際際滷

際際滷Share a Scribd company logo

Smart Cross Selling App

Volodymyr Holomb
Volodymyr Holomb

Cross-selling lead online generator for insurance and banking industries will help you to gain more appropriate leads from your database of clients

1 of 17
Gain more appropriate leads from your database of policyholders www.rbcgrp.com smart cross sales
2 www.rbcgrp.com https://www.insurancethoughtleadership.com/customer-experience/lowering-costs-customer-acquisition https://www.retentionscience.com/blog/customer-retention-should-outweigh-customer-acquisition/ https://blog.agentero.com/2018/10/04/the-optimal-cross-selling-strategy-for-insurance-agents/ Insurance has the highest customer acquisition cost of any industry 7-9x 7-9 times, it costs more for an insurance agency to attract a new customer than to retain an existing customer Companies focus more attention on acquisition than retention 44/18 About 44% dedicate themselves to the acquisition, while only 18% focus on retention Customer retention has a profound effect on the bottom line 5>>25-95 When customer retention increases by only 5%, profits increase by 25% to 95% Cross-selling is a crucial component of customer loyalty The cross-selling opportunities for insurance clients are tremendous: 61% of policyholders have only one policy with their agent 29% have two policies 10% have three or more policies Why insurers need cross-selling optimization
3 www.rbcgrp.com Four big mistakes why cross-selling fails 1 2 3 4 Cross-Selling without Analysis It isnt easy to succeed without choosing the target audience appropriately, identifying their preferences, and determining what has value Improper Identification of Target Buyers The starting point of any selling process is to find the appropriate customers Unprofitable Customers One in five 1) cross-buyers is unprofitable, although the average profit from shoppers who cross-buy is higher than those who dont Cross-selling the Wrong Products Cross-selling inappropriate, no- value products to shoppers are one more reason why it fails 1) https://hbr.org/2012/12/the-dark-side-of-cross-selling
4 www.rbcgrp.com We know how insurers can get more most appropriate cross-selling leads from their database of policyholders
5 www.rbcgrp.com We democratise AI and we launch smart cross sales
6 www.rbcgrp.com How democratized AI works to gain leads? You tag customers who have already bought the target product. We predict other customers that might buy it. Data preparation and uploading AI training Prediction & accuracy evaluation Business case modeling and data downloading Data checking This is your homework compose data for customer profiles and mark target customers We will provide tips and templates for data preparation Upload prepared data Checking the data structure Data quality check Clustering and selecting Converting and normalization Preparation of datasets for train/test Making and training NN's Saving a NN models Prediction purchase probability Calculation of accuracy metrics Generating CumulativeGain & CumulativeLift plots Modeling a business case by changing the conversion rate and average revenue per customer Download a list of leads to a file 2 3 4 5 1 Fully automatic data processing and customer similarity prediction User interaction User interaction
7 www.rbcgrp.com How democratized AI works to gain leads? 2 3 4 5 1 User interaction User interaction Upload the data Confirm the field name with the customer ID and tag of the target customers Evaluating model performance Download the data
8 www.rbcgrp.com Key features and benefits Theshortlearningcurve,uploadthefileandgetresults Preparecustomerprofiledataonceandeasilychangetargetcustomers asnewleadsareneeded Dramaticallytimereduction.Spendminutestogainthousandsofleads. Automaticdataqualitycheck Estimatefirst,thengetresultsandpay ReduceCACupto80% Increaseconversionrate morethantwotimes Identifyingthemostappropriatecustomersforcross-selling Personaldataofclientsisnotused Youcanscramblethedata ThetransmitteddataisnotsavedandisstoredinRAMonlyuntilthe resultsarereceived Simplicity Fullyautomaticdata processing High-accuracyAI- poweredmodels Safety
9 www.rbcgrp.com How much does it cost? 0,1 per lead EVALUATION GET RESULTS PAY Are you estimating 100 000+ leads per month? Request a corporate price!
10 www.rbcgrp.com Early Adopters Program values 0,01 per lead for the first 100 000 EVALUATION GET RESULTS PAY Getinsightsanddataforincreasingcross- salesbeforeanyoneelseinthemarket TimeWinning Personalmanagerforadviceindata preparation,applicationtrainingand support8/5inthefirstmonthofuse Babysitterssupport Newfeaturesdevelopmentand customisationbasedonyourfeedback Quickdevelopmentresponse
