Bajaj Finance · Prep

The Professor's Questions

(These are the key points of the questions as jotted down in my notes, not the professor's original wording.)

Compete with Amazon?
Decisions that made digitalization work
Customer-focused?
Are customers loyal?
Takeaways and what I will do

My Answers

Q1. Compete with Amazon?

Yes. Compete as the credit layer, not as a second Amazon. If the choice is "mine the wallet" or "rush into smaller towns," I would do both, in that order, and I would refuse a third option the case keeps flirting with: turning the app into a general store.

Amazon wins the next tap. BFL wins the credit decision and the repayment. Those are different games. BFL should not spend its 13.9% net interest margin fighting a price war it cannot fund.

Why yes

BFL already sits at the checkout. The case says it moves 30% of electronics and 15% of smartphones in India. By August 2021 it had $21.2 billion of assets and 50.5 million customers. Amazon Pay Later had 2 million customers.
Online is still a thin slice. In 2020, e-commerce was 5.8% of Indian retail, and Amazon had 36% of that e-commerce. 0.36 × 0.058 ≈ 2% of all retail. The other 98% still runs through stores, which is where BFL's 114,000-point network lives (Exhibit 5, June 2021: 114,450).
The profit pool is the spread, not the shelf. Exhibit 8: BFL's NIM is 13.9%, against 4.25% at HDFC Bank and 3.5% at HDFC Ltd. Market cap was $42 billion on $16 billion of advances. HDFC Ltd's market cap was $62 billion on $78 billion of advances. Investors were paying for the franchise, not for book size.
Lending is a regulated, capital-heavy business. An NBFC cannot take demand deposits, so BFL's funds are already expensive. That is a weakness on price and a moat on skill: underwriting, collections, and fourteen years of cleaning a book. Net NPA went from 7% in FY2008 to 0.12% in FY2012, and was 0.75% in FY2021 even after the overdue definition tightened from 6 months to 3 months.

What I would do, in order

Protect the moment of purchase. Make a half-finished application continue from the store counter to the app and back. The person already standing at a BFL counter should not finish the next purchase inside Amazon.
Treat the 10% wallet as profit, not as a substitute for growth. These 50.5 million customers account for 45% of India's mortgage book, 65% of its credit cards, and $100 billion of personal loans, while BFL holds about 10% of their wallet. That gap is real. It is not the whole growth plan. Cross-sell franchise only grew from 24.1 million to 27.4 million in two years (Exhibit 6, Q1 FY2020 to Q1 FY2022), while the firm adds 6 to 7 million new customers a year. Durables are replaced slowly. Panchal is right: cross-sell will not replace new acquisition.
Enter the uncovered 38% of households with the urban risk engine, not with franchisee enthusiasm. Distribution covers 62%. Rural points are already mostly franchised: 1,218 of 1,745 rural locations in June 2021, about 70%. No credit bureau, weaker franchisee service, and customers new to formal repayment can replay 2008. That year they shut 330 locations and dropped 14,500 dealers. Urban durable stores fell from 19,000 (FY2008) to 2,000 (FY2010). I would cap franchisee-originated credit until its delinquency matches owned branches.
Keep shopping, bill pay, and travel as reasons to open the app. They are frequency, not businesses. Exhibit 11: 52% of online durable buyers come for low price, 46% for convenience, and 90% prefer an aggregator to a brand's own app. BFL loses a price war. It can win "0% credit + next-day delivery + a warranty" on the purchase that needs finance. Pamnani's rule still applies: try things, do not collect hobbies. If an add-on cannot clear the same bar as a real product line, shut it.

The cost of this choice

Amazon may plant a flag in some small towns while BFL moves second. Wallet-deepening will look slow, because a refrigerator is not bought every quarter. I would rather lose a town for a year than buy FY2021's lesson twice: provisions rose from $519 million to $816 million, and ROE fell from 20.2% to 12.8%. In FY2019 it had been 22.5%. The operating ratio actually improved, to 30.7% of net interest income. Credit cost, not overhead, is what broke the year.

What I would not do

I would not try to out-Amazon Amazon on assortment, logistics, or price. A marketplace that fights Amazon also fights BFL's own retailers. And I would not "only mine existing customers." Six to seven million new customers a year are the growth math. Cutting that off hands the next town to whoever shows up with a phone.

Q2. Decisions that made digitalization work

The app in 2019 was the last chapter, not the start. Seven decisions made the earlier years work. Each one had a trap next to it. BFL missed the traps because it had already been burned, and because it paid people to cooperate instead of ordering them to.

