Customer Lifetime Value - Multidimensional approach

Customer relationship management has become very critical for all the companies and marketers are eager to find out loyal customers for increasing their sales base. But it is critical to note that all loyal customers may not necessarily be profitable. Customer lifetime value model is a framework which allows clear segmentation of the customer base to identify targeting segments which would return maximum profits on investment in marketing efforts. CLV is used to create comprehensive approach towards targeting customers and bridging the gap of knowledge towards retaining them. The churning customers are accurately identified and strategies of retention are broadly suggested as per the customer persona. This framework can be used as a suggestive for both up-sell and cross-sell. Customer acquisition and retention has always been an issue for the sellers and CLV can be used as a tool for resolution of these business problems as well as creating focused marketing campaigns for various customer segments as per their life cycle stage.

Another aspect worthy of notice is the investment done towards acquiring the customer. The CLV framework can be made more robust and sophisticated depending upon the availability of such information as it can be used to identify ROI, or the more sophisticated Value metrics. Analysts always search for multiple attributes of information which can be used to create a more robust approach of calculating CLV. Competitive information, Marketing costs, Brand health indices, NPS, Online purchase behaviors, various cost indices for retention, marketing and acquisition etc. are a few of such aspects of data. Another dimension which can be added to the analysis is around Recency, Frequency and Monetization of the purchases done by the customer. But the assumption with this analysis is that a higher valued customer tends to become even higher valued as he is likely to continue his purchase behavior. Customer feedback or touch point data could help verify this assumption for such metrics to alone define CLV. A multi attribute framework with sophisticated data mining algorithms can provide state of the art platforms for supporting strategic business decisions. The Global BI team along with Dell Global Analytics is working on creating aforementioned CLV models using Survival analysis for retention and Random Forests for Margin Index prediction.

Customer lifetime value - various approaches

Marketing problems can be analysed from various angles as there is no comprehensive solution to all the interlinked and complex questions that the business will pose for the analysts.
Strategic decision making is a very sophisticated job and requires complete focus on the validity and applicability of the "go to market" plan that the data driven approach will suggest.

Let us review some new business questions today:
"which customers are going to stay put with us", "how long will they stick to my business","how much profit can i make from such long standing customers"

Buying behaviour of an average customer is getting impacted by multiple factors, called drivers, like product quality, value of the deal, branding of the product, promotion efforts, competitive space in the market etc. Many times the companies launch price offs, combo deals etc to retain their customers. Especially in a B2B space, customer relationship management is very crucial. Loss of a single big fish may prove very costly and even result in convolution of the brand in the market.

Hence, Marketers want to predict the longevity of the customer as well as their value in terms of profit. Here Customer lifetime value analysis can play a prominent role in suggesting the targeting strategy to the sales and marketing teams.

Retention predictions are usually done using survival analysis whereas profit predictions can be done using various methods like regression, cart or random forests.

In the next post i will discuss these topics in detail. So stay tuned....

Uplift Modeling - The lamp in the dark

As I mentioned in my last post, Uplift modeling is a new age technique to help the marketing team save money, drive positive ROI and ensure successful campaigns by controlling the target audience or respondents of the campaigns.

Uplift model will plan strategy on the hold out recommendation and drive positive incremental response. This technique has a multi-aspect approach. The population base is split into a four blocker as shown in the picture:

The algorithm developed will try to split the data into four groups based on their response to the marketing activity. The effect of the treatment will be measured using this segmentation. These groups shall be individually classified by sophisticated data mining algorithms to form equations which could be thus used to create futuristic marketing strategy encompassing identification of the target audience, setting up of measurement framework i.e. the control group, optimization of the marketing budget by selection of target audience which would be comprising of the responders to the campaigns.

As seen in the picture above, marketing team would like to target persuadable or savable audience alone. Based on the price of each piece of marketing, budget optimization can be done. This will account to savings from implementation of the model.

Impact of multiple independent variables coming from purchase pattern behavior and other external data impacting the stickiness of the customer to the brand can be taken into account. In fact, the algorithm self corrects and every time new factors get incorporated, they get accounted as per their statistical significance.

The approach is unique as it combines Bayesian networks and Random forest for prediction. Both the techniques are unparalleled in accuracy and sophistication.

  

The inherent responders would be sieved out from the group of responders who converged due to the impact of the marketing activity. This approach would also help us to stop targeting the customer groups who respond negatively to any marketing activity. Also, this will minimize random targeting which could negatively impact the Dell brand value.

I shall talk in detail about both these techniques soon. So keep reading !!

Marketing Analytics - Uplift Modeling

In the world of marketing analytics, analysts are commonly surrounded by such questions as below: 

" How do I manage my campaigns? Whom to target? With which campaign should I target my audience? How should I optimize my marketing budget? "

All these questions are integral to any marketeer. Some do get nightmares when none of the campaigns seem to give any uplift, in fact the numbers are turning to red. The Green zone becomes a dream and Sales is pressurizing to help and all the blames of dropping sales are coming back to the marketing team.

Analysts give great amount of thought to these questions and dissect them into separate studies which collaboratively could provide relevant guidance on a subject so complex and dynamic. Any marketing activity doesn't happen in isolation, there are multitude of factors which play actively to make or mar the game.

Flop campaigns and negative returns are painful and Marketing teams turn to data and analytical support to understand actually where should they invest their budget so as to see a positive uplift. Hence, number crunching and algorithm development become so important for any organization who want to take informed decisions.

Uplift modeling is one such new age technique to help the marketeers save money, drive positive ROI and ensure successful campaigns by controlling the target audience or respondents of the campaigns.

Uplift model will plan strategy on the hold out recommendation and drive positive incremental response. This technique has a multi-aspect approach as depicted below:

I shall talk in detail about this in next post. Till then, happy number crunching :)