MT(T)M: Metrics that (truly) matter

The digital advertising industry turns 29 this year, looming on the big 30 next year.

For those of you are not aware, the image below is the first digital ad to have ever been utilised. It is a display banner:

What is fascinating about this banner is that, with a USD 30,000 budget, it generated a click-through-rate of 44% - That is almost 1 in every 2 impressions resulting in a click.

This is what gave birth to the "click" being a direct unit of measure for "Brand Engagement" and hence "Intent" and hence "Performance" of branding and engagement campaigns.

The average CTR (click-through-rate = Clicks/Impressions*100) for display ads currently is 0.46%.

Here is a breakdown of global CTRs by industry:

Source: Wordstream 

We are talking about a decline in CTR from 44% to 0.46% - A 100x drop.

The immediate question that arises now is - Is the CTR a good measure for engagement anymore?

Follow up questions then become:

1. What is engagement in the real sense?

2. Can clicks be misleading?

3. What other industry standard metrics are misleading?

4. Does breaking this open make things simpler or more complex?

Those are just some thought starters. Before going into some of my thoughts on this, here are more stats to help usher us into understanding the deception we function in:

• Approximately 38% of web traffic is automated/bots (Imperva)

• Of these bots, 24% are considered ‘bad bots’ or bots used for fraud and theft (Imperva)

• At least 15% of paid web traffic is un-attributable, most likely lost to fraud (Cheq)

• Rates of click fraud jumped 21% during the initial stages of the Coronavirus pandemic(source)

• Industries with the highest rates of click fraud include photography (65%), pest control(62%), locksmiths(53%), plumbing(46%), and waste removal (45% (source)

• Financial and legal services were hit by around 25% invalid traffic/fake clicks during 2020 (source)

• Companies spending $10,000 per month on Google Ads are estimated to be losing approximately $12,000-$15,000 each year to click fraud (source)

To understand the problem, lets do some math:

1. Average CTR: 0.46% (that 4.6 clicks from a 1000 impressions)

2. Assuming a lower end of bot-click percentage, 24%: 
(1-24%) x 0.46% = 0.34% - This is the real CTR post bot clicks removal.

3. Adding accidental (unintentional) clicks to this, of about 35%:
(35%) * 0.46% = 0.16% of the clicks are accidental

4. So overall click number that might be of any value at all:

4.6 - 1.6 - 1.1 = 1.9 (approx 2).

That is 2 out of 1000 impressions which is 0.2% CTR. Taking this as a fraction of the benchmark of 0.46%:

(0.46 - 0.2) / 0.46 = 56%

This is the amount of wasted clicks that that are either bot or accidental going by general statistics available publicly.

This post is not about the best practices of reducing bot clicks or accidental clicks. It is about re-imagining digital ad metrics in a way that does not permit this methodology of deception in the first place.

Deception cannot persist in a system that chooses not to attribute value to it

If not clicks, then what?

To get to the bottom of this question, I figured the best thing to do is look to the past and ask "why clicks in the first place?"

Simplest answer to this would be that clicks are, essentially, the way we do anything online.

So it is was only natural that Hotwired (the first commercial online magazine, the legends that introduced the digital banner to us all) would measure the impact of their display ad by checking out how many people...well.. clicked it.

The number was 44 out of every 100 people who landed on their page, so naturally Hotwired reported this as a metric of absolute and undeniable success.

But digital advertising or not, what did the click signify in the online world? Easy:

It indicates navigation.

The click is the means to navigate the online directory called the Internet. There is no other way to go from point A to point B in the internet without clicking through from one section to another, one page to another, one site to another etc.

So, applying that logic to advertisements, a click is valuable because it measures navigation; Navigating from the environment of the advertisement to the desired location, usually the advertiser's owned space, normally a website or an application.

Essentially, this means that we are using the "click-through-rate" as a means of identifying how many people had used the advertisement as a vehicle to navigate from the publisher's page to the brand's landing page. This navigation is useful because it adds to the funnel of people the brand is attempting to "push-down" and "convert" - Meaning get them to become a customer.

So, if not clicks then what becomes a question of "is there are better way to measure the arrival of people from advertisements to landing pages other than the click based indicator?".

Well of course. And here are some of my thoughts on what these could be and how we can have them implemented:

1. View-Through-Traffic-Rate (VTTR)

Yes, I know something like this exists (VTRs). But only for video. And it measures video views / impressions which is a measure of how many people viewed a video.

I am proposing a view-through-rate for display advertising that measures traffic generated by views (instead of clicks).

This is built on the following theory:

A click is not the only way people consume advertisements. Folks could see an ad, have it registered in their minds, and navigate to the brand site by other means, later.

This means that the traffic that landed on site after having SEEN an ad needs to be measured just as much as the traffic landing on site through a click on the ad. In fact, I would argue that traffic that lands on site after having "seen" the ad is far less likely to be bot or accidental.

I will first talk about how to implement this and then cover some caveats to this thought.

Implementing this:

1. Extract pixel from your digital advertising platform. Lets go with Meta for this example

2. Place Meta pixel on the landing page to fire when the landing page loads (higher order thinking: you could make it fire after about 5 to 10 seconds to allow for page load time and possible crashes etc.)

3. Save this pixel fire as a custom conversion calling it something like "Real Traffic To Site"

4. Set the view-through window for this as low as possible. I would recommend between 1 to 2 days.

5. Measure view-through conversions for this event - This would be the number of people who saw the ad and turned up on the brand's site within 24 to 48 hours.

On DSPs, you could have a Viewability metric attached to this by only measuring the number of this particular event that was driven by viewable ads (that is ads that had at least 50% of its their pixels loaded in the viewport for a period of at least 1 second).

