Archive for July, 2014

Jul 22 2014

Analytics for eTail

We are glad to share that our team successfully organized the second meetup titled “Analytics for eTail” at IIT Madras Research Park. The meetup serves as a platform for the business firms to understand the relevance of analytics and how it can improve their day-to-day business operations. The meetup is alive at http://www.meetup.com/Analytics-for-Business/about/ .

Ms. Parvathy Sarath, Director of Aaum Research and Analytics introduced various eTail topics to the business firms. WP_20140721_014The various eTail topics discussed are

  • Monitoring campaign Performance
  • Price sensitivity analysis
  • Optimization techniques
  • Social media analysis
  • AB testing
  • Recommendation
  • Market mix modelling
  • Sales attribution analysis
  • Heat map generation techniques
  • Loyalty measurement and analysis and
  • Dynamic Pricing

Meetup 2 - All

The meet up spanned for two hours. The participants showed lot of interest in understanding the concepts and to see how those techniques could be adopted for their business. Click here to view the presentation delivered on our geniSIGHTS solution. The participants showed keen interest in attending future Meetups conducted by Aaum.

 

Jul 16 2014

The vital click!

Attribution modelling is a widely used tool in businesses that helps the marketers to understand the impact of various marketing channels have on the ROI of their businesses. The insights from this simple exercise allows the marketers to track and analyze the multiplied touch points in sales and their impact on the conversion value. This in turn helps the marketers in effective credit allocation of their marketing budget on the various credit channels.

With the behemoth amount of data, digital marketers are now shifting their focus on attribution for the purpose of increasing their conversion rates. Low online conversion is generally overcome by optimized customer service by personalizing customer experience management, web analytics and enhancing the use of feedback. Conversion path analysis is done to convert the normal website visitor to a paying customer or even a subscription to a newsletter may be considered as a conversion. A simple layout of a path conversion is shown as follow:

conversionpath

When the website visitor is on the landing page there is a requirement to enhance his user experience by providing relevant information about the products and by establishing an emotional connection to the brand. To avoid distraction, it is required to provide focused content and targeted conversation and providing highly flexible pages for easy operation by the users.

Conversions generally involve more than one channel and conversions generally travel down the multi-channel funnel.

The channels could be:

  1. Paid-search

  2. Organic-search

  3. Social media

  4. Referrals

  5. E-mail

  6. Direct

For example, one may first read about your product on a blog-post. Then, he may see a display-ad. Later, he may read a review on some website. Curious about your product, he will see your PPC ad. After that, he visits a product comparison website. He then clicks on an organic search result which then reaffirms a customer to buy your product and he finally purchases your product.

In the example, we see that a ‘buyer persona’ visits various channels and finally decides whether he wants to purchase a product or not. A marketer would now want to assign credit to the different channels which assisted directly or indirectly in the conversion process. This set of rules that govern the assignment of credit to the various channels is known as attribution. For this process there are various attribution models available. These are:

  1. Offline-Online attribution model: This model determines the impact that the digital marketing channels have on the offline marketing channels and vice-versa. This model understands the influence that digital marketing channels have on the offline marketing channels and how to assign credit to them.

  2. Multi-Device attribution model: This model determines the impact that various devices (Laptops, Desktop, Mobiles, Tablets, etc.) have on conversions and how credit is to be assigned for various devices.

  3. Multi-Channel attribution model: This model is the most common model that is used in the industry. It studies the effect that various digital marketing channels have on the other for conversions and how the credit is to be assigned for the various channels.

In the case study available here we have tracked and analyzed the various metrics driving to multi-channel attribution modelling that is commonly used in the E-Commerce market. The model also extends further to showcase a data driven hybrid approach to attribution modelling that combines the intelligence of more than one attribution model based on the customer profile to channelize the credit assigned effectively. The outcome of such an exercise would be similar to a plot below which depicts the conversion value attributed from each channel which is used in your conversion path. 

hybrid_attribution

The assignment of credit is essential to understand which channel plays a more important role in assisting the conversions. We can determine on which channel we would want to spend more money and time. Thus, at a minimal cost we would be able to improve the efficacy of conversion rate. Thus the attribution model is becoming an integral part in the lives of digital marketers.

Jul 7 2014

Validating “income levels” based on asset based “wealth indicators”

The income variables received from surveys will typically shows a relatively very high proportion of abnormal values due to insertion of false information, substantially reducing the number of valid cases in multivariate statistical analyses, etc. Such a discrepancyon a key predictor variable like income variable is undesirableas it alters the modeling analysis, insights. In this blog, we shall focus on an effective alternative mechanism to qualify “the wealthiness” of the customers based on their assets.

The survey questionnaires can be effectively constructed to collect information on household assets with the aim of obtaining more precise measures of economic well-being on the largest possible proportion of respondents. Indeed, the non-response or fraud rate associated with the household asset items in the application is much lower than the one for the income variable.

The estimation of relative wealth using Principal Component Analysis (PCA) is based on the first principal component. Formally, the wealth index for household is the linear combination, the first principal component variable across households or individuals has a mean of zero and a variance of λ , which corresponds to the largest eigenvalue of the correlation matrix of x .

 

The first principal component y yields a wealth index that assigns a larger weight to assets that vary the most across households so that an asset found in all households is given a weight of zero . The first principal component or wealth index can take positive as well as negative values. PCA analysis was performed on the asset information collected from the respondents. Each applicant was scored from 1 to 5 based on the computed wealth indices. The score 1 denotes the bottom 20 % or otherwise the poorest of our applicants. Similarly the top 20 % were denoted by the score 5 who were our richest applicants. use of PCA for estimating wealth levels using asset indicators to replace income or consumption data. A relative wealth index was computed using the methodology described based on the following items asked in the survey:

Wealth Index

The above exhibit is the normalized representation of the data. The analysts can now calculate the calculate principal component loading value that can be used as effective wealth indicators. The below exhibit qualifies the effectiveness of these indicators on the income level. The plot showcases the wealth index computed for the individuals who have undergone the survey at household level. Based on the wealth indicator computed from the asset analysis of the individuals, they have been scored and ranked as “Poorest”,”Poor”,”Average”, “Good” and “Richest”.Do note that the average income level rises as we scale to higher wealth indicator groups. This technique not only validates the income levels but also provides an alternative powerful variable for statistical analysis.

Wealth Index District Wise

Click here to view a case study demonstration on Validating “income levels” based on asset based “wealth indicators”.