Traditional vs Alternative Credit Data, Side by Side

Traditional credit data is the record of how somebody repaid money they were lent, reported by lenders and held for lenders to read. Alternative data is anything else a lender reads that was generated for some other purpose. Accuracy is not what separates them. A person can see a credit record and has a defined route to correct it, and usually can do neither with the rest.

Search results on this question almost always start with the contrast, and that is the wrong end. A comparison of two things is unusable until each of them can be stated without looking at the other. So each side gets a complete definition first, one after the other, before a single row of contrast appears.

Most comparisons in lending end up ranking two options. Traditional and alternative credit data resist that, for a reason the arithmetic does not supply. The two are not rival answers to one question. Each is a record made for its own purpose, and each covers ground the other cannot reach.

What is traditional credit data, exactly?

Traditional credit data is the record of credit that was extended to somebody and how it was repaid. A lender advances money, the repayments arrive or they do not, and the lender reports what happened to a credit information companyA party whose business is holding records of credit that was extended and how it was repaid, and returning them to parties entitled to ask. Covered in its own right separately.. The credit information company holds what every reporting lender sent it, and returns the assembled picture to lenders who are entitled to ask for it. The loop closes: lenders put it in, and lenders take it out.

Three properties follow from the fact that this record was built to be read by lenders, and every one of them is a consequence of the purpose rather than an achievement of the technology. The first is shape. Because a lender was always the intended reader, the record arrives structured for that reading: an account, a size, a date, whether the repayments arrived when they were due. Nothing was ever aimed anywhere else, so nobody had to translate anything.

The second is coverage, and it is the harsh one. The record exists only where credit was previously extended. The mechanism that creates entries is borrowing itself, so there is no entry describing somebody who never borrowed. The missing entry is not a gap in the system's ambition. The definition is working correctly, and that limit is the whole reason the second kind of information ever got looked at.

The third is that what such a company may hold about a person, and which parties may be shown it, is set by the Reserve Bank of India at rbi.org.in, and it moves. The row further down carries the name of that authority where a value would otherwise go.

What is alternative credit data, exactly?

Most explanations go wrong here, and they go wrong by answering with a list. A list goes stale the moment somebody reads a record nobody had thought of reading. Alternative credit data is defined by where it came from: information that came into existence for some purpose other than lending, and is then read for lending. That definition holds whatever appears next year, because it describes the origin rather than the item.

Try it on something ordinary. When a payment moves through Setu Payments Limited, invented, a record of that movement exists. Three purposes made that record: the money has to move, both sides have to see that it moved, and somebody may need to trace it afterwards. Nobody created it in order to assess a loan. Read it later for that purpose and it is alternative data, and it was alternative data the whole time, from the second it was made.

Three properties follow from that origin, and they mirror the three above without being their opposites. The first is shape: the record was structured for whatever it was made for, so somebody has to translate it before a lender can use it, and every translation is a judgement made by whoever built it. The second is reach, and reach is the reason any of this exists. Such records are made by living rather than by borrowing, so they are generated for enormous numbers of people who have no credit record at all.

The third is where it sits. The alternative record rests with whoever generated it in the course of doing something else, and that party is not built to hold records about borrowing, does not answer to lenders, and may have no idea that a lender ever looked. When such a record was created, the person it describes was not the point. Nothing about it obliges anybody to make it visible or fixable to that person.

Why does what a record was made for decide everything else?

Both definitions are now on the table, so the contrast can start. The contrast starts with origin, and origin goes first for a reason that is not presentational: every other difference between these two kinds of information is downstream of the single fact that one was created to be read by lenders and the other was not. Get the order wrong and the comparison runs on consequences while the cause sits unexamined.

Look at what that one difference produces. A record is structured for its intended reader, so origin produces the shape. Whoever needed the record built or commissioned it, so origin produces the party holding it. A record made for a known audience arrives with the question of access already asked, so origin produces who may see it. A route for fixing an entry has to be built by somebody, and only a purpose that anticipated disputes builds one, so origin produces whether such a route exists. Origin produces how long the record lasts and whether it travels.