11 www.rbcgrp.com How does it work? 1) https://www.kaggle.com/datasets/ishandutta/analytics-vidhya-insurance-cross-sell-prediction 2) https://unbounce.com/conversion-benchmark-report#finance_and_insurance 3) https://www.insuranceeurope.eu/publications/465/european-motor-insurance-markets/ Among the company's clients, it is required to find potential ones to offer the option of damage coverage. For testing, we have a label with customers responses. 381k customer records1 with 12 columns and 5k listed as target customers Case #1: 4 641 990 of expected insuran c e premium Recall is the fraction of the relevant results (customers who bought the product) that are successfully defined. All relevant customers = 7+75 Recall=75/(7+75)=0.91 Models performance Expected premium: 4 641 990 Num of recommended leads 94 015 Average conversion in the insurance industry 0,079 Lift 2,5 The average premium for 束Damage covering損 250 Cumulative gains when we select ~30% with the highest probability according to NN models, this selection holds ~78% of all Damage coverage cases in test data. Lift curve when we select ~30% with the highest probability according to NN models, this selection for Damage coverage cases is ~2,5 times better than selecting without a model. Other results Processing time: 163,3seconds Recall: 91,46% CAC reduce to ~28,1% from the original Cost of leads processing: 94015*0,1= 9401,5 in the original dataset after clustering after forecasting similarity 94 015 recommended leads Columns name Role Unique ID for the customer ID Response for Damage covering option Target Gender of the customer Feature Age of the customer Feature Driving license Feature etc. Feature
12 www.rbcgrp.com How does it work? 1) https://unbounce.com/conversion-benchmark-report#finance_and_insurance Case #2: 3 737 086 of expected insuran c e premium Recall is the fraction of the relevant results (customers who bought the product) that are successfully defined. All relevant customers = 101+1168 Recall=1168/(101+1168 )=0.92 Models performance Expected premium: 3 737 086 Num of recommended leads 27 278 Average conversion in the insurance industry 0,079 Lift 15 The average premium for 束Damage covering損 137 Cumulative gains when we select ~7% with the highest probability according to NN models, this selection holds ~90% of all Damage coverage cases in test data. Lift curve when we select ~7% with the highest probability according to NN models, this selection for Damage coverage cases is ~15 times better than selecting without a model. in the original dataset after clustering after forecasting similarity 27 278 recommended leads The client portfolio of a risk insurance company has 1,600 clients who have bought a life insurance policy. It is necessary to find clients similar to them to offer them to buy a life insurance policy as well. ~2,3M customer records with 144 columns in Customer Profile 360属 1) Demographics (city, age, region etc.) Information regarding holding policies of the customer etc. 1,6 k records listed as target customers Other results Processing time: 3444 seconds Recall: 92,06% CAC reduce to ~1,1% from the original Cost of leads processing: 27278*0,1= 2727,8
13 www.rbcgrp.com A few tips for use Create a list of clients who have once left your company. In this list, tag customers who have returned after leaving and see the list of similar ones generated by the AI to add them to the return campaign. Get back churned customers Make agents cross-selling work easier. Let them concentrate on building customer relationships and trust AI to generate leads lists. Join 10+ agents, and we'll give them a free training webinar. Bring AI-powered cross-selling to your insurance agents Create a common client list with your partner. Get a list of potential customers for your partner's product or service. Similarly, get a list of potential customers for your target product. Combine these two lists and make a joint offer: partner's product + your product Create co-branded offers for clients
14 www.rbcgrp.com Have more questions? Request a meeting!
15 www.rbcgrp.com Ready to gain more leads? Request full access!
16 www.rbcgrp.com Need more complex solutions? Request a meeting about: SMART segmentation PERSONAL recommendations CHURN prediction
philipenko@rbcgrp.com www.rbcgrp.com Sincerely, Igor Philipenko, Director of Advanced Analytics at RBC Group

