• Save
Neil Mason presents on Data Mining and Predictive Analytics at Emetrics San Fransisco 2008
Upcoming SlideShare
Loading in...5
×
 

Neil Mason presents on Data Mining and Predictive Analytics at Emetrics San Fransisco 2008

on

  • 3,652 views

 

Statistics

Views

Total Views
3,652
Views on SlideShare
3,634
Embed Views
18

Actions

Likes
18
Downloads
0
Comments
0

4 Embeds 18

http://www.slideshare.net 8
http://www.linkedin.com 7
https://www.linkedin.com 2
http://www.lmodules.com 1

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

Neil Mason presents on Data Mining and Predictive Analytics at Emetrics San Fransisco 2008 Neil Mason presents on Data Mining and Predictive Analytics at Emetrics San Fransisco 2008 Presentation Transcript

  • Cutting through the NOISE!! Applications of data mining and predictive analytics A li ti fd t ii d di ti l ti Neil Mason, Applied Insights Emetrics, San Francisco, May 2008
  • The work we do in the online space Analytics strategy development and implementation Systems selection & implementation Site and Customer Analytics
  • The work we do in the online space Analytics strategy development and implementation Systems selection & implementation Site and Customer Analytics
  • The work we do in the online space Analytics strategy development and implementation Systems selection & implementation Site and Customer Analytics
  • The work we do in the online space Analytics strategy development and implementation Systems selection & implementation Site and Customer Analytics
  • The work we do in the online space Analytics strategy development and implementation Systems selection & implementation Site and Customer Analytics
  • The challenge… Survey Data Promotion Ad- Affiliates data GRP data serving data Email data Customer You Performance data data Transactions T ti ISP data PPC data Web analytics y Panel data Analyst Offline data sales data
  • Web Survey Panel data Customer analytics d data data d data
  • Rabbits in headlights…
  • Rabbits in headlights…
  • The response? Data integration Better query engines Data mining and predictive analytics y
  • What do we mean by data mining and predictive analytics? Predictive Data mining analytics Discovering previously undetected patterns and Applying historical patterns to relationships in data predict future outcomes
  • Application of predictive analytics Number of tracks Day of Country presentation Length of Time of conference presentation Expected Size of size of After lunch? conference audience
  • Application of predictive analytics Number of tracks Day of Country presentation Length of Time of conference presentation 4 Size of After lunch? conference
  • Predictive Analytics - Techniques • Statistics • e.g. Regression • Artificial intelligence • e.g. N Neural N t l Networks k • Hybrid • e.g. D i i t Decision trees • Optimisation • e g Monte Carlo Simulation e.g. Simulation,
  • The data mining process (CRISP-DM) Business Data Understanding Understanding Data Preparation Deployment Modelling Evaluation
  • The data mining process (CRISP-DM) Business Data Understanding Understanding Data Preparation Deployment Modelling Evaluation
  • Some applications of data mining and predictive analytical techniques Segmentation S t ti Propensity modelling Econometrics and forecasting Anomaly detection
  • Some applications of data mining and predictive analytical techniques Segmentation S t ti Propensity modelling Econometrics and forecasting Anomaly detection
  • Who are your visitors? Applications of visitor segmentation techniques
  • Creating meaningful segments • Demographic • Gender, age etc g • Lifestyle • Behavioural •BBrowsing i • Purchasing • Response • Attitudinal • Brand empathy • Satisfaction
  • Creating meaningful segments • Demographic • Gender, age etc g • Lifestyle • Behavioural •BBrowsing i • Purchasing • Response • Attitudinal • Brand empathy • Satisfaction
  • Behavioural segmentation strategies Deterministic Discovery based Rules Associations Hierarchies Patterns Filters Correlations
  • The framework… Who visits the Why do they visit What do they do on site? the site and what the site? do they think of it? ? ? ? ?
  • Developing the visitor segments Behavioural segmentation based on content b d tt consumption Segments profiled using other behavioural data and also additional survey and/or customer data
  • Segmentation using cluster analysis Behavioural data Vis123 Vis124 Vis125 Vis126 Vis127 Vis128 Vis129 Vis130 Vis131
  • Building the visitor profile… Profiling data Behavioural data Attitudinal data Vis128 Vis130 Vis124 Vis123 Vis126 Vis127 Vis131 Vis129 Vis125
  • Happy Trackers (6%) Happy Trackers mainly use the site for Track and Trace and little else In terms of profile they tend to have a stronger business slant and be slightly older than on average g They are not heavy users of the site and their visits are relatively light and narrow – all they do is use Track and Trace However they are happy with what they do, they rate the site functionality the best out of all the segments
