Opinion

The attribution elephant in the corner

ben sharp

In this guest post Ben Sharp argues the industry needs to make a real effort to nail down the issue of attribution to prove what is really working for them.

There’s an elephant sitting in the corner. His name is Attribution. We all know he’s there and we all know we need to do something about it, but at the end of the day, it’s just too hard.

The IAB Australia recently did a survey called The State of Industry: Marketers & Advertising Technology. I sit on the Ad Tech Advisory board and was involved in helping drive this survey. The aim was, as leaders in the advertising technology space, to understand what our customers (both brand direct and agency) understood about our space, what tools they were using and what they wanted to learn more about.

Attribution was the stand out topic that marketers, agencies and even ad tech vendors wanted to learn more about. Only 57% of the people surveyed were using attribution technology. This is in line with AdRoll’s own research from October 2014 where nearly one in three marketers didn’t know enough to effectively use an attribution model.

Even at the IAB Leadership Summit where I sat on the panel that discussed the results of the Marketers and Advertising Technology survey, there was no discussion about attribution despite a clear signal that we, as an industry, are crying out for more education in this area. So why is attribution so scary?

Online advertising bought a new level of measurement and accountability. For the first time, you could track exactly who saw your ad, who clicked on it and who later went on to convert. For the first time, marketers had a real opportunity to understand what was actually working, what wasn’t and make decisions based on those data points.

Attribution modelling allows marketers to set up rules so that different marketing touch points are assigned credit for getting a customer to make a purchase. With it they can optimise their marketing spend towards the highest performing channels. It’s important because everyone wants to make a sale as quickly and efficiently as possible. Attribution helps marketers work out how to make that happen.

But there is no hard and fast rule about what is the best way to assign that credit, which is what makes attribution hard. For many marketers, the simplest attribution method is last click or last touch attribution. This is where 100% of credit for a sale is assigned to the last marketing channel that brought the customer back to the site when they made their purchase.

This model favours direct response channels like search or retargeting. Given that we know a customer’s purchase journey looks more like a toddler running around than a linear path these days, it would be dangerous to think that you should put all your budget towards only those two channels.

In fact, a survey on attribution conducted by Econsultancy and Google as early as 2013 showed that 69% of marketers thought that last click attribution was an ineffective attribution model. Most marketers accept that even if someone didn’t click on the ad, seeing the ad had an effect and therefore, should receive some credit (remember 8% of internet users account for 85% of clicks according to ComScore). That’s not a new concept and marketers still accept television, print and outdoor advertising has impact even if it can’t be measured as precisely as digital.

In recent years we’ve seen more complicated multi-touch attribution emerge like view-through models (assigning credit to ads viewed), time decay (assigning a decreasing credit amount to marketing channels over time) and position-based (assigned credit based on where a channel was in the consumer journey – usually higher towards the front and end of the journey). These methods are hard to implement but provide a more accurate picture of what the true impact of your advertising is.

Choosing an attribution model is a very personal choice and there are a number of factors that go into it. If nothing else, you should always ask yourself two main questions before choosing an attribution model:

  • On average, how long does your customer’s journey to purchase last? This will effect how long you want to set your attribution window. If the customer is buying a small ticket item – say a pair of shoes – then you want to make your attribution window small, a number of hours or days. But, if you have a larger ticket item like a car, house or mortgage then you want to set an attribution window that reflects the longer consideration period, a number of weeks or months.

  • On average, how many touch points does your customer have before they purchase? Again, this will affect the model that you chose. If you know that your average customer usually only has two touch points with your business before they make a purchase, a simpler time decay multi-touch model might be the way to go. If you know they usually have five or six interactions with your brand, you might want to consider something more position-based.

Finally, you do need to keep testing. Attribution models are constantly changing and marketers need to be prepared to change, develop and question the model all the time.

Attribution is tricky because there is no simple answer.  Even after coming to terms with online attribution, there still isn’t a definitive way to incorporate online and offline channels into the same attribution model.

There are a thousand little elements that go into the decision to make a purchase (Megan Brownlow from PwC’s presentation gave an excellent presentation on this at IAB Leadership Summit — note: iab membership required to view) and it’s impossible to know what was the actual tipping point that led to a purchase. At the end of the day, there is still part of marketing that is the art of influence.

  • Ben Sharp is managing director ANZ at AdRoll
ADVERTISEMENT

Get the latest media and marketing industry news (and views) direct to your inbox.

Sign up to the free Mumbrella newsletter now.

 

SUBSCRIBE

Sign up to our free daily update to get the latest in media and marketing.