We're presently trying to come up with a way to filter out what we are calling "ghost clicks" (that is, clicks from a spam filter like Barracuda, not from a real person) from our Marketo activity log data, which we've downloaded to our data warehouse. This topic has been discussed several places, but apparently without much resolution (see here, and here, and here, for example).
I'm wondering if the timing of an activity can be used to spot "fake" clicks from a spam filter - so, if a "click email" event happens only a few seconds after the email send activity, maybe that isn't real.
But are the timestamps on Marketo's activity log really accurate down to the minute and / or second? If a "send email" activity is logged at 19:46:11.000 and a "click email" activity occurs at 19:46:53.000, is it true that the email was fired off from Marketo's servers exactly 32 seconds before the "recipient" (or the intended recipient's spam filter) clicked? This particular example was also telling in that the "Email Delivered" activity occurred at 19:47:04, so the click was registered prior to the delivery. But do we even know that these events are registering in the correct order? Can the timestamp in the activity log be trusted?
Solved! Go to Solution.
The Email Send activity is not accurate. If you send a email to 10 million leads, we just queue up a task to write "send email" on each lead's activity log, but it can happen at an arbitrary time (not the exact moment the email going to the lead actually leaves Marketo). The Email Delivered activity is very accurate, and in practice, extremely close to the time the actual email was sent to the lead as well.
Justin
Hi Matt,
PM will confirm, but my understanding is that the execution order is accurate and can be relied on to understand the sequence of events.
-Greg
The Email Send activity is not accurate. If you send a email to 10 million leads, we just queue up a task to write "send email" on each lead's activity log, but it can happen at an arbitrary time (not the exact moment the email going to the lead actually leaves Marketo). The Email Delivered activity is very accurate, and in practice, extremely close to the time the actual email was sent to the lead as well.
Justin
That's helpful, Justin - thanks. What about "Email Click" - is that accurate? So if the delivery "happens" after the click (at least in the activity log), then we might be good to assume that we're not looking at a "real" engagement with the content?
Not necessarily, see this: Why Does My Smart List Count Not Match the Email Performance Report?
There are definitely cases where we won't get a 250-accepted delivery event but then an open or click will occur and it is totally valid. It's just that we didn't receive a delivery event from the recipient mail server for whatever reason (see resources below, for a few examples). There is definitely no guarantee that we'll get these events back correctly from the recipient mail server and we have to handle cases where we don't. Our Email Performance Report has business logic built-in (as described in doc above) to show the right thing in these cases. In my opinion it will be very difficult to do what you're trying to do.
See these two references from SendGrid
https://sendgrid.com/blog/smtp-server-response-codes-explained/
Hi Matt,
The hard point will be to put this into action. You will not be able to find the "false clicks" through filters. You will have to extract the activity log data through the API and process it outside of Marketo.
-Greg
We have already extracted the activity log (we use a combo of Progress DataDirect and SQL Server Integration Services), so that isn't an issue - the issue is data quality. We're looking at doing a company-level lead score, rather than just an individual lead score as Marketo does - but we want to make sure we're not basing that score on bad data.
Hi Matt,
That's an interesting use case for Company score.
Mahesh Jeswani (MJ), this is for you
-Greg