Mysterious drop in image link clicks

[BenCirillo]

Have a bit of a mystery I'm trying to solve. 

We have an email newsletter that goes out every week to about 20k subscribers. In this newsletter we sell ad space. Nothing fancy, just a square image that is linked to the advertiser's website. 

Week to week, we had been seeing about 100 clicks from about 50 people on these ads. Not crazy but not to bad for what are essentially banner ads. 

Then, around the week of 12/6/2022, we saw a dramatic drop in clicks to those links, down to the single digits. After a few months of data, that pattern has stuck around, so it's not a one-time fluke. What's odd is we aren't seeing a drop in clicks on the other links, except clicks to are social media icons, which are also image-only links. 

I have bot activity set to logged but not filtered. I checked emails before and after and found some clicks attributed to bots and some not, so I know we're capturing all the activity. I checked and there's no change in the way we have been entering the links. While the clicks are low, we are getting some, so I know they're working. While I could attribute it to consumer behavior, the sudden rather than gradual drop makes that seem unlikely. 

There doesn't seem to be anything we did to trigger this change. Does anyone know of anything internal to Marketo or something in the greater email ecosystem that could be either hiding clicks on images, or that could have been artificially inflating them previously? 

[Dave_Roberts · Accepted solution]

Hey Ben,

It looks like your problem stems from December, but I just read something about a Marketo update in February that seems to have drastically dropped some folks numbers and maybe there's something in there that's relevant to your situation? 

Here's a link to the community post for more info: https://nation.marketo.com/t5/product-discussions/sharp-decline-in-clicks-and-opens-after-feb-15-upd...

I'd guess that this might have something to do with email link scanners or some kind of machined-solution b/c there's a "categorical association" you're able to make (all linked images, but not all links). I think it'd be weird for an audience of people to behave so commonly all-at-once.

Also, 100 clicks from 50 people seems disproportionate and might suggest that there's some machine clicks that you were recording unless you can safely assume that folks who opened and read the newsletter were clicking on the ads twice which seems like a strange behavior for an end-user?

When you look at the number of clicks an email gets (for all the links) does it seem like it makes sense? For example, is there an average of at least one click on all the links in every email sent? If yes, this might suggest some kind of link scanner is at play and in turn might be something which could have gotten an update (eg. some software updated, or folks switched to a new software scanner).

Just thinking from my behavior with email newsletters, I might click ONE link in the page to go out and read about something that's interesting to me but I wouldn't click the banner ad AND the social icon(s) AND a link to read more content somewhere, that sounds like a lot of work to expect of the average reader unless you've got some super engaging newsletter content AND compelling advertisements AND clear call-to-actions setup to push folks to click the Social links.

Maybe more questions than answers here, but I hope this gives you a few things to consider in figuring this out. I'd love to read about what ended up being the culprit if you're ever able to get to the bottom of this one, I'll keep an eye out here for more.

[BenCirillo]

That's the kind of thing I was looking for. Bot clicks that aren't tracked as bot clicks. To me, looking at the numbers, the more likely scenario is previous clicks were inflated rather than current clicks are underreported, and this is a pretty good explanation of how that happened. 

Our engagement is normally a little depressed over the holidays, so it's not super weird to me that we might have had naturally low numbers and then this update kept the numbers low. Or that there's a combination of effects: Changes to link scanners on one end and this update on the other. 

[SanfordWhiteman]

One of Dave's points is the more likely culprit, but have you changed how you generate the hrefs of these links perchance? If you use a {{my.token}} and the token value includes the protocol ("https://") the link won't be tracked.