banking on S Anand

Visualisation - activities to centralise

Wed, 20 Sep 2006 12:00:00 +0000

Surely we don’t have many activities to centralise? We already have a central hub for processing operations!

We heard that from a fair section of our client organisation. They initially had operations spread across their branches. Some years ago, they had established a central hub and many regional hubs. Yet,

Only a few prominent operations were centralised. Others were just regionalised.
Regionalisation was inconsistent. Some branches still did these at their own premises.
Branches still did the bulk of the work.

We made a list of activities, surveyed all their branches and hubs, and got a good sense of which activities were happening at branches vs regionally vs centrally.

Rather than make a list of these activities (they numbered over 300), we put the variwide chart to an unorthodox use. The chart below shows the activities on the x-axis, and the extent of centralisation on the y-axis.

The graph actually consists of thin vertical lines, one for each activity. The height represents the number of branches for which the activity is happening regionally. For the activities on the right, they’re happening at branches. Dark blue lines are happening centrally. Light blue lines are regionalised.

You can see at a glance that about 55% of activities are at branches, 35% are regionalised and 10% are centralised. Clearly there’s a big potential to centralise. Once we showed this slide, most of the objections went away.

Comments

Sudheer 21 Sep 2006 5:01 am: Hi Anand, Without taking the credit away from analysis, te most amazing part of your analysis is the representation. I guess the simplicity and the novelty of the picture. One thing which you could elaborate a little more is to tell us how you identified the list of activities so that they are of similar significance/effort etc. Given that branches are the last mile in terms of delivery there will be a lot more of smaller activties compared to the HO.
S Anand 21 Sep 2006 7:07 am: You’re right, Sudheer – that’s an important question. We assumed that the size of activities is proportional to the size of assets at the operations. So the vertical axis is not the number of branches, actually. It’s the asset size. This is not a perfect assumption, but for a homogenous bank like ours, a fairly good one. That takes care of the size of activities.

Normalising non-normal distributions is bad

Wed, 19 Jul 2006 12:00:00 +0000

I was working with the treasury of a bank. They were trying to estimate how much money could flow out of their savings account in a day, worst case.

I took their total savings account balance at the end of each day and found the standard deviation. I took thrice the standard deviation, and said, “You can be 99.7% sure that your daily loss won’t be more than 1.5% of the balance.”

That would be right if it were a normal distribution. But it’s not.

Banks have millions of savings accounts, each of which is like a random variable. But unless they’re independent, and they have finite standard deviations, the central limit theorem won’t work.

Firstly, savings account transactions are not independent. If there’s a run on the bank, they’d all pull out their money. Whenever a company declares dividend, a large number of savings account are credited. Salary accounts are credited at the end of the month. As a rule of thumb, you could say that if one savings account goes up, the others are likely to as well.

Secondly, savings account transactions are not normally distributed. If you take a single savings account, you won’t find a bunch of debits and credits. Every month, you’ll find one large credit for the salary, one mid-sized debit for monthly expenses, and several small debits for individual transactions (bills, ATM, etc.) Once in several years, you’ll find a gigantic debit (purchase of car or house, wedding, etc.) or a gigantic credit (retirement / pension fund, sale of property, etc.)

As a result, the savings account is likely to fluctuate a LOT more than if it were a normal distribution.

If I had just looked at the data, I’d have found several occurrences of fluctuations greater than 1.5%. The normal distribution predicts that there should be fewer than 0.3% of such cases. That’s about 1 per year. I’d have visually been able to spot nearly one a month. I’d also have been able to spot the huge 4% swings that do happen once in a few years.

People wiser than me have made the same mistake. I was interning at Lehman Brothers when they were planning to launch a new electronic bond-trading product. My task was to trace the bond price movement.

The data we had was bad. Many bonds jumped as much as 40% in a single day, due to data errors. The bulk of my task was to clean out these errors.

After cleaning up, there was still two jumps that couldn’t be explained. I went to my boss, who recognised them at sight. One was a sudden drop in price of all Government bonds in December 1998. The other was a 32% drop in price of Hikari Tsushin – a mobile phone retailer – on the day they went bankrupt.

We concluded that the daily price drop wouldn’t be more than 9%, to a 95% confidence level. If that was right, a 32% drop in one day would happen once in a million years. Yet, we had Hikari Tsushin just the previous year.

We didn’t bother about it. In fact, we didn’t even think about it. If we’d checked, we’d have found that the daily price drop was closer to 12% or something, to a 95% confidence level.

Summary: Force-fit a normal distribution on non-normal data can understate the worst-case scenario. Often you’re better off just inferring confidence levels from the raw data than from a fitted distribution.

