Interviews, insight & analysis on digital media & marketing

The top five attribution mistakes and the danger they pose to everyone

By Rob Webster, Founder, Canton Marketing Solutions

It is an often repeated statistic that 80% of revenue goes to GAFA (Google, Amazon, Facebook and Apple). Their dominance within the modern landscape is seen as a primary contributor to the myriad problems the online industry faces.

With with public trust at an all-time low, there are significant concerns about how the industry funds hate speech and terrorism and harms child well-being. As a result several organisations have worked together to launch the Conscientious Advertising Network, which is a great step forward, to solve some of these problems.

All of this, to a greater or lesser extent, comes back to our problems with attribution. This is not to say that the platforms don’t need to take responsibility for the content on their sites, that privacy concerns need not be addressed or that a number of other things need not happen.

However, one of the underlying causes is that the attribution techniques that are standard in our industry do not reward good ethical content, they do not respect privacy and worse do not even drive ROI for our advertisers.

By fixing these common issues advertising and media in general would at a stroke be on a better path. Worse, the reason for these mistakes is not technological but one of education of best practice.

Below I have looked at the top five which would, at least partially, be solved tomorrow with existing technology.

1. Using web analytics for media attribution

I will accept there are times when it is acceptable to do this, for example by SMEs with small media budgets or sites built on an affiliate model. However, this does not explain or justify the huge proportion of digital media bought using these platforms to analyse performance.

Any advertiser claiming to have any belief in advertising should not be using web analytics alone to measure or optimise their media activity.

The first reason is that it is not designed to do so. Web analytics have been designed to measure what happens on a site and to a degree how you get there but it is not designed to measure media performance. For that you need a media measurement system which often comes with an ad server such as Flashtalking, Adform or Google Campaign Manager or as specialist provider like Visual IQ or Abakus.

These platforms do have incremental cost, starting at about £500 a month and rising, hence me excluding them from the SME conversation. However, they more than pay for themselves as advertisers are able to optimise their media to the best performing channels and sites once advertising spend rises above around £10,000 a month.

Why is using web analytics so bad? Simply they are only able to measure post visit and mostly only pick up brand search at scale (see number 4) as well as Google shopping. By using such a platform, you undervalue ads in prime placements where people do not click to buy immediately.

They are also only able to show value at the absolute base of the conversion funnel from users who were probably going to buy from you anyway (see number 3).

Consider you are an ecommerce company selling premium running shoes. A website that perfectly fits your target audience exists with lots of content about running. However, because the users are there to read about running, they don’t click and buy immediately in any number.

What they do is remember your brand and a few days later go to your site (probably via brand search) and buy. Under a web analytics attribution model you would not buy this media, and thus by extension fund great journalists and content and most importantly for you) grow your business.

Instead you spend all your money on the competitive and inflated world of Google Search and Shopping missing out on this great opportunity. Even worse Search and Shopping are in many cases stealing the demand created by these content producers ultimately hurting the publisher ecosystem and advertiser value simultaneously.

2. Encouraging cookie bombing

Over reliance on post-impression tracking that is available from ad servers and not taking additional steps leads to the infamous cookie bombing.

Consider that under the simplest post-impression models all a media vendor needs do is try and be the last impression served before the purchase. Notice I said impression served, it does not matter if the user noticed the ad or if it was even in view.

This led to the practice known as cookie bombing whereby vendors would run media as cheaply as possible, regardless of viewability at a frequency cap of one to try and drop a cookie on as many users, as cheaply and as often as possible.

This gave them the most chance of being rewarded for any sales that occurred. There are many masters of this in the display field but one that needs to be called out is Facebook with its post-impression tracking option. Facebook has a very strong media platform today so with that in mind, why does it need to be called out? Because Facebook does not allow most third party attribution companies to measure its post-impression activity currently.

I would encourage all large advertisers to tell Facebook spend is on the line if they do not allow third-party validation of their post-impression numbers.

So what’s to be done? Taking into account qualitative metrics such as viewability and dwell time is one approach. Be cautious of any media owner or tactic that has an average frequency of around close to one.