  • Happy Trackers– 6%, Occasional information Top content Top searches Top campaigns • Track & trace • Redirections • Redelivery • Redirections • Recorded delivery • XMAS • Customer services • Redeli er Redelivery •SSmartstamp tt • Delivery services • 9th highest number of visits Key behaviours • 4th most buyers; redirections • Key demographics & attitudes • Older • More business than personal • Satisfaction above par • Highest site rating • Stated reasons for visit: Track & Trace
  • Price Finders (10%) Price Finders are primarily concerned about finding our information on things like airmail services and prices as well as other delivery services and costs Quite often their visit has something to do with an online auction activity but they are possibly new to the game as this segment generally haven t visited haven’t the site very often and a large proportion of them are new to the site
  • Cottage Industrialists (2%) Cottage Industrialists are frequent users of the site and they mainly come looking for information on postal prices, delivery services, parcel information and the like. Half of this segment are involved in some type of online auction related activity and over the course of their lifetime they tend to look at the broadest amount of content on the site. Quite often they will be using the search function to do this They are reasonably happy with the customer experience on the site and are more likely than on average to recommend the site to others
  • Regular Posters (1%) A small but valuable segment Regular Posters are frequent visitors to the site and are mainly buying stamps via online postage. The vast majority of this group actually bought something d i th period thi during the id This segment has a slightly more older male profile and is more likely to be coming for business reasons As well as visiting frequently, their visits also tend to be longer and heaviest in terms of content consumption ti However, they are not as satisfied with the site experience as other groups, possibly due to the processes i involved ld
  • The framework…in action Who visits the Why do they visit What do they do on site? the site and what the site? do they think of it?
  • Segmentation for email targeting Segment 3: Segment 5: Average # orders 3.3 33 Average # orders 3.3 Similar ordering patterns Avg # items 6.4 Avg # items 6.0 Avg spend £175 Avg spend £178 Avg order value £54 Avg order value g £53 Avg items per order 2.0 Avg items per order 1.8 Products: Products: Different product purchasing p p g DIY Domestic appliances pp Car maintenance Furnishings Garden tools and furniture Nursery Index I d vs all online ll li Male Ml Female F l Index I d vs allll Male Ml Female F l Different demographics shoppers online shoppers Younger (<35) 87 78 Younger (<35) 83 122 Older (>35) 127 97 Older (>35) 87 106
  • It’s often all about timing… Tinofrteaapa im fis m per g il s to make a difference The whole tree is not displayed here… Overall the propensity to order twice doubles if an email is sent within the first 3 days – emailing within 5 days still generates a significant increase in conversion from single shopper to repeat shopper
  • Understanding the drivers of conversion over multiple visits Propensity to convert…
  • It generally takes more than one visit to get the conversion Car Insurance 120% omers 100% mulative % of custo 80% 60% 40% Cum 20% 0% 1 2 3 4 5 6 Number of visits to conversion
  • Tracking visitor behaviour over multiple visits First visit Second visit Subsequent Purchase visit visit •Source of •Days since •Days since •Days since first visit first visit first visit first visit •Campaign •Entry page •Entry page •Source of visit? visit •etc •etc Keywords Campaign •Keywords •Campaign used? visit? •Day/time •Keywords used? •Depth of visit •Tool used? •Tool used? Tl d? •Email E il landing? •Entry page •Exit page
  • Building the event profile… Visit 1 events Visit 2 events Visit 3 events Purchase visit events Vis123 Vis124 Vis125 Vis126 Vis127 Vis128 Vis129 Vis130 Vis131
  • Key drivers of First Visit Buyers All First Time Buyers Index = 100 Paid & Natural Direct Landing Affiliate Other Search Index = 131 Index = 46 Index = 77 Index = 100 d Branded Non‐branded keyword keyword Index = 146 Index = 46
  • What are the main factors influencing purchases over multiple visits? Conversion amongst multi‐visit visitors Index = 100 Used tool Didn’t use tool on first visit on first visit Index = 156 Index = 69 2nd visit 4 days y 2nd visit on same i it Second visit S d i it Second visit S d i it 2nd visit more than d i it th or less from first day as first within 8 days after 8 days 4 days from first Index = 73 Index = 149 Index = 174 Index = 146 Index = 59
  • Conclusions • “Web analytics” is a journey not an event • A volume and complexity i As l d l it increases new t l such as tools h data mining and predictive analytics are needed in the analysts tool box • Operationally deployed • Testing systems, targeting systems • As an ad-hoc weapon ad hoc • DM & PA can help cut through the noise and reveal relationships and patterns that would be difficult to determine using t diti dt i i traditional queering approaches l i h • Challenges: • Data preparation and management • Selection of appropriate tools and techniques • Ability to execute!
  • Thank you! Any questions? Neil Mason neil@applied-insights.co.uk