Horizontal vs. Vertical Analysis

Horizontal analysis and Vertical analysis

One is analyzing all the customers in the population 'at a given time in their lifecycle'
Other is analyzing all the customers in the population 'as-is'

The WFR Problem India vs. China
We ran into an interesting problem last week. The web failure rate or WFR (Refer for definition & meaning: WFR Meaning) of India was 3x that of China. Immediately the first thought is 'Is China support site better than India support site'?
Other hypotheses that come top of mind - 'China has local language support in their support site', 'Chinese customers are probably more net savvy'........
Or maybe 'we need to get more data :) and drill down to an LOB level'

Why we would have never found an answer
This was a classic example where no amount of analysis/drill-downs would get us to a solution. What we were looking at was essentially was an analysis 'as-is' level - at an 'aggregated' level.

Please note: all numbers are dummy and not real

Getting everybody to the same point
Analyzed the customers into 2 buckets:
Once that had a problem: 'Contacted Customers' or 'Contacted ASUs'
The full bucket: 'All Customers' or 'All ASUs'

The clue was that web failures/Contacted Customers was same in India and China.



whereas the web failures/all active customers was comparable


This pointed to higher proportion of installed base in India contacting


Looked at the systems by their age and found the culprit - India has a higher proportion of newer systems than China. This results in Indian customers having a higher WFR because newer systems tend to have a larger # of problems in the first half of their lifecycle



An excellent example of why it is important to look at horizontal analyses getting everyone to the same point

Data, Information & Knowledge

Data -> information -> Knowledge. Which is the supposed value chain.

A close look at these terms:

Data is a set of un-analyzed observations; It is 'record of something happened'
Examples:
Observations on visits from visitors with and without coupons to a shopping cart
Observations on calls from customers with issues
Observations on sales by record at Reliance Digital

Information is data with a 'purpose' and 'meaning' - the purpose & meaning coming from 'summarization' and a 'context'; It is 'How you know it happened & What Happened'
Examples:
Visits from email tend to buy less often than visits from affiliates
Patients with metabolic syndrome are more likely to have a heart attack than patients without

Here each of the examples is summarizing data with a purpose - and the context is important. For a web analytics person the second information above is alien as he does not understand what 'metabolic syndrome' is and for a doctor, the first statement is alien as a doctor cannot understand what 'affiliate' is :). So some knowledge is necessary to qualify information
Information cannot inform in the absence of knowledge

Knowledge is a set of recipes and contexts in which the recipes become effective - It is what is possible to do, when it is possible to do, to achieve what is it necessary to do; It is 'what to do about Information'
Example:
Because visits from email tend to buy less often, we need to focus more on affiliates
Because metabolic syndrome is more likely to result in heart attack, these patients must be given extra care