This would increase the chances of the ad actually having been viewed by the user who navigates to the site later within 24-48 hours.

Utilities for the above metric:

1. Impact buys - When a brand spends a considerable amount of money buying ad slots on major publications, for example, as roadblocks - CTRs cannot be the measurement here. The aim of this kind of a buy type is not to get clicks but to generate interest. View based traffic from this ad slot within 24 to 48 hours of having seen the ad would be an interesting thing to measure and correlate back to the ad buy itself

2. Rich Media advertisements - The aim of spending more money on creating visually appealing, interactive creatives, and spending a higher CPM (cost per 1000 impressions) on these creatives being delivered is because they are "eye-catching". By only looking at CTRs, we might be missing out on how much view-based traffic these creatives might be generating

3. Generally all display advertisements - Any display banner, through any platform, should be measured upon its view-based traffic and its click-based traffic as well.

The main caveat (possible challenge) around this measurement would be:

The Problem Of Attribution: 

What if someone "saw" advertisement A from channel 1, clicked on advertisement B from channel 2 and landed on the site? Do both the channels get attribution for that traffic (channel 1 delivering view based traffic and channel 2 delivering click based traffic). Who actually is responsible for the traffic?

Solving the above is fairly simple - It is the same solution that utilised for any form of duplication with regards to conversions.

Use a central reporting system like an Ad Server.

The ad server would provide us with de-duplicated conversion numbers, and hence would automatically take into account things like "Last-View", through the core data of "Path-to-Conversion".

So measuring VTTR by extracting view-through conversions from the ad server (where the conversion in question is landing on-site) would automatically un-do the challenge mentioned above.

2. Cost Per Engaged Visitor (CPeV)

The previous proposed metric forced us to re-think the idea of the "driver of traffic" - From clicks being the only driver of traffic to now including the view of an ad as a driver of traffic as well.

The Engaged Visitor metric forces us to go one step further and validate the traffic itself. Instead of focusing purely on the ratio of "How many people did my ad drive to the site", it is more of a measurement of "how many people of value did my ad drive to the site".

In many cases, digital marketeers jump straight from the traffic based metric, which is the CTR, to direct lower-funnel metrics such as:

1. Cost per Lead

2. Cost per Sale

In my opinion, this is lazy thinking.

When the processing power of today's software infrastructure is accessible to us at scale, sticking to simple measurements such as either a click (which is too misleading) or a lead/sale (which is too stringent) disallows advertisers from measuring and/or optimising for a myriad of metrics in the middle.

Between the click and a lead is the world of "engagement".

According to Adobe, "engagement is the use of strategic, best marketing strategy, resourceful content to engage people, and create meaningful interactions over time"

I think the key phrase there is "create meaningful interactions". So going by that, site engagement would refer to actions taken by the user that the brand considers to be "meaningful".

Some of the interesting actions that can be used as strong indications of engagement could be:

1. Time spent - Use site analytics (such as GA) to determine a good time duration that indicates interest (such as 2 mins on landing page or 2 mins on site)

2. Number of pages navigated - This is a great way to identify true interest in the brand. A user how navigates 2 or more pages is clearly engaged with the brand. This is the kind of user you want to optimise for and send to the site.

3. Combination - A compound metric such as "Spent 1.5 minutes on landing page and navigated to 2 or more pages" could be a great metric to optimise for.

What we are doing in the process of ideating this is defining, for the particular brand, what an engaged visitor looks like.

Once this definition is complete (e.g. an engaged visitor is someone who spends 2 mins on the landing page), then the next steps are pretty straightforward:

1. Extract a pixel from your platform (lets take DV360 for example)

2. Implement the pixel (in this case called a floodlight) onto the site and enable it to fire only after 2 mins

3. Use that pixel fire as your "conversion", call it "engaged visitor (eV)" and optimise all line items for that particular event

4. Have a clear objective such as "the CPeV should be $5" and optimise towards this using the same tactics you would use for any CPL or CPA objective

By doing this, you have effectively shifted a typical "branding" campaign, which would have been running on click through rates, to a performance style campaign, with a more fulfilling metric (the CPeV)

Optimising this way will have a few positive halo effects on your campaign:

1. It directly impacts the metrics that (truly) matter to the advertiser such as time spent on site, or no. of pages viewed or bounce rate. It will directly improve the image of your campaign on Google Analytics

2. The algorithm will be forced to optimise towards a metric that is far more validated than the click. This means that it will learn to identify a better image of the "right user" for the brand and target them. This results in better click through rates as well which is a nice little bit of a bonus there.

3. It will result in better quality traffic to your site, resulting in your retargeting campaigns being more optimised and budget better invested

The CPeV can be customised based on the mutual understanding of on-site/app user behaviour between your agency team and the client. Some easy compound actions that can equal to an "Engaged Visitor" can be:

1. Navigated to more than 2 pages

2. Spent more then 1.5 mins on the landing page

3. Spent more than 2 mins on the entire site

4. Spent more than 2 mins on site + Navigated across more than 2 pages (has a strong impact on GA performance)

5. Downloaded a brochure OR Visited a Product Page (or equivalent) for 45 seconds OR clicked into a lower funnel page such as a "Dealer Locater"

The last option in the list above uses an OR function. In this approach, you are not compounding actions (such a point 4 in the above list) but instead, you are creating a range of actions, all of which hold equal value in your definition of an Engaged Visitor.

Defining what your brand's "Engaged Visitor" looks like is the major gap that the digital, outcome oriented industry has been neglecting. Unlocking the Engaged Visitor allows for far better optimisation protocols, better data insight, deeper audience taxonomies and better Metrics That (Truly) Matter