Capital First is looking to expand to new cities in India. The document proposes using a recommender system approach to identify the top cities where Capital First should open new branches. The recommender system would analyze data on the locations of similar businesses to recommend cities that have potential based on the presence of businesses similar to Capital First. The system is then used to generate recommendations for new Capital First branch locations and to make predictions about where competitors like Bajaj Finance may open new branches.
Which city should Capital First open a branch next?
1. WHICH CITY SHOULD CAPITAL FIRST OPEN A BRANCH NEXT?
Priyank Aranke (aranke.priyank99@gmail.com)
An approach using cutting edge computational techniques
October 2016
2. THE PROBLEM
V. VAIDYANATHAN IN CAPITAL FIRST ANNUAL REPORT 2015-16
MARCH 31, 2016
CLEARLY THERE IS LOT OF OPPORTUNITY FOR MSME SECTOR IN TIER 2 CITIES …
“…the Cabinet approved a proposal for the introduction of
the Micro, Small and Medium Enterprises Development
(Amendment) Bill, 2015”
“India’s 50 million MSMEs and its fast emerging middle
class, with a differentiated model, based on new
technologies, provides a large and unique opportunity.”
3. WHERE SHOULD CAPITAL FIRST OPEN SERVICES NEXT?
THERE ARE 496 CITIES IN INDIA WITH A POPULATION OF OVER 1 LAKH.
CAPITAL FIRST IS PRESENT IN 43.
WHICH OF THE REMAINING 453 CITIES SHOULD CAPITAL FIRST OPEN NEXT?
SOURCE: HTTP://WWW.CAPITALFIRST.COM/BRANCH-LOCATOR
6. THE HIGH LEVEL APPROACH
IT’S CALLED RECOMMENDER SYSTEMS
▸ Why Amazon and Netflix recommendations are so good
▸ They have data on purchase history of millions of people
▸ So they can figure out people who have tastes similar to
you
▸ Then they recommend to you what ‘people like you’
have liked
7. THE HIGH LEVEL APPROACH
WHICH CAN BE ALSO APPLIED TO ‘RECOMMEND’ NEW CITIES
▸ Here’s how:
▸ Collect data on thousands of Indian towns and cities
▸ Find out the businesses who have locations similar to
you
▸ Recommend the locations which ‘businesses like you’
have discovered
8. FRANCHISE DATA
PROPRIETARY, HAND-COLLECTED AND CAREFULLY CURATED DATA ON 26
FRANCHISES AND 1496 CITIES IS INPUT TO THE RECOMMENDER ALGORITHM
Franchise No. of cities
Axis Bank 462
Bajaj Finserv 292
Cafe Coffee Day 220
Capital First 43
Domino’s Pizza 245
Dunkin’ Donuts 23
Eicher Motors 291
Equitas Mf 36
Gruh Finance 155
Hero MotoCorp 603
Hypercity 12
Inox Leisure 52
Janalakshmi Fin. 166
Franchise No. of cities
Kalyan Jewellers 59
Kotak Mahindra Bank 537
More Store 153
Ola Cabs 87
PVR Cinema 39
Repco Home Finance 102
Shoppers’ Stop 34
Sony Electronics 145
Sriram Vehicle Finance 794
Tanishq Jewellers 108
Toyota 220
Uber 27
V-Mart 104
DATA AS OF JUL–OCT 2016
9. THE RECOMMENDER ALGORITHM
RECOMMENDER ALGORITHM
THE RECOMMENDER ALGORITHM GENERATES TOP LOCATIONS WHERE CAPITAL FIRST
SHOULD OPEN BRANCHES - BASED ON LOCATIONS OF OTHER SIMILAR BUSINESSES
SEE REFERENCES SLIDE FOR TECHNICAL DETAILS ABOUT THE RECOMMENDER ALGORITHM
10. AND THE OUTPUT OF THE ALGORITHM IS…
DATA AS OF OCT 2016
RECOMMENDED NEW CITIES F0R CAPITAL FIRST TO OPEN BRANCHES NEXT
11. IN ADDITION TO RECOMMENDING WHERE YOU SHOULD OPEN THE
STORES NEXT, THE TECHNIQUE CAN ALSO BE USED TO:
▸ Find out which of the existing stores are under or over-performing
▸ The model outputs a score for each city which indicates the business
potential of that city. You can compare that score to the actual sales in
that city to determine whether the store is under or over-performing.
