Revenue Discrepancies Troubleshooting

There are many critical differences in the way different systems track web data and those differences drive dramatically different results. We have documented a number of these differences here.

In the current Google Upgraded URL world, when looking for revenue/conversion discrepancies, look for changes to the Tracking Template or Custom Parameters around the same date that the discrepancy began. Typically, these changes result in a change of tracking logic which can cause discrepancies if the change is significant and interrupts this previous logic.

Note: Please make sure you are narrowing the discrepancy down to a single date for a single sample keyword/ad. We can expect a conversion discrepancy of 0-2 conversions per date for a single keyword/ad due to tracking logic differences between two different tracking solutions. When this is expanded to a larger date range for a larger number of keywords/ads (i.e. a month's worth of data for a single account that covers many keywords and ads), then the discrepancy will appear larger. This is expected behavior. In these types of instances, all that can be verified is that your current revenue integration is set up correctly.

If you are seeing a conversion discrepancy greater than ~2 conversions for a single keyword/ad on a single date, please file a Customer Support ticket for further assistance.

Date Of Click Vs. Date Of Conversion

If visitors click on an ad on Friday, discuss the item with a friend on Saturday, and make a purchase on Sunday, Date of Click attribution will assign the conversion credit to Friday, whereas Date of Conversion attribution will assign the conversion to Sunday. 

The best practices in search marketing are to use Date of Click conversion data to match publisher cost and click data with revenue and conversion data; it is also required for accurate day-of-week bidding. Order entry systems and accounting systems generally use date of conversion, making matching difficult. Internal tracking systems should store the date of the click so as to be able to report on this.

System Date of Click/Date of Conversion
Google Ads Conversion Counter Date of Click
Google Analytics Date of Conversion
Marin Tracker Date of Click
Google Ads 360 Date of Conversion

One-Per-Click Vs. Many-Per-Click Conversions

Let's look at an example:

Visitor A clicks on an ad, makes a purchase, and then makes another purchase. Visitor B clicks on an ad and makes a single purchase.

One-per-click will count a single conversion for each visitor. Many-per-click conversions will count 2 conversions for Visitor A and 1 conversion for Visitor B. The best practices in search marketing are to use one-per-click attribution for conversion count but compute the total revenue across all purchases. The reason is this is because one-per-click more accurately matches how you are paying for that traffic; in other words, when you map your conversion rate, a conversion rate exceeding 100% does not make sense. Many-per-click conversions are more accurately thought of as an order or transaction count. The difference can raise discrepancies when comparing against an internal order entry system. To prevent this problem, save the entry timestamp and referring URL with your internal orders in order to differentiate between separate paid visits.

System One-per-click/Many-per-click
Google Ads Conversion Counter Both
Marin Tracker One-per-click (many-per-click available as a custom report)
Google Ads 360 Processed by Marin to result in one-per-click

Other differences:

Issue Marin way Other way(s)
Multiple transactions on the same paid click Marin will attribute one or more transactions to the last paid visit and count it as one conversion (the revenue will be totaled). Unique order IDs should be included in the order header sent to Marin in order to identify situations where a page was reloaded but there is not in fact a new transaction. Some in-house tracking solutions will count this as multiple conversions. Google Ads conversion counter makes a distinction between “one-per-click” conversions (that is what Marin provides) and “multiple-per-click” conversions.
Multiple transactions over multiple paid clicks Marin will attribute one or more transactions to the last paid visit and count it as one conversion (the revenue will be totaled). Google Ads conversion counter generally functions the same way; however, if a visitor makes two paid clicks and then performs two transactions (in that order), both the paid clicks will pick up a conversion.
Time zone Marin uses the time zone of the MarinOne Client (which is, in turn, generally determined by the publisher accounts which are linked to MarinOne). Atlas and Google Ads 360 are based on US Eastern Time. For example, transactions at 10 p.m. Pacific for a California marketer will be attributed incorrectly to the next day.
Cookie duration Marin typically uses a 30-day cookie. However, this is a client-specific setting so this is adjustable. Google Ads conversion counter uses a 30-day cookie window. If a visitor does a search on March 1st and makes a purchase on April 2nd, the purchase will be unattributed in Google but will be attributed in MarinOne (for now).
Cookie party Marin’s redirect uses third-party cookies. Marin’s javascript-based tracking uses first-party cookies. Google Analytics uses first-party cookies. As a result, Google Analytics will report higher numbers than third-party tracking solutions if some visitors block third-party cookies.
Paid clicks with valid keyword IDs in URLs MarinOne will only report revenue up through the totaling hierarchy (groups, campaigns, overall) if there is a valid and unique keyword associated with the keyword ID to attribute it to, and if the keyword’s campaign is linked into MarinOne. This issue could cause discrepancies vs. an order entry system, which may not discriminate between paid and unpaid visits.
Pixels are getting cached Marin pixels are set to avoid caching but some browsers may cache them by mistake. This is only an issue for testing and auditing; testers should be sure to clear their image caches before testing to better simulate external visitors.
Different conversion types Marin can handle distribution of revenue across multiple conversion types, for example, a newsletter sign-up, etc. Other systems may use different or no types, resulting in different totals.
Attribution over multiple visits Marin typically attributes revenue to the most recent paid visit for that conversion type. Other systems may attribute revenue to the first paid visit, or spread it over some number of prior visits, resulting in different totals.
Definition of “last” visit With landing page click capture, this is the last visit where the requested page contained a Marin keyword ID parameter. With the Marin redirect, this is the last visit that went through the redirect. Google Ads conversion counter looks at the last Google paid click. If somebody does a search on Yahoo between their paid Google click and their purchase, Marin will attribute the conversion to the Yahoo visit but Google will attribute it to the last Google visit. Other systems may use the last visit, whatever the source (eg natural search).
Placement of click capture technology

