Predictive Marketing: 1.3 – Channel Attribution

Marketing Channel attribution and its impact on Predictions

Based on the type of industry, you must be using Marketing channels that are believed to be the most effective.

For E-Commerce or Retail based businesses, the most popular Marketing channels could be your company’s Website, Social Media Ads, Adwords, Ad Affiliates, Printed Catalogs, Email campaigns, Flyers etc.,

What is Channel attribution?

When you have multiple Marketing channels, you would attribute every sales transaction to a channel that influenced/caused it. This is needed for any basic tracking or measurement of channel effectiveness.

Why is it relevant for predictive marketing?

For predictive marketing, you might choose one or more channels to run your campaigns. The campaigns are delivered to the prospects that have the highest likelihood of customer conversion or to the customers that have the highest likelihood of repeat purchase. But, when a channel is picked up for prediction likelihood, your customer purchase history is the primary source for data mining.

If sales were not attributed to the correct channel, incorrect data set will be processed. Accurate channel attribution is one of the pre-requisites to keep in mind especially when you are considering any predictive marketing in the roadmap.

Let’s review the diagram and understand the key takeaways.

The first three columns from the left, Channel spend, CPA (Cost Per Acquisition) and channels are self explanatory.  

Let’s review the last two columns on the right, “unique customer buying journey” and “sales”. Every channel generates some sales that get attributed to an appropriate channel.

What is so complex there?

It wouldn’t be as complex and tricky, had your Channels and customer set are always mutually exclusive to other channels. The catch here is that, your prospects and customers get exposed to the campaign you have been running across multiple channels.  In addition, this exposure happens in a variety of sequences influencing his/her purchase. It is not necessarily a static sequence for a given Channel nor for a Customer and hence the complexity.

From your Customer’s perspective, your organization is all one big entity and if they are convinced about buying a product, a sale would happen on your end. But, to you as a Marketer with a responsibility and interest in improving the Channel effectiveness, you would want to drive precise data capture and related insights into what Channel influenced your Customer the most, for every sale generated.

So, in the diagram, it is showcasing an example path where it might take many paths from Customer’s perspective and your challenge is in attributing to the most accurate one based on your business rules.

Let’s jot down a few takeaways before we proceed further.

1. Channel attribution is a complex process when multiple channels are in play.
2. It is neither a linear path nor a mutually exclusive one.
3. While it might not be 100% no matter how many techniques you implement, there is still a sense of ensuring good accuracy.

How does inaccuracy in Channel attribution hurt my predictive campaigns?

This is the crux of our today’s discussion. When your sales are attributed inaccurately to another channel or falsely attributed to your channel, two cases arise.

• Case 1: You are overstating the Sales in your channel and ROI numbers are all impacted.
• Case 2: You are understating the Sales and ROI numbers.

You might wonder that, this is all still within the scope of Channel ROI and where exactly is it hurting predictions?

As we reviewed at the top, the data set for predictions will be filtered for all the sales that is attributed to a channel in consideration. In addition to your Customer purchase history, Prediction analysis requires the data about Ads placed, catalogs sent, Number of catalogs sent and the dates etc., related to the Channel in question, to derive the stats of  graduation path by a customer, average time it takes for someone to buy, average number of Catalogs to receive before they act etc., and many more. So, potentially, all the patterns reviewed by the Machine Learning algorithms point them to a new learning path.

Effect of overstatement on predictions: The algorithm is potentially losing the patterns that could have pointed them to show a Sale will not happen but your overstatement of Sales suggests that your Customer is buying in a different graduation path than it originally is. This will be abstracted from everyone’s vision due to the inaccurate channel attribution.

Effect of understatement on predictions: The algorithm is trying to find patterns that might strengthen the premise that, with all of the other pieces of information, a Sale should have happened but the attribution is happening to another Channel in error.

In summary, while Channel attribution cannot be 100% accurate,  it is very important to ensure for the correctness to the extent required, so the prediction algorithms learn the right lessons and right takeaways to give more accurate insights for better corrective or proactive actions.

Thank you for your attention and time. Please drop me a note if you want me to expand on any other topics. Thank you also in advance for all the shares, likes, comments to my posts. If you liked it, please share it with your network or leave a comment.