—Reallocated budgets and adjusted target ROAS values in bidding strategies;

—Disabled specific products/categories/brands that spent a lot but did not generate sales or had very low ROAS;

—For subcategories that spent a lot and delivered mediocre results, we created new, improved asset groups with more relevant, personalized texts, images, and URLs, and more accurate search queries in audience signals;

—Tested different audience signals, such as buyer audiences, to improve Google’s algorithm performance in targeting and better understanding their interests, habits, behavior, and the websites they visit;

—Created separate Performance Max campaigns for the most successful brands/categories;

—Improved texts and changed images in asset groups;

—Tested campaigns with and without the system’s auto-replacement of ad texts and final URLs;

—Tested running ads in the “only shopping” format, with empty asset groups.

Adjusting Budgets and Target ROAS

A key area of Performance Max campaign optimization was reallocating budgets and adjusting target ROAS in bidding strategies. Every two weeks, we conducted a detailed analysis of the campaigns, checked their performance, and made the necessary adjustments.

The main goal was to find the optimal target ROAS for each Performance Max campaign, where sufficient traffic volume and acceptable actual ROAS were achieved. A target ROAS that was too low led to broader reach by showing ads to a less converting and wider audience. Conversely, a target ROAS that was too high significantly narrowed the audience and reach, which, in turn, led to a decrease in sales volume.

Thus, we gradually determined the optimal target ROAS values—the “golden middle” for each campaign—and currently, we adjust them less frequently.

Reallocating budgets also played a significant role in optimization. Periodically, we adjusted budgets, increasing them for effective campaigns and reducing them for ineffective ones, focusing on actual ROAS, average check, and the share of impressions they generated.

Thanks to these measures, the ads operate with maximum efficiency, and sales continue to grow steadily.

Testing and Other Supporting Campaigns

We tested other campaign options to improve overall performance:

—display campaigns to drive sales of priority products;

—display remarketing campaigns to increase awareness of current promotions;

—a separate branded search campaign;

—separate Performance Max campaigns for the most effective brands (however, these campaigns performed worse compared to those where brands were combined in one campaign based on margin);

—a Performance Max campaign for Black Friday—a temporary campaign in which we included products with the largest discounts, allowing for a significant increase in sales during the most profitable week of the year for e-commerce;

—a separate standard shopping campaign (for backup and better Performance Max campaign performance).

Microsoft Ads

Separately, it’s worth mentioning the launch of ads in Microsoft Ads.

In September 2024, we suggested to the client to test advertising in Microsoft Ads as well. Based on experience, this can work well for e-commerce due to lower competition, a lower cost per click, and better audience loyalty compared to Google.

We set up the Merchant Center in Microsoft, passed the moderation of all products, correctly imported campaigns from Google, checked and adjusted their settings (since in Microsoft they differ slightly from Google, and some important settings are very well hidden).

Next, we adjusted budgets and bidding strategies, set up conversions (to collect data and train the algorithm faster, we set two main conversions: “start of order” and “successful order”).

The initial budget was minimal for testing—$7,000. We launched the ads, waited for 3-4 weeks for the algorithms to learn, and the results stunned us. In the first month, the ROAS was 800-900%. We made a few adjustments to the campaigns and continued to monitor their performance.

The first result was not a coincidence: the campaigns continued to deliver stable sales with a return on ad spend of 800-900% each month. The client agreed to increase the budget to $13,000. This led to further sales growth, although the ROAS decreased to 700%, which is expected when increasing budgets.

However, advertising in Microsoft Ads remains more profitable compared to Google, but unfortunately, further scaling is limited by the significantly lower traffic volume in Microsoft.

Difficulties Encountered

Decline After Setting Up a Branded Search Campaign

We always set up a branded search campaign first for all our new clients because it allows us to fully capture and not lose hot customers who already know about the brand, effectively manage the budget, maximize impressions, and appear higher in search ads than competitors who are trying to capitalize on our branded keywords.

We launched a new search campaign targeting all the client’s branded keywords. As expected, this campaign started delivering excellent sales and ROAS results immediately. However, after a few days, we noticed a decline in sales for some Performance Max campaigns.

After thoroughly analyzing their settings, signals, feeds, placements where ads were displayed, etc., we did not find any obvious issues. However, the cause of the decline in performance turned out to be much deeper: the search categories report, which only started showing accurate data during that period, revealed that these Performance Max campaigns had previously been receiving most of their conversions specifically from branded keywords. Their algorithms had already been optimized for this and similar traffic, and the launch of a separate search campaign disrupted everything. The campaigns began to conflict with each other.