More Related Content

Smart Cross Selling App

  • 1. Gain more appropriate leads from your database of policyholders www.rbcgrp.com smart cross sales
  • 2. 2 www.rbcgrp.com https://www.insurancethoughtleadership.com/customer-experience/lowering-costs-customer-acquisition https://www.retentionscience.com/blog/customer-retention-should-outweigh-customer-acquisition/ https://blog.agentero.com/2018/10/04/the-optimal-cross-selling-strategy-for-insurance-agents/ Insurance has the highest customer acquisition cost of any industry 7-9x 7-9 times, it costs more for an insurance agency to attract a new customer than to retain an existing customer Companies focus more attention on acquisition than retention 44/18 About 44% dedicate themselves to the acquisition, while only 18% focus on retention Customer retention has a profound effect on the bottom line 5>>25-95 When customer retention increases by only 5%, profits increase by 25% to 95% Cross-selling is a crucial component of customer loyalty The cross-selling opportunities for insurance clients are tremendous: 61% of policyholders have only one policy with their agent 29% have two policies 10% have three or more policies Why insurers need cross-selling optimization
  • 3. 3 www.rbcgrp.com Four big mistakes why cross-selling fails 1 2 3 4 Cross-Selling without Analysis It isnt easy to succeed without choosing the target audience appropriately, identifying their preferences, and determining what has value Improper Identification of Target Buyers The starting point of any selling process is to find the appropriate customers Unprofitable Customers One in five 1) cross-buyers is unprofitable, although the average profit from shoppers who cross-buy is higher than those who dont Cross-selling the Wrong Products Cross-selling inappropriate, no- value products to shoppers are one more reason why it fails 1) https://hbr.org/2012/12/the-dark-side-of-cross-selling
  • 4. 4 www.rbcgrp.com We know how insurers can get more most appropriate cross-selling leads from their database of policyholders
  • 5. 5 www.rbcgrp.com We democratise AI and we launch smart cross sales
  • 6. 6 www.rbcgrp.com How democratized AI works to gain leads? You tag customers who have already bought the target product. We predict other customers that might buy it. Data preparation and uploading AI training Prediction & accuracy evaluation Business case modeling and data downloading Data checking This is your homework compose data for customer profiles and mark target customers We will provide tips and templates for data preparation Upload prepared data Checking the data structure Data quality check Clustering and selecting Converting and normalization Preparation of datasets for train/test Making and training NN's Saving a NN models Prediction purchase probability Calculation of accuracy metrics Generating CumulativeGain & CumulativeLift plots Modeling a business case by changing the conversion rate and average revenue per customer Download a list of leads to a file 2 3 4 5 1 Fully automatic data processing and customer similarity prediction User interaction User interaction
  • 7. 7 www.rbcgrp.com How democratized AI works to gain leads? 2 3 4 5 1 User interaction User interaction Upload the data Confirm the field name with the customer ID and tag of the target customers Evaluating model performance Download the data
  • 8. 8 www.rbcgrp.com Key features and benefits Theshortlearningcurve,uploadthefileandgetresults Preparecustomerprofiledataonceandeasilychangetargetcustomers asnewleadsareneeded Dramaticallytimereduction.Spendminutestogainthousandsofleads. Automaticdataqualitycheck Estimatefirst,thengetresultsandpay ReduceCACupto80% Increaseconversionrate morethantwotimes Identifyingthemostappropriatecustomersforcross-selling Personaldataofclientsisnotused Youcanscramblethedata ThetransmitteddataisnotsavedandisstoredinRAMonlyuntilthe resultsarereceived Simplicity Fullyautomaticdata processing High-accuracyAI- poweredmodels Safety
  • 9. 9 www.rbcgrp.com How much does it cost? 0,1 per lead EVALUATION GET RESULTS PAY Are you estimating 100 000+ leads per month? Request a corporate price!
  • 10. 10 www.rbcgrp.com Early Adopters Program values 0,01 per lead for the first 100 000 EVALUATION GET RESULTS PAY Getinsightsanddataforincreasingcross- salesbeforeanyoneelseinthemarket TimeWinning Personalmanagerforadviceindata preparation,applicationtrainingand support8/5inthefirstmonthofuse Babysitterssupport Newfeaturesdevelopmentand customisationbasedonyourfeedback Quickdevelopmentresponse
