Cross-channel attribution: Getting the right media mix
During and following a recent Efficient Frontier webinar I hosted there were a lot of relevant questions. Marketers are inevitably wrestling with the problem of attaining the optimal media mix. Understanding what each channel’s value is and the optimal mix for the goals at hand adds an extra layer of complexity for already busy marketers. The days of the “last-click wins” mentality and one-off online strategies have drawn to a close. Attribution analysis enables a more accurate understanding of what leads to conversions, and these insights can lead to dramatic improvements in ROI. I’ve summed up some of the pervasive themes of the webinar and common questions amongst the group of leaders of how Efficient Frontier can aid in their marketing decisions:
Q: Tell me what attribution is exactly.
A. Attribution is a heuristic method that quantifies the influence of all media channels on conversions.
Q: This is for DR (direct-response) only leads? What about brands?
A: Our attribution methodology applies to both DR as well as brand-focused advertisers. However, the key difference in our method is that it provides a way to measure as well as optimize to an advertiser’s goals. While it is true that brand marketers are not directly focused on revenue metrics they do care about reach and engagement metrics. Further, in every organization brand marketers are held accountable to the amount of business they generate. The key to proper attribution for brand marketers involves three steps:
- Metric and objective selection, e.g. the goals you really care about. It could be total engagement on Facebook, or total website visits, or a combination of several metrics such as page views, engagements, email capture etc.
- Finding the true value of these metrics in relation to the business or revenue they generate.
- Applying attribution and optimization algorithms on top of these metrics.
Step 2 is tricky because the connection between online branding metrics and offline revenue generating activity is often tenuous, and the best one can typically do is apply econometric models. If this is not feasible then a blended metric can be created, which is a combination of each individual engagement metric.
Q: So it’s a click model? But most people don’t click on display. It does not seem to measure advertising it seems more of a click response optimizer.
A: This is not a click model. This is a model + algorithmic approach that measures the effect of every measurable interaction in the user’s conversion path. In display, view through impressions can be measured and attribution can be applied to impressions. In the webinar, I mentioned that the key concept to attribution is to measure the jump in conditional probability of conversion (or another goal) over a preceding sequence of actions. This applies to display view-throughs, too. One can set up controlled experiments to measure the value of display impressions to understand their value in the conversion path. Efficient Frontier’s attribution platform can help advertisers attribute and optimize to display impressions too.
Q: What is the interaction?
A: In the webinar, I mentioned that channels are constantly interacting with each other and that effect has to be captured. Example: Let’s say that the search budget is $ 1 MN and display is $ 2 MN and that the conversion rate on search clicks is very high. As a result, you decide to shift the budget from display to search where search budgets are now $2 MN and display is $ 1 MN. This might cause the conversion rates on search clicks to drop more than you expect because the display impressions that helped the search clicks before are being served a lot less than before. The search and display channels are interacting all the time and this cross elasticity has to be captured in a good model.
Q: Does Efficient Frontier have tagging that works across social and display?
A: Social impressions can’t be tagged or tracked due to Facebook’s policy, but we can track social clicks and display impressions and clicks.
Q: Isn’t Google Analytics “last click” attribution?
A: Yes, it is, but Google does provide first and last click reports. (http://www.google.com/ads/innovations/searchfunnels.html)
Q: What is the distribution of Efficient Frontier clients using each of the models?
A: We have a variety of advertisers using these models spanning, but not limited to: finance, auto, travel, retail and more.
Q: Does the idea of attribution models only apply to agencies that support multiple digital channels?
A. No, it also applies within one channel. An example: in search where non-branded keywords often help branded keywords.
Q: Does Efficient Frontier take into account each company’s overall strategy into this experimentation?
A. Yes, and this can be challenging when doing the attribution. For instance, if the attribution suggests that the budgets need to be dramatically changed then it might not be feasible to do so as it can put the line of business at risk. In such a case the spend recommendations coming from the attribution should be used directionally. For instance, if search is more efficient than display and the media mix recommendation shows that search budgets need to increase by 25% one can start with a 5% increase and observe the overall performance. If performance improves then more budget can be shifted to search. Still, these approaches can be challenging as the effect of slow budget shifts can be hard to measure due to performance volatility, seasonality, marketplace changes etc.
Q: How much experimentation is needed to find optimal point?
A. Theoretically, the experiment is always on. Optimization requires constant experimentation because things are always changing.
Q: Does the Efficient Frontier attribution system take into accounts both converted traffic and non-converted traffic?
A. Yes, as long as non-converted traffic is captured via one or more metrics.
Q: Do advertisers need to rely on clicks? How does Efficient Frontier track impressions versus clicks across channel?
A. Only display impressions can be tracked. Social and search impressions cannot be tracked due to the reasons mentioned above.
Q: What are the technology hurdles that need to be overcome to achieve the vision? Where are the capability gaps today?
A. First the platforms such as Facebook need to allow advertisers to track view through impressions. Second, mathematically there is no method that can accurately predict performance for the entire media mix. The methods are only locally accurate in their predictions and this necessitates an experimentation exploitation approach. I believe that, ultimately, attribution is not the solution for media mix modeling because it does not capture the interaction between channels very well. Attribution is a more effective solution for ad and conversion optimization.
Q: What are the limitations of only tracking digital touch points and not offline media points? Are we just making digital even more siloed in its accountability? Are there any tools that track online and offline touch points together?
A. The limitations are that it is hard to directly measure the online to offline effect and vice versa. There are some tools that track offline phone calls but the best one can do for TV is use panel data. Apart from that, one can build econometric models, but using these is not a scalable solution because it requires a lot of human input and the quality of the model depends strongly on the person who built it. It is also a look back into a snapshot of time in the past and may not reflect the dynamic advertising environment looking forward.
Q: What impact do you think Google’s move to secure searches by signed-in users will have on attribution modeling?
A. Little apart from paid searches because these can still be tracked. However, for organic searches some estimates say that up to 30% of traffic will not be accurately captured. I don’t think this is a big problem because not capturing the organic traffic has potential for misattributing the effect of organic searches. Advertisers have little immediate control over the position they show up for an organic search, so when it comes to media mix allocation little will change.
Q: How does Efficient Frontier track TV campaigns?
A. Efficient Frontier doesn’t, but we use indicator metrics provided by our advertisers such as spend, GRPS and TRPs to build econometric models.
Q: Is there a scalable solution that can track the display impressions in the cross channel tracking?
A. Efficient Frontier’s solution can track display impressions in cross-channel tracking.
Next Efficient Frontier Webinar: Tuesday, December 13, 2011 at 9 A.M. PST / 5 P.M. GMT
Making Facebook Marketing Work
How to Provide Value to Your Fans and Real Business Results to Your Brand
Join Efficient Frontier and featured guest Forrester Research Inc. as we discuss the findings in Forrester’s November 2011 report, “It’s Time to Make Facebook Marketing Work”, and provide you with specific next steps and concrete examples for successful Facebook marketing with tangible ROI.