As a result, the system started directing branded traffic mainly to the search campaign, while Performance Max campaigns had to retrain and refocus on new audiences and less conversion-effective keywords, negatively affecting overall performance.

We could have waited for the algorithms to retrain and stabilize the situation, but this could have taken a long time, and the client needed to show stable, growing results without declines.

Therefore, we turned off the branded search campaign—and the results returned to previous levels. This was one of the rare cases where the branded campaign proved unnecessary, and the negative impact of its launch significantly outweighed potential benefits.

Bot Activity on the Website

As mentioned earlier, the account had two main conversions that the campaigns were optimized for: purchases through the cart and clicks on the phone number.

At the end of October 2024, during the weekly analysis of product and brand performance, we noticed that for some of them, the number of clicks on the phone number suddenly increased significantly—to the extent that, for example, 90% of users from the total traffic were clicking on the phone number.

Obviously, these are unrealistic indicators, likely caused by bots, competitors, or click-fraud services on one of the low-quality platforms where Performance Max campaigns mistakenly believed our potential clients were located.

The client confirmed that there was no actual increase in the number of calls.

Google Ads recorded an increase in the number of conversions from phone number clicks and automatically adjusted its display priorities based on the logic: the more conversions—the higher the priority for promotion. However, the algorithm does not consider the quality of conversions and treats all primary goals in the account as equal.

As a result, Performance Max campaigns began reallocating budgets to products that received dozens of bot clicks on phone numbers but did not generate real purchases through the shopping cart.

This disrupted the proper functioning of the machine learning algorithms and shifted the focus of the ads to products that had no real commercial value.

To correct the situation, we excluded phone number clicks from the primary conversions in the account, leaving only shopping cart purchases—this order path is longer and less vulnerable to bots and accidental clicks. Additionally, we reallocated budgets between campaigns.

Within two weeks, all campaigns stabilized: sales returned to the previous level, and ROAS increased.

Fake Orders for $100 Million

At the end of February 2025, we faced a problem when the client’s technical team updated the CMS, and the internal marketer, not considering the specifics of working with Google Ads, clicked on an ad and started testing the website after the update. During the testing process, he made several orders for hundreds of products with a total amount of almost $100 million.

This negatively affected the ad performance. Due to the fake orders, the Google Ads system received incorrect data, which caused the algorithms to retrain. As a result, there was a disruption in campaign optimization, leading to an increase in cost per click, a decrease in the number of clicks, and a drop in the number of orders and ROAS.

How We Solved This Issue:

• First, we sent a message to the client with detailed technical recommendations on how to work with the website to ensure that their team’s actions wouldn’t affect the ads and to prevent similar issues in the future.
• In the global settings of smart bidding strategies, we excluded all data collected by the system on the day and during the hours when false conversions occurred, so that the algorithms would no longer focus on them.
• We temporarily increased the budgets for the campaigns that were least affected.
• We increased the budget for the regular shopping campaign and created separate Performance Max campaigns in Shopping-only format for the brands whose sales were the most important for the client at that time. This was done to control ad placements and prevent Google from displaying ads on ineffective placements (YouTube, GDN, mobile apps, etc.) during the retraining phase.
• We adjusted the bidding strategies.

Thanks to our quick and well-coordinated changes in the account, by the third week after the system error, we were not only able to return to the previous performance levels but also exceeded them. The client was very pleased with our response and the quick recovery of sales.

Results

• The monthly marketing budget increased by 17.33% (from $75,000 to $88,000).
• The total revenue from cart purchases in Google Analytics grew by 69.85% (from $471,000 to $800,000).
• ROAS in Google Ads increased by 301% (from 327% to 628%).

Next Steps: optimize and scale the advertising accounts in Google and Microsoft, aiming to smoothly reach a total monthly revenue of $2,000,000.

Conclusion

Search Ads in eCommerce in the US market requires constant monitoring and regular optimization, especially when working with large budgets. It is important to respond quickly to changes, carefully analyze campaign performance, and adjust strategies in a timely manner. Every decision should be data-driven and well-justified, as even a minor mistake can significantly impact the outcome. Only well-planned and precise optimization will ensure stable results and the most effective use of the budget.

Order a Personal Consultation for Your Business