Why You Shouldn't Do Attribution Modeling

As a digital/web analytics practitioner and marketing geek, I was excited to see Mike Pantoliano speak at Mozcon and learn something new about Google Analytics.
Mike's session essentially gave you the tools and showed you how to mimic Google Analytics Premium's attribution modeling feature. The title of the session was Attribution Modeling: Why You Must be Doing It and How to do It Easily with Google Analytics. You can see Mike's slides here.
So what is all this about? By default, Google Analytics attributes success to the marketing clickthrough. If a conversion has multiple marketing touches, only one channel gets credit.
But Mike showed how with even the free version of Google Analytics one can move from this:


To this, an attribution model that distributes credit among all the marketing clickthroughs:


And his method was exporting the multitouch attribution report in GA and processing it with an an excel file.


Mike's session was great. Even as an experienced GA user and GA qualified individual, I learned something new. But, I'm here to tell you why you probably shouldn't be doing attribution modeling (yet), and you can get more out of other analytics techniques before adding more attribution complexity.

Your conversions aren't created equal

One of the biggest issues with web analytics data is that it measures all conversions and revenue as equal.
Wait, you're thinking that revenue=money and we are all trying to get money, right? Wrong. Profit is what matters. For the most part, you don't see profit in your GA reports. Applying an attribution model is complex and should not just be just be applied to your revenue but should also take into account your ad spend and internal costs.
Optimizing for today's profit isn't the answer either. You should be optimizing for future growth and profitability.
Everyone understands web traffic quality through metrics like bounce rate and conversion rate, but few people talk about customer and conversion quality. Here are two examples, in ecommerce, new customer orders are often worth much more than repeat customer orders, and the sales from new customers are more likely to have been influenced by your marketing. For lead-gen it is even more obvious. Not all leads are created equal and qualified leads are worth much more than less qualified leads.
Even if two types of marketing are being attributed the same number of orders or same number of submissions, the marketing may result in very different impacts on future profit.

Your leads aren't all diamonds.

Instead of jumping into a complex attribution model, first segment your conversions and leads by quality and look for ways to combine transaction data with your web analytics data to forecast future revenue and profit.

Your marketing channel definitions are too simple

Google Analytics defines the direct, referral and organic search marketing channels (medium in GA terminology) out of the box. With AdWords auto-tagging and utm tracking codes, you can build out a bunch of marketing channel reporting fairly easily.
You can segment your marketing further though.

Navigational vs competitive search

The simplistic view is that search is two channels, paid and organic. The truth is that user intent matters as much as whether a search click was paid or organic.

Start by breaking your paid and natural search channels into navigational and competitive searches. For many brands, revenue coming through navigational searches is not incremental and has more in common with direct load traffic than competitive search traffic. Unfortunately “not provided”, makes this sort of analysis even more complex.

Marketing emails vs triggered email

Make sure your transactional and triggered campaigns are measured separately from your marketing campaigns.
Your triggered campaigns are often very effective and have little to no on-going cost compared to the effort you put into designing and sending marketing email campaigns.

Discovery affiliates vs landing page affiliates

Affiliates are often overcredited with last-click attribution. Make sure you understand the differences between affiliates that help clients discover your brand and products and affiliates that sneak into the sales funnel closer to the transaction.
Before you apply an advanced attribution model, make sure you are segmenting and analyzing your channels based on user intent and not just marketing vehicle.

You are link and click obsessed

If you are a marketer that comes from a search background this is almost always true. You are obsessed with the link and the click. The search engines themselves have provided great tools for tracking clicks and attributing success to them.
But the truth is not all clicks on your marketing links are driving incremental website visits. Especially for bigger brands, many of those people would have visited your website anyway without your marketing link.
Channels that are creative heavy are often under-attributed as Google Analytics can't give any credit to view throughs, even with advanced attribution models. This includes email, display, social, and of course offline marketing where clients are often seeing your marketing messages and can't click on anything.
Before using an advanced attribution model, consider whether there are other ways to measure success of channels. With email and display, consider doing a hold out test to measure the true impact of the marketing.

Sometimes a client survey or a how did you hear about us field in a form is a far better attribution methodology than any modeling done with Google Analytics data.

You only want to make your channel look better

There I said it. You don't actually care about whether your channel is getting under-credited or over-credited with a simple attribution model- you just want to find a way to amplify your success sell it to management. Everyone can't just pick the biggest number they can have a tool create and say that is their impact on their business.

If you are going to make your channel look better, another channel has to look worse. To actually understand which attribution model is closest to the truth, and which marketing touches are having the most impact in the sales cycle, you have to run a big statistical regression on the impact of the marketing touches. Are you planning on doing this? Probably not.
Before doing advanced attribution modeling, start thinking about your business objectively and do real analytics with a simple attribution model.

Conclusion

All that said, used properly, the Google Analytics multi-touch features and attribution modeling can provide great insights. But, any attribution model you put in place will have its flaws and just be one version of the truth. I believe most marketers can find easier ways to optimize and refine their strategy by using simpler segmentation, analysis and combining their web data with transaction data, before moving onto a more complex attribution model.


Reference :- http://www.seomoz.org/ugc/why-you-shouldnt-do-attribution-modeling

EBriks Infotech :- SEO Company India