Print

How Web Analytics helps to identify eCommerce website breakdown and what actions should be taken by the digital marketing team (practical case)

Artem Akulov

We will share a case from our practice – how we discovered a breakdown in the payment system of an online store, what reports were used and what actions were taken. The article will be useful not only for digital advertisers but also for owners or managers of e-commerce projects. In the end, there is a useful feature and advice on Google Smart Shopping campaigns, which are rarely written about.

Briefly about the project: our client is from Russia, he has an online store on the Shopify platform for the USA market. Our agency works on sales through digital marketing within the specified budgets and target profitability. The main revenue is generated from PPC, on average 75-80%.

How the problem was discovered

It’s easy to detect if something is wrong with the payment system when all payments fall off at once. The revenue graph immediately drops to zero. You can’t miss this.

But it’s different when only some of the payments get canceled. As it happened to us – only payments by credit cards were not functioning, but everything worked with online payment systems (in our case, PayPal). The graph declined smoothly.

Shopify Analytics Report: At First Glance, Revenue Decline Looks OK

The breakdown occurred on March 8th, in the middle of the day. First two days of the recession were not taken seriously. That wasn’t that level where you have to take it seriously. The PPC worked properly. It looked as if the chart would any day hit the rising wave again. But from that moment on, we kept a closer eye on the report.

On the third day, we saw a continuing recession, and we started to look for the reason.

The first thing we checked was the availability of best-selling products in stocks as the lack of these items caused a prolonged recession previously. Sometimes best-selling items can be sold out faster than the new deliveries are planned (the client transports goods to the USA from Russia, where the production is located under its own brand, transportation takes up to 3 months).

The best-sellers were checked in the main Google Shopping Campaign, which generates most of the sales. Everything was in stock except one item —which is not a clear sales leader. Rather, all revenue was distributed evenly across several products. In addition, sales revenue generated by this product did not match the volume of the downturn. Conclusion: the sold-out product did not have an impact on the main recession.

To check the availability status of a product in the Google Shopping campaign, you have to go to the Product section and sort the products by Conversion Value, the availability status will be in the Product Status column

As a next step, we’ve checked if there were any distortions in the traffic by the channels. All channels worked without anomalies. It turned out that at the top level, user behavior did not change, interest in our products has remained the same. The next logical step was to go down to the deeper levels of the funnel. We checked the conversions from stage to stage: Visiting the product card → Adding a product to the cart → Checkout → Transaction. And here deviations were found.

Custom dashboard in Google Data Studio: transactions went down, while checkouts remained the same

Checkouts and transactions tend to correlate among themselves due to a constant conversion rate. The correlation was interrupted on 03/08 – a clear sign of a problem.

Such a problem can be seen only in reports where data is presented in dynamics. Tables or other dashboards with static numbers won’t help. Combined charts work best – you can check all stages of the funnel at once.

Engineers of the agency are setting up such reports in Google Data Studio for each project. This allows us to receive basic information on the KPI of the project online and track other important indicators.

How the problem in the payment system was identified

We already knew that the reason for the recession was a breakdown of the payment system. Suppose we didn’t, then what should we do after looking at the recession graph? After all, there may be several reasons: sending form stopped working, payment page won’t open, some features of the website have fallen off, conversion tracking tag has been removed accidentally, the price is not shown at the checkout stage, etc. All this might happen. There is only one way- each hypothesis must be followed and tested, starting with the most expected one in your opinion.

Shopify platform provides a great report on abandoned checkouts for e-commerce stores. That’s what we checked at first.

Shopify report on abandoned carts with completed checkout

In this report, you may want to dig deeper into each abandoned order and see all the data of each one of them. The valuable information we’ve found — the history of payment processing errors.

One of the users tried to make a payment 4 times, but he did not succeed.

We found many refusals like that. There were visitors who tried to pay 7 times from 3 different cards, waited for some time, and then tried again. These were clearly targeted and loyal customers. The very first error message just happened on March, 8th, i.e. matches entirely with the beginning of revenue recession. The problem was found.

The error was reported to the client. It turned out that he received email notifications from QuickBooks (software responsible for accounting). Support service wrote about problems with the QuickBooks Merchant Services Account. But the request was not promptly taken into work and was skipped afterwards. The problem could have gone unnoticed for a while.

The error in all payments was as follows:

[{“code”=>”PMT-3000”, “type”=>”account_error”, “message”=>”The merchant account could not be validated.”, “infoLink”=>”https://developer.intuit.com/v2/docs?redirectID=PayErrors”}]

On March, 11th the problem was fixed with the help of QuickBooks technical support, payments worked. But that’s not all.

What had to be done after troubleshooting

After everything was fixed, it would be logical to relax and calm down. But two resulting problems appeared, which could be passed by due to inattention or ignorance. First, we’ve got a lot of lost customers. Second, ad campaigns using smart strategies received incorrect conversion data within three days. These problems had to be fixed as well.

Email newsletters

Emails were sent for all those from abandoned checkouts immediately on the day the payment error was fixed. With the promise that now everything is working and purchase can be completed, and to smooth the situation – a discount has been offered. We managed to return 13% of customers.

Adjustments of Google Ads campaigns

The main revenue of the store is generated by Smart Shopping Campaigns,  bidding strategy — “Maximize conversion value”. Smart Shopping campaign sets bids on auctions, chooses places, devices, time, audience, etc. It does it in a way that increases conversion rates and value for a given budget every time.

The indicator that everything is going well is the conversion rate (transaction rate in our case) and targeted profitability. If you have a sufficient number of conversions and required profitability, you can continue to work with the founded audience, if you have no conversions – you have to cut off the audience.

Within three days, Smart Shopping campaign connected with the right audience who wished to make a purchase but could not because of technical errors. The machine learning algorithm recognizes this audience as an incorrect one and excludes it. Hence, even after fixing the problem, we could not get back to the previous conversion rate.

But this could be fixed. In Google Ads, you can exclude data on which algorithms were trained. In practice, it is not a sufficiently known tool, and it wasn’t written much about it though the benefits from using this option are significant. 

Data exclusions in Google Ads

Data exclusion interface in Google Ads

To exclude data, go to the Shared library → Bid strategies → Advanced Controls → Data exclusions → Click on the “+” button, then choose settings you need and save. Details can be found here.

After being excluded, the data does not get erased and can always be found in reports. But the algorithm stops learning from it.

Data exclusion is useful in case of any technical failures of a website or the tracking tag – when data is distorted or not being transmitted at all. But you should not get carried away with it, otherwise, it will backfire.

Today, optimization of advertising campaigns comes down to improving the quality of data on which machine algorithms are trained. Managing this data could help a business to succeed. 

We have shared one of the data management tools with you. We hope this article will help you find necessary reports and improve advertising campaigns. Also, you can always request the setup and support of your advertising in the agency.

If you have any questions, write them below in the comments to the article or use the special form to send a question to our website in case it contains confidential information.