Ecommerce
&
MachineLearning
byDavidJones-TechnicalDirector-ResolveDigital
GoogleSearch-Netflix
GmailSpamFilter-Siri
FraudDetection-Amazon
Machinelearningisnotjustforthe
bigguys
anymore.
MachineLearninginEcommerce
Emailmarketing
Cross-selling
Personalisedrecommendations
AbandonedCartEmails
LeadScoring
—
1.CollectData
2.TrainModel
3.MakePredictions
—
CollectData:
Purchases
ProductViews
Likes&Ratings
Wishlists
Anydatathatlinks
customerstoproducts
2.TrainModel
—
prediction.io
OpenSourceMachineLearningServer
EcommerceTemplates
APIforDevelopers
Productionready
3.Makepredictions
Realworldexample:
UnitedCellars
16kproductviews
60korders
3kproductratings
79000rowssmalldata(bybayareastandards)
Withwine
tastepreferencesare
diverse
1.PersonalisedProduct
Recommendations
2.SimilarProducts
A/BTEST
FilterPredictionsAvoidanythingthatlessensrelevance
Don’tshowoutofstockproducts
Howdoyourcustomers’shop?
WithWinepeopleusuallypreferredorwhite
A/BTestResults
45%longeraveragesession
22%increaseinconversionrate
37%increaseinaverageordersize
71%MoreRevenue
MachineLearning
Generatesrevenue
Butalso
Improvescustomers’experience
3
Things
Learned
1
Thehumantouch
isstillimportant
2
prediction.io
isawesome
3
Smalldata
canbeenough
ThankYou
~
DavidJones,@d_jones
resolve.digital/ml