▸ Predicting which cities a given competitor would target next
▸ Since the recommendation engine works on publicly available data,
we can use it to predict the locations which a competitor would
target next. This will help you plan your response in advance.
▸ These predictions have worked in the past. See the next slides for
my past successful predictions and future predictions on where Bajaj
Finance will open its next branches.
MANY WAYS TO USE THIS TECHNOLOGY
12. IN AUG 2016, USING THIS APPROACH, I PREDICTED THAT BAJAJ FINANCE WOULD OPEN IN 25
NEW CITIES. BY OCT 2016, BAJAJ FINANCE HAD OPENED A BRANCH IN 22 OF THESE 25 CITIES.
SUCCESSFUL PREDICTIONS ON BAJAJ FINANCE BRANCH LOCATIONS
Predicted in
August 2016
Added in
October 2016
Agra ✓
Ambala ✓
Bhopal ✓
Dehradun ✓
Erode ✓
Goa ✓
Jabalpur ✓
Jalandhar ✓
Jamshedpur ✓
Jodhpur ✓
Kanpur ✓
Kolhapur ✓
Lucknow ✓
Predicted in
August 2016
Added in
October 2016
Ludhiana ✓
Mangalore ✓
Patiala ✓
Patna ✓
Raipur ✓
Rajkot ✓
Ranchi ✓
Salem ✓
Tiruchirappalli ✓
Amritsar —
Guwahati —
Mohali —
13. FUTURE PREDICTIONS ON BAJAJ FINANCE BRANCH LOCATIONS
PREDICTED NEW CITIES FOR BAJAJ FINANCE LOAN PRODUCTS
City Doctor Home Property Business Personal
Amritsar ✓ ✓ — ✓ ✓
Goa — ✓ — ✓ —
Guwahati ✓ — ✓ — ✓
Jalandhar ✓ ✓ — ✓ —
Kanpur ✓ — ✓ — —
Lucknow ✓ ✓ — ✓ —
Mangalore ✓ ✓ — ✓ —
Mohali — — ✓ — ✓
Mysore ✓ — — ✓ —
Patiala — ✓ ✓ ✓ —
Patna ✓ ✓ ✓ — —
Raipur ✓ ✓ — — —
Trivandrum — — ✓ ✓ —
DATA AS OF OCT 2016
14. DATA AS OF OCT 2016
FUTURE PREDICTIONS ON BAJAJ FINANCE BRANCH LOCATIONS – ON A MAP
15. TO KNOW FURTHER
▸ To get real time recommendations every month:
▸ Subscribe to my blog: https://chainsofindia.wordpress.com/
▸ Follow me on Twitter @aranke_priyank
▸ I would be happy to discuss the data and the algorithm behind the
model and how it can used in your business. Please feel free to contact
me at aranke.priyank99@gmail.com
Priyank Aranke (aranke.priyank99@gmail.com)
Thank you for your time.
16. REFERENCES
▸ Recommender Systems:
▸ https://en.wikipedia.org/wiki/Recommender_system
▸ https://en.wikipedia.org/wiki/Collaborative_filtering
▸ Data sources:
▸ Slide 2 – Capital First Annual Report 2015-16
▸ Slide 3 – 2011 India Census, Capital First branch locator
▸ Slide 8 – Respective Franchise websites
▸ Source code: https://github.com/priyankaranke/recsystemsforfranchise/blob/
master/Rec_systems_for_franchises.R
▸ Locations data (for 26 businesses and 1476 locations) available for reference by
request