Clean the book before you scale it. After the crisis, net NPA was 7%. They shut 330 locations, stopped 14,500 dealers, fired several hundred people, stopped sharing the manufacturer's subsidy with retailers, and killed personal-computer loans — borrowers were selling the computers and disappearing. The trap is to grow your way out of bad credit. They did not, because Nanoo Pamnani said he would shut the business, and a newly spun-out firm had to show shareholders a clean win. Urban durable stores went from 19,000 to 2,000, then were rebuilt to 26,900 by June 2021. The cut was the strategy.
Change the game to same-day joy, not to a lower price. Loan approval took 3 to 5 days; rivals did it the same day. Project WOW cut it to 3 hours. Bajaj's line: if parents can tell their children the television will be installed that day, word of mouth does the rest. By March 2008 they were booking 621,000 durable loans a year across 19,000 dealers. The trap is speed without rules, which is how the book had reached 15% non-performing assets, three times the industry. They put the rules in software. Later, a rule change could go out to the whole merchant network with one click. On Diwali the system flags red if point-of-sale processing is down for five minutes, because they originate 300,000 applications that day.
Add only businesses large enough to deserve the effort, and kill the rest. "A Mercedes, not a Ferrari or a Ford." Pamnani: try everything, avoid hobbies. They ended above 50 products, but they also exited infrastructure lending — repayment depended on the government paying on time, not on the customer's own risk — and they wound down life-insurance sales through the field force when productivity failed. A young business had to grow 100% a year. A mature one had to grow at least 20%, or fix the model, or close. The trap is Olay's trap: a pile of products and no kill rule.
Make customer data a company asset, and pay the team that collected it. In 2015 Jain forced every system into one warehouse. "Giving up customer data was like giving away the crown jewels." The trap is the silo. Two things blocked it. It was top-down: people who said "our market is different" were told to join the competition, and Mody says nobody left. And there was a price: the business that uses another unit's customer pays an internal fee. Durable and mobile-phone finance originated 90% of customers and earned 35% of its profit from that fee. Sharing became revenue, not a favor.
Let vendors supply the limbs. Keep the brain. Technology partners provided software and hardware. BFL's own people owned process design and underwriting rules. The trap on one side is a vendor's process you cannot change. The trap on the other side is building everything yourself and spending years on capital projects. They wanted the brain without the heavy capex.
Split origination, risk, and collections, then give risk and collections joint ownership of losses. Day-to-day approval is not the same job as watching the portfolio. If losses rise, risk tightens the rules and collections gets better at the same time. They call a customer right after a check bounces, "when he still feels a bit ashamed." Retailers' approval rates are published and auto-cut when their customers go bad, so the dealer sends better people to the counter because he wants larger, faster approvals. A home-loan customer, they found, has one-fourth the delinquency of someone without one, and gets a better price or a larger line. The trap is the sales team owning the growth number and hiding the bad loans.
Digitize the process only after you have deleted steps, and put the designers inside the business. In November 2019 they framed an "omnipresent" model because income and cost were both growing in a straight line, and they feared cost would soon outrun income. The history does not yet show that: rupee operating expenses compounded at 29% for the fourteen-year span, income at 36%, profit after tax at 51%, and operating expenses fell from 58% of net interest income to 31%. The fear was about the next chapter. Covid made it urgent. Jain told his top 100 leaders to imagine they had a billion dollars of capital and a smaller economy, and to redesign their function. The CTO's admission matters: as a B2B company they had never really built the B2C layer. They plotted every offline process, cut what did not add value, then put it on the app. User-experience people sat inside the businesses, because "business originates ideas." The trap is a digital lab next to an unchanged company.

They had three different adults in the room: Sanjiv Bajaj, the owner; Pamnani, a retired chairman of Citibank Asia; Jain, the operator. They also had a scar. A firm that has already seen 15% bad loans does not treat risk as a slide in the appendix.

Q3. Customer-focused?

Not yet. BFL is excellent at the customer's moment of purchase. It is not organized around the customer. It is organized around products, risk, and ROE.

What looks customer-focused

WOW was designed around a parent's same-day joy, not around a rate card.
The Rs 1 scheme, against a normal 33% down payment, pulled 2,000 people to one southern retailer on the first weekend. People walked out with a 43-inch television instead of a 32-inch, or a double-door fridge instead of a single-door. The customer spent more because the entry fee collapsed.
"People who are not desperate for loans seek good service." That is why they pushed salaried personal loans up to Rs 2 million at the front line with the algorithms in the back, and became the number-two player in a business owned by banks.
The app was built so a dropped journey can resume on another channel. On the durables marketplace a customer can pick from over a million items, take a 0% loan, and get delivery in 24 hours.

What the organization actually is

Each product is its own business unit. The head is a mini-CEO for profit, sales, risk, and people. Technology and operations are the only things they must share. That is a product company with a common data pipe.
Cross-sell is a push. A separate, highly paid team sells the next product to people risk has pre-approved. Existing customers cannot walk up to that team and ask for another loan. The firm decides when the customer is allowed to buy more.
The CTO said they had been a B2B company and had not focused on the B2C layer. After Covid, Jain told the top 100 that the answer was to accelerate digital and "become a consumer-focused company." You do not announce that you are becoming one if you already are.
The sharpest line in the case is Jain's: the stupidest thing a company focused on ROE can do is pay twice for the same customer. That is an acquisition-cost sentence. The customer is an asset you should not buy twice. It is not a person whose next need you are organized to hear.
Collections are effective and not warm. They call when the customer is still ashamed. I would not romanticize this. It is one reason the book survived. It is also not what "customer-focused" usually means.

