AI & MACHINE LEARNING
IN ECOMMERCE
Andraž Štalec
25 % increase in retail e-commerce sales in 2017
RECORD NUMBER OF TRANSACTIONS & REVENUE
Statista 2018
WHY IS THIS HAPPENING?
Need to understand the bigger picture
THE TIDE IS SHIFTING
Media consumption vs spend 2016
eMarketer 4/16 & IAB 2016
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Print Radio TV Desktop internet Mobile internet
Time Spent Ad Spend
3 things that will redefine your digital focus
BUDGETS WILL
SHIFT TO
ONLINE
BUILD
AWARENESS
AND INTENT
ONLINE
DELIVER BETTER
ROI
TO SUSTAIN THESE SHIFTS COMPANIES WILL RELY ON
DATA TO GROW AND OPTIMIZE THEIR BUSINESSES
ENTERING THE 3RD PHASE
OF DIGITAL TRANSFORMATION
HUGE ADVANCEMENTS IN ARTIFICIAL INTELLIGENCE
AI AS A UTILITY
Brains not distracted as humans are
iOS maps notifications
iOS maps notifications
Artificial intelligence
AI is not a plug-in
Enhance
Scale
Automate
Build processes manually
4 pillars of AI future
Data Technology Process People
4 pillars of AI future
Data Technology Process People
The problem of No Data vs the problem of Big Data
TRADITIONAL RETAIL VS ECOMMERCE
Entry barriers offline are
much higher than online
Entry barriers
Number of competitors
Number of competitors is rapidly
increasing
Growth of online user
population is slowing down
Number of users
Returning customers
Number of returning customers
is in decline as new competitors
are entering the market
Cost of advertising is rising and
one-time customers are not
profitable anymore
One-time customers
Main difference between offline and online retail
One-time customer revenue stream
CPA
GM CPA <= GM
The only way your business could be profitable is achieving CPA
much lower then your Gross Margin.
One-time customer revenue stream
CPA
GM CPA > GM
Due to increased competition CPCs are increasing and CRs are
decreasing.
Shift to retention model
CAC
GM
CLV <= CLC
Sum of GM from all customer transactions should be higher then
sum of all costs connected to the customer
GM GM
GM
GM
GM
MRC MRC MRC MRC MRC MRC MRC MRC MRC MRC MRC
CLV
CLC
CLVP = (P × AOV) × AGM) × ALT
The tricky part is calculating your CLV
The SaaS Ecommerce Funnel
Acquire Activate Retain Grow Referrals
2nd purchase1st purchase
How do you measure churn in ecommerce?
Customer aquisition
Retention
Customer churns
Retention for other ecommerce sites
3 months 6 months
The SaaS Ecommerce Funnel
Acquire Activate Retain Grow Referrals
1st purchase 2nd purchase Customer
Engagement
Score
Customer Engagement Score
01
02
03
043%
35%
10%
50% 05
10%
Days since last
session
Sessions with
product views
Interest-based
browsing
Adds to cart Email engagement
rate
Active vs. churned customers
Active vs. churned customers
active
churned
WITH NEW FOCUS NEW KPIs ARISE
Top Ecommerce metrics
2 31 4 5 6 7
CLV CAC Customer
profitability
score
Repeat rate Retention rate AOV CES
UNDERSTAND YOUR DATA
Before you begin
Gather and prepare the data
Transactional data from
client‘s ERP
3rd party data for
enrichment
Behavioral and
demographic data from
Google Analytics
Ecommerce segmentation
K-means clustering based on RFV
Create distribution for new features like
ALT
Add them to clusters to create more
acurate segments
Clustering:
2
3
1
High value
High Frequency
Grey zone Home run
Grey zoneDead zone
WHY THE HELL ALL THIS TALK ABOUT DATA?
2 main usecases of AI
Insights
Use AI to deliver better and deeper
actionable insights for your business.
User experience
Use AI to deliver better experience to
users, automate processes and scale
the business.
Customer clusters
segmented by purchase
behavior, CES & LTV Size of clusters
How are your customers
segmented?
