‘Why CPA networks seem to generate lots of sales based on the reporting, but in reality, we don’t see as many sales?’ This is one of the questions our clients ask us more and more often. This problem could be caused by many factors, particularly by different methods of calculations of the number of transactions. Let’s take a closer look at some of these factors in our client’s example.
The client has been cooperating with a CPA* partner for a while already. As the network owner works with a large number of webmasters simultaneously, it isn’t always possible to track whether these webmasters take all the client’s preferences into account or not. This results in controversial situations between the CPA networks and their clients.
According to the CPA partner’s reporting, they generate 15% of online sales to our client. However, the client’s internal reporting (CRM and Google Analytics data) shows that the share of sales is 5% only.
As we know, the CPA partners usually get paid per each sale; therefore, there was a possibility that in this situation, our client overpays 3 times. In this case, a client has to understand the differences in conversion calculation methods and gather evidence that not all of their preferences were taken into account and that the client’s methodology is more exact, considering their unique features. This is necessary to build a constructive dialogue with a partner.
That’s why the client turned to us, asking us to help them explain the reason for discrepancies in reporting and provide recommendations for future cooperation with the CPA partner.
*CPA networks are systems that serve as mediators between advertisers and webmasters (those who place ads) with CPC (Cost Per Action).
To identify the causes for discrepancies in the number of conversions to pay only for the sales that are indeed generated by this channel.
STEP 1. We used Google Analytics API and PYTHON from Google Analytics to collect the following data:
- Client_ID — the unique user identifier;
- Session Duration — the duration of a session in seconds;
- Hit Timestamp — the variable that fixates an exact time of each action;
- Count of Sessions — the serial number of a user’s session. Each session from a unique user receives its own incremental (the one that increases with a particular step) index, starting from 1 for the first session. The following sessions don’t change the indexes of previous ones.
- Session Source and Channel — this parameter describes the traffic source for a website.
- Transaction ID — the unique transaction identifier on a website that matches the order number in CRM.
STEP 2. We used the transaction ID to combine the unloads from Google Analytics and CRM with the CPA report’s results.
STEP 3. Based on this data, we created reports than enabled us to detect the following inaccuracies in system’s work:
1)The session was disrupted with the partner’s cashback widget, provided that the traffic from such a tool isn’t taken into account.
As we have the following information about sessions:
We got the time intervals between a user’s session and saw that for a certain webmaster, the time between the start of their session and the end of a previous one is less than 1 second. After filtering the sessions of this webmaster, we saw that a session starts on a checkout page. We turned to the partner to clarify this thing, and they confirmed that this was the cashback extension. Such extensions could be useful from the marketing’s point of view; however, specifically for our client’s business, they weren’t relevant at this point. This inaccuracy was removed.
2) The next inaccuracy occurred in the attribution model: the transaction was recorded on the CPA network each time it occured in a path.
In the beginning, we started looking for the cause of the discrepancy in Google Analytics data and the partner’s report.
After we created a report, we saw that part of the transactions listed in the partner’s report is assigned to other sources in Google Analytics.
We assumed that a different attribution model could cause this. The data from Multi-Channel Funnels Reporting in Google Analytics proved that. Multi-Channel Funnels Reporting API and Google Analytics Add-on helped us to track the paths for each transaction.
Then we added a column to the unload that displayed if the partner recorded this transaction for their channel or not.
This enabled us to prove that the CPA partner’s attribution model is the last click in their system. That is, if the partner’s channel appeared somewhere on the way to conversion, the conversion was assigned to this channel. If the partner’s channels appeared multiple times, the conversion was assigned to the last of them.
3) Some webmasters had longer Cookies Storage Period than the one that was agreed upon.
We used the following information about the users, which conversions were recorded by the partner:
- the date of the last session from the CPA/affiliate source;
- the transaction date.
We reduced the current cost of the CPA network services by 58%.
We reviewed the contract with the client and terms of cooperation with the CPA partner, and created a win-win strategy:
– We listed in detail which tools can be used and which cannot; we also listed the consequences of using the unwanted tools;
– We stated that the Cookies Storage Period should be comparable with a user’s purchase decision period.
This enabled us to eliminate the future risk of ineffective expenses.
We succeeded. The costs spent on analytics paid off 10 times during the first month the obtained results were used.
During their negotiations with the CPA network representative, our client used the results of audit (charts, spreadsheets, concrete examples of differences in methods used fo transaction calculations) as the evidence base for their arguments.
The key moments that helped us achieve a positive result:
- Correct Google Analytics settings. Although our client didn’t have Data WareHouse, the analytics collections tool in Google Analytics was set up correctly. It enabled us to create necessary information sections, even if it required a lot of work.
- The client was involved in the process, and we build a trust relationship with them.
The client’s time expenses on audit:
- 3 hours of the client’s time spent on the briefing, presentation of the data, and discussion of the audit results;
- 12 hours the analytics spend to create the evidence base, check up the website’s settings and the work of statistic collection system;
- 4 hours of work with alternative traffic sources.
If you plan on using a large number of channels, plugging in to the partner programs or CPA networks, it’s important for you to:
- correctly set up the collection and the storage of the user interaction data;
- analyze the reporting and identify the triggers that could help you define whether the reports are accurate or not.
And the most important thing.
Don’t close your eyes to data discrepancies. Look for the causes until you find an explanation. A constantly present error could lead to unnecessary expenses, and, therefore, to marginality decrease.
We are #ADINDEX, an integrated internet marketing agency. We are a part of Premiere Google Partners and work on the market since 2013. Our key clients are DEKA, Aromateque, HOLZ, Cyfra, UkrTelekom, Lasunka, TOYS. Our goal is to help the clients’ businesses grow.
If you need help with marketing or web analytics, contact us at email@example.com or call +38 097 735 31 33. We’ll develop a strategy for you, create secure web analytics, and help you boost sales. You can also book a consultation with the agency’s CEO, Vadym Pylypenko.