Data Science for e-commerce

Veldkant 33A, Kontich ● [email protected] ● www.infofarm.be

Data Science Company

DataScience for e-commerce

Infofarm - Seminar25/11/2014


Agenda

• About us

• What is Data Science?e-commerce vs Data Science vs BigData

• Example Data Science applications in e-commerce

some inspiration to see your opportunities…

• Applying Data Science

how to get started with all this?

Veldkant 33A, Kontich ● [email protected] ● www.infofarm.beData Science Company

About us


Speakers

• Niels Trescinskie-commerce Consultant

– Fenego (Intershop)

– Elision (Hybris)

• Günther Van RoeyTechnical (IT) Consultant

– InfoFarm (BigData & Data Science)

– XT-i (software development and integration)

– PHPro (website development)


InfoFarm - Team

• Mixed skills team– 2 Data Scientist

• Mathematics

• Statistics

– 4 BigData Consultants

– 1 Infra specialist

– n Cronos colleagues

with various background

• Certifications– CCDH - Cloudera Certified Hadoop Developer

– CCAD - Cloudera Certified Hadoop Administrator

– OCJP – Oracle Certified Java Programmer


InfoFarm + Fenego & Elision – e-commerce!

Highly focused one-commerce

Business Knowledge

Highly focused on Data Science and

Big Data

Technical Knowledge


Introduction: what is Data Science?


What is data science?

• Data Scientist: “A person who is better at statistics than

any software engineer and better at software

engineering than any statistician”

- Josh Wills

• “Getting meaning from data”

Finding patterns (data mining)

• Complementing business

knowledge with figures


Data Science & Big Data

• Relevance for e-commerce - use data to:

– Increment conversion

– Increment operational efficiency

– Understand your customers’ needs

– Make better offers

– Make better recommendations

– …

• Many successful online business thank their position to

smart data usage:

– Google was the first search engine that didn’t index by keyword

– Amazon is the e-commerce leader thanks to BigData

– NetFlix is a world leader in personalized recommendations



• Most of us don’t run a business like the ones referred to in stereotypical Big Data cases

• Big Data does not necessarily means or requires much data

• Data Science is very affordable to companies of all sizes

• Typical Data Science projects are 10’s of man-days of work



• Non-structured data: weblogs, social media content, …

• Secondary use of data sources is the key

– eg: Weblogs

• Are there to log webserver activity

• But can also tell you how people find, compare and choose products!

– eg: ERP / Cash register software

• Prints bills

• But can also tell you what products are typically bought together in a shop

• Many data is present, valuable information is hidden in it!


Topics not covered in this seminar

• Very interesting topics that we will gladly

elaborate upon another time:

– Statistical Tools (R, SPSS, …)

– Mathematical models

– Machine Learning Techniques (Clustering, Classification, …)

– BigData Tools & Platforms (Hadoop, Spark, …)

– Data processing tools (Pig, Hive, …)


Example Data Science applications:

#1: Recommendations


Recommendations – Why? How?

– Why?• Attempt to cross-sell or up-sell

• Provide customers with alternatives that might please them even more

– Traditional approach• No recommendations at all

• Products in the same category

• Manually managed cross-selling opportunities per product

– Why are these approaches fundamentally flawed?• They all start from the seller perspective, not the customer!

• “We know what you should be buying”

• Manual recommendations are too costly and time-consuming to

maintain – even impossible with large catalogs


Recommendations

– Online vs Offline

• Main focus on online, but why?

• Who knows best what products to recommend?

• Learn from your data, don’t take decisions based on a feeling.

– Time based recommendations

• Recommend or cross sell different products depending on

– season?

– holiday?

– weather?

– Customer based recommendations

• Learn from your customers and their past.

• Android vs iOS smartphones.


No product recommendations

at all(Link to category

without match with specific product)

Which roller would be appropriate?

No primer + paint combo?

Recommendations – Traditional approach

Showing (too)similar products?

No color alternatives?No glossy/matte

alternatives?


Recommendations – what does Amazon do?

Cross-selling as realized with other (similar?) customers

Starts from customer point of view!

Recommendations based on perceived customer journeys

Re-use the product comparisons that

previous customers did!

DATA DRIVEN!


Recommendations – Other ideas

• Data Science ideas

– “x % of the people who looked at this item eventually bought product X or Y”

– Get cross-selling information from ERP in the physical shops and let this feed the

webshop recommendations!

– Similar product in different price ranges

(“best-buy alternative”, “deluxe alternative”)

– ...

• This is very achievable for a webshop of any size

– Just generate ideas, and test to see what actually increases sales!

• Secondary use of various kinds of non-structured data = BigData !

– Weblogs of e-commerce site (use to deduct customer journeys)

– ERP info with bills and/or invoices (use to deduct cross-selling in physical shops)

– Product information (product categorization, …)



#2: Physical stores vs webshop


Impact physical store on online?

– Are online sales higher when physical store is nearby?

– Where to open a new store?

– How to approach your customers to motivate


Impact physical shops - Why bother?

• Determine strength of online brand vs physical brand

– Is online sales driven by brand awareness?

– Or is there quite a balance between the two?

– Omni-channel strategy?

• Know what would be the impact of opening/closing a

physical shop, also on the online business

– Support management decisions with facts & figures

• Depends heavily on sector/product/case/…


Impact physical shops - example

• Analysis for a retailer: Physical shops vs online sales


Impact physical shops - example

• Impact of opening a physical shop on local online sales

(brand awareness?)