Sourced statistic from: www.forex.academy

Comments

Sriram 24 Jul 2006 7:53 am: what you are talking about are outliers which would be there in even the most normal of the normal distributions. One always has to take them out. This is not an “un-normal” phenomenon.
S Anand 24 Jul 2006 11:20 am: No, I’m not. The normal distribution asserts that the probability distribution falls of exponentially. These transactions are power-law distributed. The distribution falls off as a power law. Specifically, this means that the normal distribution has a finite standard deviation. Power law distributions have an infinite standard deviation. You can check this from the wikipedia articles on power law distributions.
RaviM 3 Aug 2006 7:09 am: When I was doing Risk Management for a mutual fund, we were facing the same problems. So we used to take two approaches to calculate VaR for various products- Assuming Normal distribution of stock price changes(the problem here is that you are not sure what is the current Std dev of any Stock… which can change drastically on a single day in case of events. I still remember Mastek which dropped 50% on a single day). We also used to take historical data to figure out what are the past occurances when such events has happened. but here the disadvantage is that we assume that stock price movements are historically dependent, which is also not correct
Naresh 21 Sep 2010 8:45 pm: Complex problems have simple, easy to understand wrong answers. The world does not run on a normal distribution. If we could predict the future behaviour (even heavily populated datasets) using a normal distribution, the world would be a different place, wouldn’t it? The answer to this problem is to use a time weighted average of data with the latest data having the highest weight, or using the exponentially weighted moving average formula developed by Riskmetrics. Use a lambda of .94 for daily observations. You can alternatively use an empirical lambda. Use it to predict earthquakes (?), junior miners, venture investments, come what may. And then you will find that too is inadequate as a risk tool.
The Curse of the Bell Curve – Part 2 | Jigsaw Academy | Training for careers in Analytics 16 Jan 2013 8:55 am (pingback): […] in Checking/Saving account – This is an interesting article written by a friend during his consulting days. He illustrates how we tend to blindly apply the […]

ATM breakeven

Fri, 13 Jan 2006 12:00:00 +0000

Banks install ATMs to lower their branch costs, and to attract new customers. When working out the economics of ATMs, we found that lowering branch costs alone could not be a viable reason to install an ATM.

The bank argued as follows:

“Every time someone withdraws money from an ATM, they avoid going to the branch. With enough people going to the ATM, I can afford not to increase my branch size, and that saves me money. Since it costs me Rs 20 every time a person withdraws cash (in terms of salary, rent, etc.) and an ATM costs about Rs 2,200 a day, I’ll break even if there are 110 cash withdrawals from the ATM.”

The argument misses a crucial point: every ATM transaction does not replace a branch transaction. People visit ATMs more frequently than branches, thanks to them having smaller queues and being open 24 hours. As a rule of thumb, people visit ATMs twice as often as a branch to withdraw cash.

A teammate didn’t believe me. We argued.

“When I used the branch, I would withdraw money for the entire month at the beginning of the month. I continue the same with an ATM.”

“But I withdraw cash whenever I need money. And in smaller chunks. Sometimes, I just withdraw Rs 200. That way, I get to carry less cash too.”

“Ah, you may be the exception, as always. Very well, I will find out.”

He went to a fairly representative branch, and asked them how much money would people withdraw before their ATM was installed. Since ATMs impose a limit of Rs 15,000, he discarded transactions above Rs 15,000. The answer was: people used to withdraw about Rs 3,600 every time they came to the branch. Then he asked, what’s the average ATM withdrawal. Answer: Rs 1,900. In other words, people seemed to withdraw only half as much from an ATM as from a branch. (And therefore, on average would withdraw twice as often every month.) My teammate was finally convinced.

So, in order to break even, the ATM must be used about 220 times a day, not 110 times. This is nearly impossible. ATMs are used mostly in peak hours: morning while travelling to work, during lunch, and evening when travelling back to work. Apart from these hours, the ATM is practically unused. This gives roughly a 4-hour window. The time between two ATM transactions is at least a minute. So a very busy ATM might be able to make the 220 transactions in that time. Most ATMs will not.

In fact, we found that only 4 ATMs managed to break even, among their 250. The cost-saving argument alone is difficult to justify an ATM.