Roadmap to ML in Ecommerce
Prevent churn
and re-engage users
based on deviation
in their behavior
Predict LTV
of newly acquired users and
bid accordingly
Customer
transitions
Between clusters
SaaS Ecomm metrics
like churn, retention, ALT, CES
Anomaly
Detection
Customer
Support
Product
Recommend
ation
Personalizati
on
SearchDynamic
Pricing
06 01
02
0304
05
6 usecases of AI for enhanced User experience
BARS ARE RISING
Product quality + Customer support + Transparency
GOOD ENOUGH IS JUST NOT GOOD ENOUGH!
2 areas that we need to focus on for growth in 2018 and beyond
EXCEPTIONAL
LIFETIME
EXPERIENCE
BETTER
PERFORMANCE
AND ROI
Thank you
01 02 03 04
Email
andraz@red-orbit.com
Twitter
@andrazstalec
LinkedIn
Andraz Stalec
Web
www.red-orbit.com
Red Orbit, Jožeta Jame 12,1000 Ljubljana, Slovenia
W: www.red-orbit.com |E: info@red-orbit.com | T: +386 (0)590 75 680

Machine Learning & AI in e-commerce

  • 1.
    AI & MACHINELEARNING IN ECOMMERCE Andraž Štalec
  • 2.
    25 % increasein retail e-commerce sales in 2017 RECORD NUMBER OF TRANSACTIONS & REVENUE Statista 2018
  • 3.
    WHY IS THISHAPPENING?
  • 4.
    Need to understandthe bigger picture THE TIDE IS SHIFTING
  • 5.
    Media consumption vsspend 2016 eMarketer 4/16 & IAB 2016 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% Print Radio TV Desktop internet Mobile internet Time Spent Ad Spend
  • 6.
    3 things thatwill redefine your digital focus BUDGETS WILL SHIFT TO ONLINE BUILD AWARENESS AND INTENT ONLINE DELIVER BETTER ROI
  • 7.
    TO SUSTAIN THESESHIFTS COMPANIES WILL RELY ON DATA TO GROW AND OPTIMIZE THEIR BUSINESSES
  • 8.
    ENTERING THE 3RDPHASE OF DIGITAL TRANSFORMATION
  • 10.
    HUGE ADVANCEMENTS INARTIFICIAL INTELLIGENCE
  • 12.
    AI AS AUTILITY Brains not distracted as humans are
  • 13.
  • 14.
  • 15.
  • 16.
    AI is nota plug-in Enhance Scale Automate Build processes manually
  • 17.
    4 pillars ofAI future Data Technology Process People
  • 18.
    4 pillars ofAI future Data Technology Process People
  • 19.
    The problem ofNo Data vs the problem of Big Data TRADITIONAL RETAIL VS ECOMMERCE
  • 20.
    Entry barriers offlineare much higher than online Entry barriers Number of competitors Number of competitors is rapidly increasing Growth of online user population is slowing down Number of users Returning customers Number of returning customers is in decline as new competitors are entering the market Cost of advertising is rising and one-time customers are not profitable anymore One-time customers Main difference between offline and online retail
  • 21.
    One-time customer revenuestream CPA GM CPA <= GM The only way your business could be profitable is achieving CPA much lower then your Gross Margin.
  • 22.
    One-time customer revenuestream CPA GM CPA > GM Due to increased competition CPCs are increasing and CRs are decreasing.
  • 23.
    Shift to retentionmodel CAC GM CLV <= CLC Sum of GM from all customer transactions should be higher then sum of all costs connected to the customer GM GM GM GM GM MRC MRC MRC MRC MRC MRC MRC MRC MRC MRC MRC CLV CLC
  • 24.
    CLVP = (P× AOV) × AGM) × ALT The tricky part is calculating your CLV
  • 25.
    The SaaS EcommerceFunnel Acquire Activate Retain Grow Referrals 2nd purchase1st purchase
  • 26.
    How do youmeasure churn in ecommerce? Customer aquisition Retention Customer churns
  • 27.
    Retention for otherecommerce sites 3 months 6 months
  • 28.
    The SaaS EcommerceFunnel Acquire Activate Retain Grow Referrals 1st purchase 2nd purchase Customer Engagement Score
  • 29.
    Customer Engagement Score 01 02 03 043% 35% 10% 50%05 10% Days since last session Sessions with product views Interest-based browsing Adds to cart Email engagement rate
  • 30.