Impact physical shops – now what?

• Use this correlation information:

– As extra input for determining new shop locations

– Publish folders focusing on online in non-covered areas

– Use popup-stores to get brand awareness and drive online sales

– Discounts per region

– Google Adwords campaigns focusing on regions with limited

brand presence

– Customer segmentation based on this information



#3: Dynamic Pricing


Dynamic prices

– End of life products?

– Relevancy of products.

– (Local) competition.

– Customer!


Dynamic Prices – some ideas

• Auto-combination special offers based on cross-selling

info

• Monitor stock & manage promotions accordingly– Example: stock of calendars in December

(value decreases over time…)

– Example: Customer history: needs incentive to buy?

Why not give a small discount if bought

together?

Testing will show if and for which products and

customers this increases revenue!


Dynamic Prices – some ideas

• Pricing vs competition

scraping competition websites

• Analysis of tenders vs deals

– What type of deals do we typically win, and which not?

= Data mining on CRM data!

– How can we optimize our chances to make a deal?

Which tenders should we invest in? What offer should we make?

• Remark: in B2C scenarios, can be difficult / unwanted to

use dynamic prices. Mind the legal impact!



#4: Personalized offerings


Personalized offering

– Loyal (online) customer vs new customers.

– Browsing habits and patterns.

– Spending patterns.

– Personalized discounts and/or content?


Personalized offerings

• Customer should be central in the webshop

– Provide a truly personalized shopping experience

– Like high-end physical shops with personal approach to VIP customers

• Gather data about your customer

– Surfing history – what products where looked at? How long? …

– What products were bought? When?

– Brand preference?

– Product-segment preference? (budget, high-end, best-buy?)

– Abandoned shopping carts

• Take action based on information mined from this data

– Triggered e-mails, personal recommendations, …


Personalized offerings – some ideas

• Imply social media– Are there any connections of our customer that wrote product

recommendations that might convince him to buy?

– Do we know the shopping behaviour of some of the customers’ connections? Are they in line with his/hers? Can we use this to make better recommendations?

• Anticipate customer behaviour– Use all customer contact moments

eg: if customer calls customer service, they should know what products the customer was looking at during his last visit to the webshop

– Prediction model (surfing behaviour vs % deal making)eg: Low chance? Go to checkout immediately. High chance? Offer extra cross-selling opportunities



#5: Gather external data


Gather external data, zoom & magnify

– Explore search trends within Google.

– Detect what is hot on social media.

– Magnify to the results and set clear goals/actions.

– Take action!


Gather and use external data

• example: how to sell a Smartwatch?

– It’s a new product, how to market it effectively?

– eg: SEO in line with trending topics on twitter, facebook posts, …

– eg: SEO in line with used search terms

• Added value: combining external data sources with own data

• Some ideas

– Find and follow your contacts on LinkedIn

previous/future employers of your contacts may be great prospects for

your B2B business!

– Use weather info to adapt the featured product offering

Data Science exercise: do we find any correlation between the weather

and the product sales figures?



#6: Anticipatory shipping


Anticipatory shipping

– Patent pending by Amazon.

– Ships an order before it is placed.

– Order history, search, wish list and click behaviour!


Anticipatory shipping

• High-tech? Actually not complex at all …

• Steps:– Gather many info on past orders

(customer info, country, product info, price, product group, product combinations, time of day, season, …)

– Build a prediction model predicting “cancelled or not” based on all this information

– Assess the quality of the model by training it with 90% of your historical orders and testing it with 10% of your historical orders

– Pass each new order’s info and predict the likelihood of this order getting cancelled (0 .. 100%) and act accordingly



#7: Customer Service optimizations


Customer service

– Losing sales/conversion/money by poor customer service.

– Optimize information for all communication channels.

– Which issues are your customers concerned with?

– Allocate resources better!


Customer Service – Some Ideas

• Text mining– Mood analysis: detect negative messages on social media, forum, …

Put TODO on action list of customer care to contact with certain priority

– Auto-classification of e-mails, letters, messages: Is this e-mail a question or a complaint?Is it about the quality of the product or financial (wrong invoice, …)?Automatic routing of messages to the right person! (operational optimization)

• Social media– Social media status of customer (scoring based on profile)

What’s would be the impact of this customer being unhappy about our service?

• Omnichannel insights– What did this customer buy of look at?

– How did he rate the last bought products?

– Which contacts (mail, phone, …) did we have and what seems to be the most effective deal trigger?


Applying Data Science


Applying Data Science

• Data Science does not replace business knowledge– Need to find balance between the two– Confirm or deny assumed business knowledge

– Detect changing trends early (customer behaviour, …)

• Not a development cycle, rather exploratory process:– Formulate hypotheses

– Data mining and modeling

– A/B testing (test new idea on x % of your customers/products/…)

– Conclusions: did the test group show better conversion?

– Rollout or cancel and start over!

• Potential issues– Privacy law and other legal restrictions

– Feedback loops, information leakage, wrong assumptionseg: trying to gather customer preferences when an order could as well have been a gift to someone else (perfume, …)


Questions?

Data Science for e-commerce

Data & Analytics

data sciencehow

big data cases big data

secondary use of data

position tosmart data

bigdata consultants

bigdata netflix

software development

better offers