In some platforms you can analyse the incremental impact different tactics are having with unviewable ads taken out. Expose control tests can also be run and with that data a percentage of post-impression sales counted. Lastly bringing in learnings from econometrics or offline predictive modelling can be helpful.

Such controls are not easy, but they can be done. Consult an expert such as Canton to help because the impact is huge. Without doing this, long tail poor inventory is over valued compared with quality inventory.

This can and likely will include poor content you do not want to appear against unless controls are rigorously put in place. Furthermore, such techniques are how bad actors take money out of the advertising ecosystem.

3. Overvaluing retargeting

A last impression or last click model will favour retargeting over upper funnel channels and tactics. Retargeting is also a large part of the cookie bombing problem mentioned above.

A good rule of thumb is that between 60% and 90% of all retargeting sales attributed on a last impression or click model would have happened anyway because these are your most qualified audience.

This clearly aids Google and Facebook with their large retargeting programs. It also encourages money going to long tail publishers regardless of quality over and above premium content producers. Additionally, it encourages practices that annoy consumers such as high frequency and carrying on running ads after the user has bought something.

It is this behaviour that is so visible to users and causes much of the distrust of the industry.

Please do run retargeting, it’s a valuable channel to ensure your hot prospects convert. However, be aware that it numbers are not as good as they are first look. Also ensure that your retargeting is run with adequate controls over frequency, quality, viewability and most importantly consumer consent.

4. Counting brand search in your attribution

Brand search is the search generated from an advertiser’s brand terms. If you were to use Amazon as an advertiser, any search term containing the word Amazon such as “Amazon cheap TV” is brand search, whether it is paid or natural.

Paid brand search is valuable (in most cases), as it allows the control of landing pages and search copy and also protects against competitors. What it is not is a demand-generating channel. In short I would not even call it advertising (though it is marketing).

Every person who deliberately types your brand terms into a search engine already wants to go to your site, that is the user intention.

As someone that cares about digital advertising it is heart breaking to see quality publishers drive a user towards a purchaser, only to lose the credit because the user decided to navigate there using the Google search bar.

If someone just types the advertisers name into the browser, it will go via the search engine in most cases. It means the impact of all your other channels that are creating the demand in the first place have their value diluted unfairly.

So advertisers, please, take brand search out of your attribution funnel. By all means continue to execute it to protect your valuable search engine results pages. If finance need to see a report that shows the value of this, then give them every sale that other channels have not touched.

However please don’t hurt your efforts at upper funnel advertising by attributing their sales to brand search.

5. Comparing apples and oranges

The old truism is that there are lies, damned lies and statistics. In no sphere is this truer than media measurement, though we often call it by a different name, comparing apples and oranges.

In simple terms, not all channels and tactics can be compared on the same metrics. Its simply not comparing like for like. Some examples I have come across just make me shudder.

“Display results are not as good as search results”. Apples and oranges answer. Of course not, search includes brand search which as per the previous point is not comparable as the user already wants to go to the site. A better comparison is prospecting with generic search.

“Prospecting results are not as good as retargeting results”. Apples and oranges answer. Of course not, someone in your retargeting pool has already shown a user intention towards your brand. They are much more likely to buy anyway. If you can see the incremental results things should look different.

There are numerous other examples it would take too long to go through here. The point is that every channel has to be understood in the context of user intention and what we are trying to persuade the user to do.

The only truly comparable number would be marginal or incremental sales, the holy grail of media attribution. Incremental sales are something it is impossible to measure with 100% precision today but by following best practice and using available technology and with media understanding we can get close.

A better future

I am not saying that we will never be able to achieve perfect measurement and attribution however I hope and believe that if advertisers and media practitioners at least understand the concepts outlined above we can build a better future for advertising.

A better understanding of the value of advertising would be transformative for the industry. I would encourage all advertisers to invest a proportion of their media budget on attribution and invest in third party technology and independent expertise to take their business forward.

If they do it will lead to better returns for advertisers, a fairer media allocation for publishers and lower the distasteful elements in our industry.

At Canton if we can be famous for anything it would be being a part of fixing media measurement.