IBM Smarter Commerce 2013 : Rapid Fire Analytics Attribution Strategy

campaign attribution
Facebooktwittergoogle_pluspinterestlinkedinmail

Last week, we presented our analytics attribution journey at the IBM Smarter Commerce conference, and it was very well received. In this post, I would like to discuss the presentation in detail along with the link to the actual presentation.

 

Objective

Understanding the attribution and contribution of marketing channels has always been one of my favorite analytics challenge. Solving for attribution can not only increase the changes of success for marketing progress it also helps answer the following questions –

What channel influences conversions more directly than the other?

How to divide the credit between marketing channels?

What channel combination drive the highest lift?

How can I strike the right balance between competing marketing objectives (branding vs revenue vs ROI)?

How am I performing against my forecast and what do I need to change to stay on track?

What is the optimal marketing mix to drive improve results?

If you are asking these questions, then you are on the right track. Most analysts think that solving for one of these questions will help them solve the attribution problem. This is not true because attribution is not a destination, it is a journey. You pick one attribution project, solve it and move to the next one. As you progress up the attribution ladder, your latest challenge becomes more complex and requires sophisticated tools and techniques.

A simple last click vs first click attribution can be based on marketing touch points credit allocation. However, when you want to derive a statistical inference from the same data then you will need statistical analysis tools such as SPSS.

Let’s walk down the attribution journey we have experienced at Rackspace.

Rackspace is a leader in hybrid hosting and cloud computing. We have 10 data centers across the globe serving 200,000+ customers. At Rackspace, we acquire customers through chat, phone calls and online shopping cart.

Phase 1 – Attribution by integration

Solving for chat based leads attribution was our first challenge, mainly because of disconnect between the CRM and web analytics. When a visitor initiates a chat on Rackspace.com the chat/sales rep takes their information and enters into the SalesForce CRM manually. Due to manual data transfer, we lacked insights on what marketing channels are bringing these leads.

We worked with IBM professional services team and integrated SalesForce with IBM Coremetrics. The integrate tied the web activity with the CRM data with the help of a unique identifier cookie. With the help of this integration, we now have insights on the marketing channels that drive high-quality leads.

Our shopping cart presented us with another challenge. Most of the cloud based products we sell through our shopping cart are utility based pay per use products. The initial product purchase holds a zero dollar value, and it is impossible to calculate marketing ROI by using the product sale value.

To solve this problem, we used IBM multichannel import functionality to import the internal billing data into IBM Analytics. We automated the import process using regular data transfer between systems. The multichannel import now allows us to calculate true ROI for our cloud marketing programs.

Phase 2 – Attribution based budgeting

Deciding the marketing budget has traditionally been a non programmatic process. Marketers rely on some historical media mix to decide on the marketing budget for the month or the quarter. It is more of a gut based exercise.

Interestingly, as the complexity of data increases the traditional media models are becoming obsolete. On and offsite attribution along with the bid social data throws off the unidirectional attribution. There is a huge need for bidirectional model.

We used IBM Lifecycle analytics to build a model to solve for attribution based budgeting. Lifecycle gives you insights on the marketing channel breakdown for the orders generated in the first and subsequent visits. We leveraged this data and built a model (we call it monthalizer) to forecast the orders in advanced. Think of the model as a control panel where you are controlling channel levers to forecast the orders generated 4 to 5 weeks in advanced.

The monthalizer model has given us an upper hand when negotiating budget with our finance. We are no longer at the receiving end of the financial insights, and we participate in budget discussions as a team.

Phase 1 & 2 has helped us solve the influencer and credit allocation questions.

Phase 3 – Statistical Attribution

Popular attribution models are based on one or more touch points (first, or last or latent click) and they don’t tell the whole story. Display could get more credit than paid search just because it was the first channel. Similarly, paid search could get more credit even though referring sites closed the deal.

Statistical attribution provides a more precise campaign value ruling out the equal credit anomalies.

When you look at the channel individually then SIRA (Statistically inferred realtime attribution) validates the marketing touch points. Slide 25 is where the actionable insights pop up. According to SIRA, in 208 cases referring sites is the winning offer even though last touch puts more weight on natural search (SEO). SIRA also puts more weight on natural search in 26 cases over email marketing, which is the last touch point.

Statistical attribution helps in understanding true cross-channel attribution and prevents errors. This insight has helped us justify our marketing investing in organic activities.

Phase 3 has helped us answer the marketing channel mix question with a statistical confidence.

Phase 4 – Data Driven Marketing Planning

What we have learnt so far is determining the value of marketing touch works in hand with optimizing the marketing sequence. The next step is to automate the process, so we can recommend accurate future actions consistently.

IBM is developing capabilities to allow us to use the attribution modeler data as an input to Budget Optimizer. Budget optimizer will feed the optimal marketing mix into the Performance Forecaster which in turn feeds the marketing scenario planner. It completes the full circle of attribution and automates the entire process of delivering optimial business results.

I would like to summarize the post by reiterating that analytics attribution is a journey. Take one project to completion and move to the next one. Your new projects will be more complex than the previous ones and will need better tools, techniques and partnerships with analytics vendor, internal subject matter experts (seo program managers, sem analyst) and data scientists.

Leave a Reply

Be the First to Comment!

Notify of
avatar

wpDiscuz