- Customer Acquisition Cloud
- Customer Engagement Cloud
- Industries We Serve
- Contact Us
For the third entry in our attribution series we want to focus on an area that is often annoying and messy for both reporting and decision making, and provide you with some tools to help you from tearing your hair out while having to wait 30 days to make a decision. Conversion lag.
Conversion lag represents the time between when a user was brought to your website through an ad, and when that user ends up making a conversion. The reason this effect is important is that it creates a distance in time between marketing cost and marketing value. The reason this effect is annoying is that the greater that distance is, the longer you need to wait to properly understand the effect of your investments.
This is a sample conversion lag pattern for a client with a 30-day cookie window and 54.81% (50.29% + 4.52%) of Conversions occurring on the first day after their last click.
This example is using click-time attribution as that is the default view of data in the Adwords interface. What is important to note is that conversion lag exist for both transaction time and click time data, the difference is only in how it shows. For transaction time data, the conversion lag will show in the way that today’s investments will generate conversions on future days. This results in situations such as paused campaigns continuing to receive conversions for $0 spend, something we expanded on in our previous article. For click time attributed data, you instead see the conversion lag in the form that recent data is “always incomplete” until you get out of your cookie window. So using the example conversion lag above, yesterday’s revenue will right now show about 55% of what it will prove to be in 30 days from now.
The way conversion lag impacts click time data can sometimes be quite stressful, as for a client with high conversion lag, it looks like results are always going down. It is not a great feeling to come into work, look at last 30 days performance and always see a downwards graph trending towards oblivion. Even if you know it to “probably be false” and skewed by conversion lag, it is nicer to be able to observe the true progress and improvements in the account.
This makes it hard to make decisions, as when evaluating “how good results were yesterday” you will always know that they are lower than they should be, and through this reasoning you may miss points where results are really declining, unless you keep track of “how good yesterday should look on average with your current goals”. One way to solve this problem and make it easier to evaluate performance is to make reporting include estimated revenue based on previously observed conversion lag. While the previous conversion lag is not guaranteed to be stable, this can still be a great way to get a good view into how results will look at a later point in time when the cookie window has passed.
So how can you make these estimates? It is actually really easy, and it is something you can build into your own reports so you can keep better track of performance pacing. The basic principle is that if you know yesterday to have an expected 80% of conversions visible, you can take your current conversions for yesterday, say 100 conversions, and divide them by 80%. The result is then 125 conversions which is a reasonable estimate of what we can expect to see for yesterday in 29 days from now. This process is repeated for the day before yesterday and so on.
To save you some time, we have prepared a process using Adwords Conversion Lag reports, and a template file you can use as a base to add into your reports.
Tools → Attribution:
Paths → Time Lag:
Copy following data from Adwords UI:
The far left column showing “Days to Conversion” lines up with the far right column of “% of Revenue to Date.” For example, in this image we expect to see 90.8% of the Revenue reported from 20 days ago.
Note: This template is made for a 30-day cookie window, but can quite easily be tweaked for shorter or longer cookie windows.
Since search runs a “pay-per-click” model, you will want to make sure you are on the “time from last click” tab when pulling your conversion lag estimates. Otherwise you will over-estimate them. Also, this assumes that you are optimizing and evaluating data on a last-click model. If you are using another attribution model ( read more about these in part 1 of our attribution series), your actual conversion lag landscape will be somewhere between the “from first click” and the “from last click” graphs.
Now that we have covere`d three of the main areas you need to keep track of in attribution, we want to use this knowledge to see how it translates into the important area of testing. Whether you are testing a new software or different strategies within your account, it is a very important component to continuous growth of performance. For this reason, we will dedicate our next post to this area to help you make sure your tests can be used for decision making.
Feel free to contact us for more information or for a free SEM strategy review.