  • 11. 11 www.rbcgrp.com How does it work? 1) https://www.kaggle.com/datasets/ishandutta/analytics-vidhya-insurance-cross-sell-prediction 2) https://unbounce.com/conversion-benchmark-report#finance_and_insurance 3) https://www.insuranceeurope.eu/publications/465/european-motor-insurance-markets/ Among the company's clients, it is required to find potential ones to offer the option of damage coverage. For testing, we have a label with customers responses. 381k customer records1 with 12 columns and 5k listed as target customers Case #1: 4 641 990 of expected insuran c e premium Recall is the fraction of the relevant results (customers who bought the product) that are successfully defined. All relevant customers = 7+75 Recall=75/(7+75)=0.91 Models performance Expected premium: 4 641 990 Num of recommended leads 94 015 Average conversion in the insurance industry 0,079 Lift 2,5 The average premium for 束Damage covering損 250 Cumulative gains when we select ~30% with the highest probability according to NN models, this selection holds ~78% of all Damage coverage cases in test data. Lift curve when we select ~30% with the highest probability according to NN models, this selection for Damage coverage cases is ~2,5 times better than selecting without a model. Other results Processing time: 163,3seconds Recall: 91,46% CAC reduce to ~28,1% from the original Cost of leads processing: 94015*0,1= 9401,5 in the original dataset after clustering after forecasting similarity 94 015 recommended leads Columns name Role Unique ID for the customer ID Response for Damage covering option Target Gender of the customer Feature Age of the customer Feature Driving license Feature etc. Feature
  • 12. 12 www.rbcgrp.com How does it work? 1) https://unbounce.com/conversion-benchmark-report#finance_and_insurance Case #2: 3 737 086 of expected insuran c e premium Recall is the fraction of the relevant results (customers who bought the product) that are successfully defined. All relevant customers = 101+1168 Recall=1168/(101+1168 )=0.92 Models performance Expected premium: 3 737 086 Num of recommended leads 27 278 Average conversion in the insurance industry 0,079 Lift 15 The average premium for 束Damage covering損 137 Cumulative gains when we select ~7% with the highest probability according to NN models, this selection holds ~90% of all Damage coverage cases in test data. Lift curve when we select ~7% with the highest probability according to NN models, this selection for Damage coverage cases is ~15 times better than selecting without a model. in the original dataset after clustering after forecasting similarity 27 278 recommended leads The client portfolio of a risk insurance company has 1,600 clients who have bought a life insurance policy. It is necessary to find clients similar to them to offer them to buy a life insurance policy as well. ~2,3M customer records with 144 columns in Customer Profile 360属 1) Demographics (city, age, region etc.) Information regarding holding policies of the customer etc. 1,6 k records listed as target customers Other results Processing time: 3444 seconds Recall: 92,06% CAC reduce to ~1,1% from the original Cost of leads processing: 27278*0,1= 2727,8
  • 13. 13 www.rbcgrp.com A few tips for use Create a list of clients who have once left your company. In this list, tag customers who have returned after leaving and see the list of similar ones generated by the AI to add them to the return campaign. Get back churned customers Make agents cross-selling work easier. Let them concentrate on building customer relationships and trust AI to generate leads lists. Join 10+ agents, and we'll give them a free training webinar. Bring AI-powered cross-selling to your insurance agents Create a common client list with your partner. Get a list of potential customers for your partner's product or service. Similarly, get a list of potential customers for your target product. Combine these two lists and make a joint offer: partner's product + your product Create co-branded offers for clients
  • 15. 15 www.rbcgrp.com Ready to gain more leads? Request full access!
  • 16. 16 www.rbcgrp.com Need more complex solutions? Request a meeting about: SMART segmentation PERSONAL recommendations CHURN prediction

Editor's Notes

  1. Early adopters discount 50%
  2. Early adopters discount 50%