My line in class: BFL is customer-competent, not customer-centered. Amazon calls itself the earth's most customer-centric company. BFL calls its ambition omnipresent, and manages itself to ROE. The 2019 app is an attempt to close that gap. It is not evidence that the gap was never there.

Q4. Are customers loyal?

No. They are loyal to the offer, and some of them are loyal to access. They are not loyal to the firm. The 10% wallet share is the evidence.

Where repeat behavior is real

The EMI card, from about Rs 5,000 (around $70), lets a repeat customer check out like a credit-card holder without new paperwork. For people who cannot get a bank card, it is often the only card they have. That creates repeat purchase. I would call this access loyalty: they come back because the other doors are closed.
The cross-sell pool is large in absolute terms: 27.4 million of 50.5 million customers, about 54%, sat in the cross-sell franchise in June 2021. Something is being sold twice.

Where loyalty fails

The front door is a 0% loan. The case says even people who can pay in full do not mind taking it. That is deal loyalty. The day another counter offers the same 0%, the reason to choose BFL is gone.
The wallet number is decisive. These customers already hold 45% of India's mortgages, 65% of its credit cards, and $100 billion of personal loans. BFL's share of that wallet is about 10%. Where the customer has a real alternative, nine rupees in ten sit somewhere else. Exhibit 8 puts home finance at only 12% of BFL's own loan mix. BFL is not the mortgage lender behind that 45%.
New relationships do not become second relationships at the same speed. The firm adds 6 to 7 million customers a year. The cross-sell franchise grew by about 3.3 million over two years (24.1 to 27.4). Total franchise grew from 42.6 to 50.5 million in that same window, about 7.9 million net. Gross adds and net adds are not the same series, so I would not call the gap "churn" in class. I would say the direction is clear: most new customers are not showing up in the cross-sell pool.
About 30% of customers added each year are new to credit, mostly from smaller markets. That is a good acquisition machine. It also means the book is constantly refilled with strangers. A loyalty compound would not need that high a share of first-timers to keep growing.
Online, the customer is loyal to price and convenience. 52% and 46% in Exhibit 11. Rewards, travel, and bill pay are there so the app gets opened. The loan itself does not create a habit. A durable sits in the house for years.

Class line: people repeat what is easy and unavailable elsewhere. The moment it is easy elsewhere, the repeat stops. BFL's loyalty is real at the locked door. It is unproven at the open door.

Q5. Takeaways and what I will do

What I take

Winning the transaction is not the same business as winning the gap between two transactions.
A digital app on top of siloed customer data is just the silo, moved online. The warehouse in 2015 mattered more than the app in 2019.
Speed and standards have to be separate jobs, and both have to own the loss.
People share customers when sharing pays them. A speech about "one firm" does not do it. The 35% profit from transfer pricing did.
You can be very good at the customer's moment and still not be a customer-centered firm. Look at whether the customer can pull the next product, or only receive a push.
Do not fight a platform on the shelf it already owns. Fight on the decision it cannot yet make. For BFL that decision is credit. For us it is judgment and trust.

Near term — this quarter

Write down where a client's truth actually sits today. In our own language, the failure mode is an empty platform: the memory lives in one person's head, and when the person leaves, the client leaves. Before any new interface, I want one client record, one owner, and a rule that the person who brought the family in is credited when someone else does the next piece of work. That is BFL's transfer price, translated. Success looks like this: a second colleague can brief the family without re-collecting the story, and the originator is not punished for the handoff.

Mid term — this year

Draw the current client journey on one page. Delete the steps that do not change a decision. Only then put what remains into a system. BFL plotted the offline process before it coded the app. I also want origination speed and judgment standards split, the way approval was split from risk. Publish a short quality gate the front line can see, so people select better work instead of being told no at the end. And write the kill rule before a new service launches. Pamnani's "no hobbies": if it is not large enough to deserve a real effort, it does not get a team. Success looks like fewer steps in the journey, and at least one service we stop because it missed a standard we wrote in advance.

Long term

Omnipresence for a family office is not an app with extra products on it. It is being present at the moments that actually move a family — a liquidity event, a succession, a cross-border decision, a next generation coming in — in person and in the system. The natural cycle is generational, which is even longer than a refrigerator. If we do not design frequency on purpose, quarterly reviews, governance, education, the relationship goes dormant, and the next platform meets the family at the next event. The product rule stays the one in our own blueprint: the product serves the allocation, the allocation does not serve the product. Success looks like a rising share of the family's important decisions made with us, not a longer menu.

How these do not stay as intentions

One metric for "the second decision with us," next to new names. BFL had 10,000 metrics. I do not want that. I want a few that would have shown the 10% wallet early.
A person whose job is to stop the thing if trust or quality slips. That was Pamnani on Tuesday at 8:00. Without that chair, the franchisee expansion will always have a reason.
Economics for the originator. If sharing a client cuts someone's number, the warehouse will stay a speech.
A date. Belden's lesson still applies here: a pay mix and a kill rule that have no date will not happen.