Getting to Grips With Attribution and Channel ROI
During summer, it’s a known fact that ice cream sales increase. It’s also known that shark attacks increase in the same period. Therefore the solution to reduce shark attacks in summer is to ban ice cream sales, right?
I’m sure you can see the absurdity of the proposition – while there is correlation (i.e. a general relationship when one variable goes up, so does the other), we know full well there is no causation (i.e. ice cream sales do not cause shark attacks.) However, in online marketing, the correlation and causation for efforts resulting in conversion – whether it is online sales, lead generation, sign ups - are far less clear.
As digital spend rapidly increases in South East Asia and in particular a rise in local e-commerce, marketers will become increasingly focused on the effectiveness of their digital channels to demonstrate return on investment. As fragmentation in channels and sub channels takes place, marketers must understand what really contributes to conversion actions.
What’s the Problem With “Last-Click Attribution”?
Let’s take a look at the following simple example:
A user visited your website three weeks ago through a banner ad campaign, but they left your website without buying anything.
One week later, that same user came to your website by searching for your company name, clicked on an AdWords ad, and bought a $500 gold watch.
The most simplistic model is “last-click attribution”: the channel that delivered the conversion action gets 100 percent of the credit. So in the above example, it was the AdWords campaign that drove the sale of the $500 gold watch.
If you were to follow pure last-click attribution you reduce banner ad spend, whilst increasing AdWords spend. However, unlike ice creams to shark attacks, there was causation: if the user had not clicked the banner ad in the first interaction, there might never have been the sale.
Thus, by following last click attribution and removing the banner spend is actually detrimental to campaign performance in this example.
Now let’s expand the customer journey to include email newsletters, exposures to banners, organic search, mobile site visits, an app, affiliates, a video channel, and of course social media interactions – and you start to see how complex the problem becomes. And just to throw a spanner in the works, the user then converts in offline channels!
Say Hello to Fractional Attribution and Assisted Conversions
In Q4 of 2011, Google Analytics updated its conversion reporting section to now include an “Assisted Conversions” report, which has popularized the notion of fractional attribution: credit for conversions should be distributed across assisting channels, not just the last channel (or even the first channel.)
When a conversion occurs there will be up to three types of interactions:
1. First interaction: This triggers the conversion process
2. Last interaction: This represents the channel that drove conversion to a close
3. Assisting interactions: Any channels in between the first and last interaction
Fractional attribution attempts to weight different channels (e.g. opening an email is worth more than being exposed to a banner ad) and weight more recent actions (i.e. more recent and closer to last interactions are worth more) into the equation. The models are still fairly arbitrary in the absence of extensive statistical testing, which is unlikely and impractical in all but the highest volume transactional sites that run highly multi-variate tested campaigns. Practically speaking, fractional attribution applied with some basic testing and understanding of the approximations is still better model than last click attribution. You can typically receive fractional attribution reports from reporting packages like
Adobe Omniture and Adometry (or ask your agency.)
Google Analytics’ Assisted Conversions attempts to provide some indication of how often, the value and whether the assisting channels are better at initiating traffic or closing conversions – because ultimately you need both types of actions occurring. For reference, read the Google Analytics’ “Analyzing Channel Contribution” summary.
It’s not intended to be the ultimate in statistical reliability since it can only track where GA tracking code is implemented, but makes a reasonably elegant attempt at guiding marketers and agencies on budget allocations. It does not tell you how much budget to allocate, but it does help to define the role for each channel and the potential value that it delivers. For the sake of clients that are not selling online, marketers need to define some value for a conversion action such as a newsletter sign up or an enquiry.
The Future of Attribution and ROI
With the expansion of digital touch points, budgets are at increasing risk of being spread too thinly and therefore ineffectively. Better attribution analysis helps marketers focus budgets towards channels that deliver desired results and identifiable ROI.
Additionally, technologies like mobile enabled NFC and in store digital loyalty programs will help to link offline behavior with online profiles, bringing us one step closer to marketing analytics nirvana.
And John Wannamakers’ quote, “Half the money I spend on advertising is wasted; the trouble is I don’t know which half” may one day be reduced to one third or a quarter. And no unnecessary banning of ice cream sales will be needed!