  • 31.
    Active vs. churnedcustomers active churned
  • 32.
    WITH NEW FOCUSNEW KPIs ARISE
  • 33.
    Top Ecommerce metrics 231 4 5 6 7 CLV CAC Customer profitability score Repeat rate Retention rate AOV CES
  • 34.
  • 35.
    Gather and preparethe data Transactional data from client‘s ERP 3rd party data for enrichment Behavioral and demographic data from Google Analytics
  • 39.
    Ecommerce segmentation K-means clusteringbased on RFV Create distribution for new features like ALT Add them to clusters to create more acurate segments Clustering: 2 3 1 High value High Frequency Grey zone Home run Grey zoneDead zone
  • 40.
    WHY THE HELLALL THIS TALK ABOUT DATA?
  • 41.
    2 main usecasesof AI Insights Use AI to deliver better and deeper actionable insights for your business. User experience Use AI to deliver better experience to users, automate processes and scale the business.
  • 42.
    Customer clusters segmented bypurchase behavior, CES & LTV Size of clusters How are your customers segmented? Roadmap to ML in Ecommerce Prevent churn and re-engage users based on deviation in their behavior Predict LTV of newly acquired users and bid accordingly Customer transitions Between clusters SaaS Ecomm metrics like churn, retention, ALT, CES
  • 43.
  • 44.
    BARS ARE RISING Productquality + Customer support + Transparency
  • 45.
    GOOD ENOUGH ISJUST NOT GOOD ENOUGH!
  • 46.
    2 areas thatwe need to focus on for growth in 2018 and beyond EXCEPTIONAL LIFETIME EXPERIENCE BETTER PERFORMANCE AND ROI
  • 47.
    Thank you 01 0203 04 Email andraz@red-orbit.com Twitter @andrazstalec LinkedIn Andraz Stalec Web www.red-orbit.com Red Orbit, Jožeta Jame 12,1000 Ljubljana, Slovenia W: www.red-orbit.com |E: info@red-orbit.com | T: +386 (0)590 75 680

Editor's Notes

  • #3 We live in a tremendous time … yet, retailers find it harder to sustain growth & profitability Source: https://www.statista.com/statistics/379046/worldwide-retail-e-commerce-sales/
  • #5 Our world is becoming fully digitalized Consumers are moving from offline to online
  • #9 Entering the 3rd phase of digital tranfsormation: 1st = making information accessible 2nd = mobile & social media + how they connect people https://www.youtube.com/watch?time_continue=31&v=shDMy232MAA
  • #10 3rd = world of AI Leading companies moving from mobile first to AI first strategy
  • #17 Road to AI ni implementacija AI toolov ampak postaviti sistem (procese) na roke (manual) in jih potem avtomatizirati in izboljšati z AIjem; tako kot postavljaš GH in potem avtomatiziraš
  • #18 Enter the AI era: There are 4 pillars of AI future
  • #20 With explosion of data, online advertising and ecommerce are becoming increasingly measureable and actionable. eCommerce, like traditional retail involves taking in a lot of data and trying to make the best decisions based on the data that you have. In a retail store data is relatively limited, you might know how many customers come in the store each day (or at least be able to estimate it), but it’s hard to keep track of how many customers try things on, how long the average customer stays in a store, or how many items they look at during each visit. Online, it‘s much easier to track customer metrics because the data is easily accessible. You can tell exactly how long the average person spends on your site, which items they look at, what they add to their cart, and what they end up buying in the end. Unlike brick-and-mortar retail where the complexity comes in when you’re trying to collect data, online the complexity is making sense of all the data this is coming in.
  • #24 Need to shift to retention model, where customer profitability is a key metric CLV = SUM of all GMs CLC = CAC + MRC CAC MRC = cost of running business – servers, licences, client retention, staff One key component is keeping MRC as low as possible -> renention should not depend on advertising
  • #25 P = average monthly purchases AOV =  Average Order Value AGM = Average Gross Margin ALT = Average LifeTime in months
  • #32 Steep function
  • #40 Create clusters using RFV or AOV and Freq Add ALT so you get segmnets based on LTV
  • #44 https://techblog.commercetools.com/top-5-machine-learning-applications-for-e-commerce-268eb1c89607
  • #46 Section titles.