Marin offers three approaches to capturing the click event, in order of most accurate first:

  • Redirect based
  • Landing-page pixel generated on the server-side
  • Landing-page pixel generated by javascript

Marin may appear to overcount or undercount due to differences in deployment methods vs. other methodologies.

For landing-page approaches, if a visitor stops a page load on a complex page or over a slow connection, the pixel may not have a chance to fire for them. Placing the pixel at the top of the page will be more accurate.

Still, other visitors may block javascript in their browsers. Therefore the server-side approach is slightly more accurate than the javascript approach.

When a conversion type (ie. Total Conv. column) is set to One Per Click, all conversions will aggregate back to the first click but revenue will not. For example, if a user clicks through paid search and converts on Nov 1, this click and conversion will be seen in MarinOne on that date. If the same user goes back to the website on Nov 5 through a non-paid click and does another conversion, this conversion will aggregate into the first conversion on Nov 1, but the revenue amount for that conversion will apply to Nov 5.

This explains why there could be a revenue amount for a specific date, but no conversions when viewing the data as Date of Conversion.

Latent Conversions And Revenue Upload

In some online marketing categories a “conversion event” occurs during the same visit as the click event. In contrast, for some categories, the conversion event may be hours or days after the last click. This so-called “conversion latency” can introduce accuracy questions when comparing systems that account for that latency in different ways. For example, a system based on Date of Conversion will more closely match the total conversions for a day. Tracking based on Date of Click will cause recent days to appear to have a slightly lower conversion rate.

Neither attribution method is incorrect but they are suitable for different things: Date of Click attribution is superior when calculating bids for paid search terms, as it better reflects day-of-week differences. Secondarily, it provides backup protection against runaway conversion counting or a failure in the cost reports, since the algorithm can assume that conversions will never exceed clicks. Date of Conversion attribution better matches internal accounting systems. The following highlights different tracking systems and the attribution approach for each:

Date Of Click

Google Ads Conversion Counter
Marin Tracking
Marin‟s Google Ads 60 Integration with Data Transfer Option

Date Of Conversion

Google Analytics
Omniture
Core Metrics
Yahoo Conversion Tracking
Microsoft Conversion Tracking

One measure of overall latency is the share of conversions at the end of a given day as a % of the eventual conversions that will accrue to that day using Date of Click attribution. If this amount is below 90%, we consider that significant latency.

Here are some selected tips for working with this information inside MarinOne:

  • If your business faces significant conversion latency, we recommend tracking both values. Date of Click conversions should be loaded as conversions for these clicks and Date of Conversion conversions should be loaded as conversions for date.
  • Set up bidding on conversions for these clicks as this would more closely match the click traffic coming from the publishers. Adjust the trailing date exclusion parameter* to exclude the prior x days. Set x as short as possible to get 80% of the conversions – typically 1 if anything.

    *Note: The trailing date exclusion parameter will not be visible in your MarinOne client account. To adjust this, please reach out to Marin's Customer Support team.
     
  • If you are doing your own tracking, when a click comes in, record add keyword-URL *and* the timestamp of the start of the visit. The date of that timestamp is the date of the click. When uploading revenue, upload in append/add mode (as you may be incrementing prior days for which there is already existing data with conversions that just came in). If using Date of Conversion exclusively, upload in overwrite mode.

Written by Marketing @ Marin Software

Last published at: June 30th, 2023