Comments

Jayant 15 Jan 2006 1:31 am: Good argument. This is typical of any service/utility. When the cost comes down, people use the service more frequently. Banks never make money on ATM. They are used for customer stickiness. In fact, retail customers keep accounts which has highest ATM density. Now that ATMs can take deposit, interesting will be to see %age of money kept by bank in ATMs and received by ATMs as deposits vs branches. Then answer what is the right mix of cash to be kept and at what frequency ATMs need to be cleared? My suspicion is banks will make money on float is ATMs are cleared more often
Arun 15 Jan 2006 2:08 am: Interesting analysis. Another thing that interests me is these inter-bank ATM tieups. HDFC does not charge ATM charges for salary account holders if they withdraw from other bank ATMs (if they use the visa debit card, that is). Normally, if you withdraw from a different bank’s ATM, the charge is Rs. 50. When I was in Meerut, HDFC’s ATM was in some godforsaken place, and I used to withdraw at least once a week from a different ATM. And I wasn’t being charged either. I must be the most loss making ATM customer for HDFC :-)
Madhu 15 Jan 2006 3:13 am: Also in some sense the money kept in the ATM is locking of capital for the bank on which it could earn some amount of interest. How is that taken in to account in the cost?
Madhu 16 Jan 2006 6:16 am: Just a thought, how is the cost variable (per transaction)? would that also not be a fixed cost, say salary of a teller or something like that?
Jayant 16 Jan 2006 9:54 am: Madhu, if I am not mistaken, it is the largely rent (in a mall, etc) for the ATM (assuming bank owns the ATM; electricity, etc should be small). Anand, correct me if wrong.
Madhu 16 Jan 2006 11:54 am: That is for the ATM, I meant for the bank branch. Also there will be a particular variable cost in accessing info from an ATM as well.
S Anand 17 Jan 2006 7:59 am: Arun, the inter-ATM withdrawal is usually not a problem. Some banks enter into agreements not to charge each other. Others charge Rs 10 or so. Given the low frequency of such usage, and especially only by high profit customers, banks usually don’t mind the small charge here.
S Anand 17 Jan 2006 8:18 am: Madhu, the money in ATMs is like money in a bank branch. It stays on the banks accounts, and can be “lent” overnight, notionally. The bank doesn’t lose the interest.
S Anand 17 Jan 2006 8:22 am: Regarding the branch cost, over 85% of the cost is manpower-driven (i.e. salary and rent). But that’s not to say it’s a fixed cost. Thanks to an ATM, a branch can move some of tellers out, So an ATM can save real money, even for apparantly fixed costs.
Arun 17 Jan 2006 11:59 am: Ah, ok. Yeah, I guess the frequency of inter-bank transfers isn’t really that much. And if they don’t charge each other, then it’s not an issue at all. Not everyone’s as lazy as me! ;-)
S Anand 17 Jan 2006 2:34 pm: If Bank X’s customer withdraws from Bank Y’s ATM, Bank X typically pays Bank Y Rs 10. But Bank X has (notionally) saved Rs 10 because an ATM transaction anyway saves the bank Rs 10. So banks that don’t charge customers for inter-bank ATM withdrawals aren’t doing their customers that much of a favour, actually.
Arun 17 Jan 2006 3:50 pm: But they do charge Rs. 50 normally, which is quite high, really. It’s only in some salary accounts (depends on the company, i guess) that they don’t charge. So, i would think they are either paying the other bank more than Rs. 10/- or they are making money off the customer for not using their own ATM.
Arun 17 Jan 2006 3:52 pm: In other words, I think they are actually doing a favour to salary account holders (probably because salary account holders keep more of their money lying with the bank and are more profitable in other ways)
S Anand 17 Jan 2006 4:29 pm: The banks I worked with were paying only Rs 10, so guess the banks are skimming their non-salaried customers. Or creating an artificial feature for the account.
Arun 18 Jan 2006 4:31 am: Ah, Ok. Then methinks it’s a bit of both.
Kakul 15 Mar 2006 8:24 pm: Did you consider the money the bank will make on float? due to less amount being withdrawn
S Anand 24 Mar 2006 6:29 am: Yes, we did look at increased float. But it was negligible. In fact, I was finding float to be marginal in several instances. See my post on demand drafts on 23rd March, for instance.
Rahul 3 Nov 2010 11:00 pm: Hi Anand, Great Work! Can you tell when does a typical bank branch break even? What are costs associated with setting up of a branch? Building, Salary of employees, etc How many years does a branch take to reach its peak of acquiring business (deposits+loans)? How does it reach profitability? Thanks
Prasanna Bontha 18 Jul 2014 6:24 am: The costs of setting up an ATM can reach its break even if the up time of the ATM is maintained at its highest. The location of the ATM can also fetch interchange charges as well.

Keep your dollars

Fri, 16 Aug 2002 12:00:00 +0000

Now, if you earn dollars, you can keep the dollars. Interesting that this move comes at just the time when the dollar is falling.

Banking and finance portal

Thu, 25 Jan 2001 12:00:00 +0000

A neat banking & finance portal by Northern Light. Covers mergers, reforms, technology, regulations, country-specific info, etc.