Scribd
Upload
63 views11 pages

Lead Scoring Case Study Presentation

The document discusses building a logistic regression model to assign a lead score between 0 and 100 to help an education company identify the most promising leads and improve their low lead conversion rate of 30%. Various analyses were performed on the data including univariate analysis of categorical and continuous variables, bivariate analysis of correlations, and model evaluation to find an optimal threshold. The model achieved good performance with an accuracy of 80% on the test data showing it can effectively identify hot leads and increase the company's conversion rate.

Uploaded by

Devanshi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
63 views11 pages

Lead Scoring Case Study Presentation

The document discusses building a logistic regression model to assign a lead score between 0 and 100 to help an education company identify the most promising leads and improve their low lead conversion rate of 30%. Various analyses were performed on the data including univariate analysis of categorical and continuous variables, bivariate analysis of correlations, and model evaluation to find an optimal threshold. The model achieved good performance with an accuracy of 80% on the test data showing it can effectively identify hot leads and increase the company's conversion rate.

Uploaded by

Devanshi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

By – Devanshi Sinha & Priyanka Solanke

PROBLEM STATEMENT
• An education company named X Education sells online courses to industry professionals. On any given day,
many professionals who are interested in the courses land on their website and browse for courses.
• X Education gets a lot of leads, however its lead conversion rate is very poor. For example, if, say, they
acquire 100 leads in a day, only about 30 of them are converted.
• To make this process more efficient, the company wishes to identify the most potential leads, also known as
‘Hot Leads’.
• After successfully identifying this set of leads, the lead conversion rate should go up as the sales team will
now be focusing more on communicating with the potential leads rather than making calls to everyone.

BUSINESS OBJECTIVE
• To help the company select the most promising leads
• Build a logistic regression model to assign a lead score between 0 and 100 to each of the leads which can be
used by the company to target potential leads where a higher score would mean that the lead is ‘hot lead’.
• The model should be able to adjust to if the company's requirement changes in the future.

Build a logistic Find the most potential


Current lead
regression model and leads (using the lead
conversion rate of the
assign lead score to score) to improve the
company 30%
each lead conversion rate
ANALYSIS APPROACH AND METHODOLOGY

Reading and Data Analyzing the Data Model Model


Understanding Cleaning Data Preparation Building Evaluation &
the data Deployment

• Shape • Checking • Univariate • Treating • Using • Evaluating


• Info null/missing Analysis of outliers automated the model
• Datatypes values all the • Creating feature using ROC
• Statistical • Handling variables dummy selection curve,
summary null values • Bivariate variables approach: accuracy,
• Removing Analysis of for RFE sensitivity,
redundant all the categorical • Building recall, etc.
columns if variables variables Logistic • Assigning
any • Correlation • Splitting Regression lead score
between data into Model to find most
different train and • Using promising
variables test dataset manual leads
• Scaling the approach of • Validating
continuous feature model using
variables selection test data set
using p- making
value & vif predictions
UNIVARIATE ANALYSIS: CATEGORICAL VARIABLES

• The lead count from lead origins 'API' and


'Landing Page Submission' are quite significant.

• Lead origin from ‘Lead Add Form' has a high


conversion rate (about 90%) however the count
of leads is very low.

• The source of leads coming from 'Google', ‘Direct traffic'


and 'Olark Chat' are more however their conversion rate is
low.

• Leads from 'Reference' and 'welingak website' have a very


high conversion rate but the count of leads is very low.
UNIVARIATE ANALYSIS: CATEGORICAL VARIABLES

• Most leads have last activity as 'Email Opened' and


'SMS Sent'.

• The 'SMS sent' category has a pretty high conversion


rate.

• Most of the leads are 'unemployed' but with a low


conversion rate.
• Working professionals have a pretty good conversion rate
but the lead count is quite low.
UNIVARIATE ANALYSIS: CATEGORICAL VARIABLES
• Students with last notable activity as 'SMS Sent' have
a high conversion rate.
• There are many leads with last notable activity from
'Modified' and 'Email Opened' but the conversion rate is
low.

• Most of the leads are from 'Mumbai'.


• Leads from 'Mumbai', 'Thane & Outskirts', and from 'other cities
of Maharashtra' have a pretty good conversion rate.
UNIVARIATE ANALYSIS: CONTINUOUS VARIABLES

After capping the outliers in 'TotalVisits' and 'Page Views Per Visit'
columns to their 95th percentile:

• The median value of 'TotalVisits' and 'Page Views Per Visit' is similar
for both converted and non converted leads.

• The converted leads spent more time on the website, as seen by the
significant rise in the median and IQR.
BIVARIATE ANALYSIS

From the given heat map, we can observe that


variable "Total visits" is highly correlated with
"Page Views Per Visit" with correlation coefficient
of 0.77.
MODEL EVALUATION

Precision and recall trade off Logistic Regression Model Parameters:


• By determining the area under the curve (AUC) of
ROC curve, the goodness of the model is defined. The
larger the AUC, the better the model. AUC for our
model is 0.88.
• Based on accuracy, sensitivity and specificity our
optimal threshold value is 0.34
• Based on precision and recall curve trade off we found
optimal cut off point as 0.41.
MODEL EVALUATION

EVALUATION METRIC TRAIN DATASET TEST DATASET


Accuracy 80.83 80.38
Sensitivity 81.07 79.47 F1 SCORE

Specificity 80.69 80.91 74.39


Precision 80.03 70.36
Recall 69.50 79.47

• Since CEO in particular, has given a ballpark of the target lead conversion rate to be around 80%. Our
Logistic Regression Model performs admirably and has a high level of sensitivity, which is exactly what
we need.
• The results in the table show that the model is not over-trained and is performing well on the test data set.
CONCLUSION

The X Education company needs to focus on the following factors to improve conversion rate of leads:

 The total time spent on the website impacts the conversion rate.
 Leads from lead origin 'Lead Add Form' have a high conversion rate.
 Leads both sourced from 'Olark Chat' also have decent conversion rate.
 Leads with lead source 'Welingak website' have a high conversion rate.
 Leads with last notable activity as 'SMS sent' have a high conversion rate.
 Leads who are working professionals have a pretty high conversion rate.
 Keep in mind leads with other last notable activities as: modified, page visited on website.

You might also like