Redefining “Impressions”

Over the past few years, I’ve seen a lot of reports and case studies from advertising and creative agencies, where number of impressions i.e. the amount of people who have been exposed to a creative, is used alongside site visits and views to measure the success of any given campaign.

Frequently, I’ve come across a report which contains something like the following (example from a YouTube marketing campaign):















Of the numbers above, I have no particular issue with clicks, views, CPV or CPC - I know what all of those mean in real terms, and with proper tracking in place, I can quantify the value of those interactions by measuring assisted conversions under a sensible attribution model.

However, the metric here which continually frustrates me is impressions. Is 2.64 CPM good? Where did those impressions take place? How many of those impressions actually resulted in, well, an impression?

Unfortunately, it’s almost impossible to say.

What is actually measured by “impressions”?

The convention of tracking impressions as a measure of success - or more specifically, as a measure of “reach” - comes from a time before we had Google Tag Manager, biddable media, or even computers. Readership and viewership were originally used to quantify the value of specific TV and print advertising slots, because a rough estimate of the amount of people likely to see something was the only metric we had at our disposal at the time.

When advertising dollars moved online, impressions became the equivalent of readership and viewership - the number of people that the content is displayed to. However, if we really drill down into what triggers an impression, it’s simply...

Status: HTTP/1.1 200 OK

Back in the heady days of 2005, when inventory was comparatively low and users were unfamiliar with the conventions of online advertising, it’s highly likely that an impression really did mean an impression -- that your ad loading on a page meant somebody saw it.

But, fast forward ten years, and the noise has dramatically increased.

We browse the web on a multiplicity of dimensions and resolutions; we work on browsers with multiple tabs open at once, loading dozens of pages simultaneously; we omniscreen; we run ad-block extensions; we are adept at finding pages in search, loading results in fractions of a second and clicking through to desired results efficiently; and most importantly, we’ve come skilled at mentally blocking the ads we don’t want to see, through a process of habituation.

Technology and user behaviour have changed dramatically, but measurement hasn’t.

While a successful page load may once have meant someone saw and mentally processed your ad, today it doesn’t really mean anything at all. Indeed, for platforms like YouTube advertising, or paid content distribution with the likes of Outbrain or Taboola, an impression just means a succesful page load where a fragment of the content was visible somewhere on the page.

So, in short, I think it’s a pretty useless metric, but nevertheless one which we hold to religiously, likely because the lack of concreteness allows us to obfuscate genuine performance with figures that look good on paper, but don’t actually mean very much

I think we can do better.

So, how do we improve?

I want to reclaim and redefine impression, turning it from a mostly useless metric into a meaningful one. An impression should mean an impression in the sense of “an idea or opinion of what something or someone is like”, not just the fact that for some browser somewhere in the world, a 200 response code was served.

Take the analogy of meeting new people at a party. If someone makes an impression on you, it’s because you spoke to or interacted with them in some way and were left with a feeling about them based on this interaction. We should judge content and advertising in the same way. Right now, we’re often judging our campaigns on whether they turned up to the party, not whether they actually had fun and interacted with people while they were there.

shutterstock_157016321 (1).jpg

image credit: shutterstock

For me, an impression shouldn’t just be the number of people who possibly saw an ad or piece of content, but rather a metric that dictates how many people were actually compelled into thought or action by the content itself. It shouldn’t be a lesser metric than clicks or views, but rather a distinct metric that demonstrates a level of engagement and interaction synonymous with the idea of “leaving an impression” - I’m calling this metric Meaningful Impression (MI).

What does this mean in practice?

I propose a few new ways of measuring “meaningful impressions” based on ad and content type.


For PPC advertising - specifically AdWords placements in the Google Search results, I think we need to more or less disregard the idea of impressions as having value independent of the clicks they drive. It’s my contention that unless a user actively clicks onto your site from either an organic or paid listing, there is little to no value in having your URL appear in a Google search result.

When measuring organic search traffic, we typically don’t look at the number of impressions to quantify success, and I don’t think we should when measuring PPC either.

However, there is one common instance where a PPC ad may provide a valuable impression that doesn’t lead to a click, and this is when the co-occurrence of a PPC ad alongside an organic listing seems to assist in driving more clicks to the organic.

Due to (not provided), this is hard to quantify to an absolute degree, but I would start as follows:

  • Calculate the estimated number of new organic visits a page receives when a PPC is turned on

  • Add the overall number of PPC clicks driven from keywords where there is an organic listing ranking well

  • Divide this cumulative number by the overall spend on relevant keywords

This will give you a rough figure for CPMI - cost per meaningful impression.


Display is currently very difficult, because there is some legitimacy in the idea of being impressed by an ad which is observed but not clicked, but equally it’s not reasonable to suggest that all page loads result in such an impression.

Sadly, I don’t think there’s anything we can do right now to improve measurement here, limited by the technology. However, there is currently a movement towards measuring “viewable impressions”, which I’m very supportive of. 

In the mean time, view through conversions added to overall clicks are probably the closest thing we have to measuring a meaningful impression, although this metric will likely under-represent the number of overall meaningful impressions in most instances.


With video, however, I think we can do a lot better.

An impression on a video is ostensibly a view of any length, whereas a view is a qualified impression based on the amount of time watched. For YouTube, views of ad units e.g. sponsored links in the sidebar, is also lumped into the figure for overall impressions.

As anyone who has done much video advertising or video marketing will know, view count is not really a very robust metric to determine how many people have really been impressed by a video. Many views may be only five or six seconds long, and videos are often played in the background on a hidden tab.

In my view, a meaningful impression on a video looks like one of the following:

  • A user actively watches most of the video

  • A user completes some sort action off the back of the video

While, currently, it’s hard to measure the above absolutely (the hidden tab issue is hard to solve with current technology), we can move a great deal further towards a robust metric, by reporting meaningful impressions as:

Engaged Views + Engaged Actions

I suggested a formula for Engaged Views in this post from last year, but to summarise - it’s basically views multiplied by the average percentage viewed. Engaged actions is then a cumulative metric to describe positive next steps taken off the back of a video which indicates a level of engagement or interest with the content. In short, one of the following:

  • A comment

  • A share

  • A click to subscribe/submission of email to subscribe

  • A click to visit an external website or another page

  • A like/+1/thumb-up

Note that it’s likely, for most videos, that you’ll end up with a greater number of views than engaged impressions. I think this is appropriate, because as mentioned above, many views are not impressions in the true sense.


Impressions shouldn’t be about page loads, they should be about impressions. Shifting towards better metrics will be hard, and the technology still has a way to go, but those of us who care about actionable metrics should start experimenting and improving our reports now. If you’re an agency, I’d like to see you testing out the metrics above and seeing how your campaigns measure up when put under a slightly more detailed microscope. If you’re client side, I’d love to see you start holding your agencies to account better on the numbers they’re giving you - instead of accepting CPM as a figure, ask for details on the CPMI as detailed above, or another metric which gives better clarity about the number of real, genuine impressions you achieved.

Have you got any other ideas of better ways to track meaningful impressions? If so, I’d love to hear about them in the comments. 

Get blog posts via email