eCommerce Verticals

eCommece Verticals

CIO: what's in it

Objective:

His objective is to increase data/information fluidity and accuratie in the company.  By fluidity mean the right information to the right people:

Right information: not just content but grain, frequency, medium of provision

Right people:

• ·         Data analyst: role is to pre-chew data and provide added value analysis to decision makers. Added value analysis is not just what but why, so what and what to do with it
• ·         Data scientist: role is to explore the data to find patterns using analytical tools (functions/formulas). Will perform both descriptive analysis (look at past events to perform segmentation, correlation, clustering, multi variate analysis) or predictive (build models to predict the future with regression, decision tree)
• ·         Business audience:
• o   Product managers
• o   CRM
• o   Marketing team
• o   Finance
• o   Ops
• ·         Executive: need summarize yesterday data to track health of business vs projections/target and LY + create finance model for new opportunities

Role:

CIO role is to set the strategy appropriate to the maturity of the company, make it evolve and grow with the company.Strategy is around not just tools but as well teams and skill sets and infrastructure.

Why is it important to have a CIO?

Need to have a consistent and centralize data roadmap, enforce one data truth. If it is spread leads to growing pain
Once company reach a certain size then analysts and data scientists can be decentralize to business. CIO office is then just a service provider and is pure IT.

Team:

• DWH: nucleus  of the data, where it is stored. Need a data acquisition team + Architect + Administrators Presentation team: In charge of distributing the data to the audience whether in the form of report dashboards or data sets
• Data analyst: primary consumer, provide the added value on top of data
• Data scientists

What would be the right organization for 150 people company today?

Cheap efficient fast turn-around.

• Cheap: Plethora of free tools with Adoop/Hive/R and mySQL, or tools included in corporate licenses (Microsoft SQL servers with Analysis studio, PowerPivot etc) or cheap solution
• Efficient: don’t address the enterprise approach in its entirety: create a robust, clean and agile DWH and plug on top departemental solution like Tableau, Quilk.
• Fast turn-around: easy to learn, easy to maintain no deployment.

---
Technical
Architecture
Culture

Stored in one location, is clean, accurate and available on time.
Distribution of the data

CRM Marketing Analytics

CRM analytics can provide :

·      Customer segmentation groupings (for example, at its simplest, dividing customers into those most and least likely to repurchase a product);

·      Profitability analysis (which customers lead to the most profit over time);

·      personalization (the ability to market to individual customers based on the data collected about them);

·      Event monitoring (for example, when a customer reaches a certain dollar volume of purchases);

·      What-if scenarios (how likely is a customer or customer category that bought one product to buy a similar one);

·      And predictive modeling (for example, comparing various product development plans in terms of likely future success given the customer knowledge base).

Web Site personalization: There are a number of personalization software products available, including those from Broadvision, ResponseLogic, and Autonomy.

In addition to use of the cookie, the technologies behind personalization include:

• Collaborative filtering, in which a filter is applied to information from different sites to select relevant data that may apply to the specific e-commerce experience of a customer or specific group of customers
• User profiling, using data collected from a number of different sites, which can result in the creation a personalized Web page before the user has been formally
• Data analysis tools used to predict likely future interactions

Marketing Lift: Measuring Campaign effectiveness

How much more products or consumers are converting due to the marketing/advertising efforts.
A basic test/control methodology should be used to find out the lift in conversion rate due to advertising.

In simple words you found out the purchase intent without the advertising (Control group) and then with the advertising (Test Group). The % difference between the two provides the lift.

Lift is calculated as (test-control)/control

Agile DWH vocabulary

http://www.agiledata.org/essays/dataWarehousingBestPractices.html

Iteration 0 Start with high level vision of the architecture
For each of the iteration of the architecture/delivery do some model storming: sketch one of the requirements from output up stream to data model,transformation and sourcing.
Test a the skeleton of the architecture to prove that the architecture will work. Go though accessing sources, that the ETL strategy works, and that the db regression testing works.
Prioritize the requirements, the one of the end users. Satisfy each of them and stitch the final solution together using regression testing.

A development approach is Test Driven Development (TDD): Break down the delivery of the iteration into sub-part, design a test for each of the sub-part, and develop the code to fulfill the test of the sub-part, then put together the code of each of the sub-part, integrate and perform regression testing.

Vocabulary:
refactoring
Regression testing
TDD