An everyday version makes the mechanism obvious. A wedding photographer's album and a caterer's delivery register both describe the same evening. The album was made for the household to look at, so it is arranged by moment, it is handed over, and if a name is spelled wrongly under a photograph somebody will fix it. The register was made so the kitchen could be paid, so it is arranged by dish and quantity, it stays with the caterer, and nobody is going to correct it on the household's behalf because the household was never its reader. Same evening. Two records. Every property different, and all of it decided before either was written.

What was this record made for? The answer to that single question forecasts the other six.

Who holds each record, and who gets to see it?

A credit record sits with a party whose entire business is holding it. The arrangement has a consequence people miss. Holding records is the business, so there is somebody definite to write to. The person described can obtain their own. There is an address, a process and a party whose job includes answering. None of that is generosity. Access of that kind is what happens when a record has a known audience from the day it is designed.

An alternative record sits wherever it was generated. The record might sit with a party that moves money, or with a party that sold something, or with a lending service providerA partner that sources or services a loan for a lender without holding the loan itself. The Reserve Bank of India sets what such a partner may and may not do. that assembled it on a lender's behalf. There is frequently no single place it lives and no party who thinks of themselves as its keeper. The person described may never learn that the record exists, let alone that a lender read it. Not knowing is a different situation from being refused access, and in some ways a worse one. Nobody can ask about something they do not know is there.

On what basis a lender may look at somebody's credit record at all, and what consent has to sit under any of it, is set by the Reserve Bank of India at rbi.org.in, and it moves.

If an entry is wrong, what is the route to putting it right?

Correction carries more weight than the other six criteria. Entries are wrong sometimes. A repayment lands against the wrong account, an account stays open on a record after it closed in the world, two similar names get crossed. None of that is unusual and none of it says anything about anybody. The question is not whether an error can happen. The question is what exists to deal with one when it does.

For a credit record, something exists. There is a defined route by which an entry is disputed and put right, and it runs through parties who are obliged to be part of it. How that route works, who carries the work of correcting an entry, and within what arrangements, is set by the Reserve Bank of India at rbi.org.in. The structural fact is the one that matters here: the route is there, it was built on purpose, and it was built because the record was always meant to be read about a person.

For an alternative signal there is usually no such route, and often nothing to point a route at. A signal read once inside a lending rule was never an entry, so it may leave nothing anybody could dispute. And a person who does not know a record was read has no reason to look for a route in the first place.

The comparison is not accurate against inaccurate but correctable against not correctable. A reader who ranks these two kinds of information on how well each one predicts has ranked them on the criterion that matters least to the person being described, because prediction says nothing about whether a wrong entry has any way out.

How far back does each one reach, and what does an empty field mean?

Depth and recency are two questions rather than one, and the two kinds of information answer them in opposite directions. A credit record reaches back as far as somebody's borrowing goes and not one day further, so it can be deep or it can be nothing at all, and which of those it is was settled by events that had nothing to do with information. Whatever generates an alternative record was not generating it five years ago, so such a record can often describe this month in detail and nothing further back.

One reading error here matters more than the rest. An empty credit record means no credit was previously extended, an empty alternative field usually means the lender was not shown one, and neither emptiness is a finding about the person. They are not even the same kind of nothing. The first is a true statement about a history. The second is a statement about what arrived at a lender's desk. What arrives depends on what was shared, what was collected and what a rule happened to ask for.

The machinery makes it easy to treat either emptiness as a finding, and that is the error to name early. A blank field sits in the same column as a filled one and looks like a value. A blank field is not a value. A blank field is the absence of information, and reading an absence as a conclusion is how a record about nothing becomes a statement about somebody.

Which record travels with the person to the next lender?

Portability is the criterion readers rarely reach on their own, and it is the one with the sharpest consequence. A credit record travels. Any lender entitled to look sees the same record, and portability is what makes a repayment history worth building at all: the effort a person put into repaying on time turns into something that arrives ahead of them at the next lender, without them carrying it.

An alternative signal frequently does not travel. The signal was read inside one lender's assessment, it may have been assembled specifically for that assessment, and the second lender has no claim on it and often no way to obtain it. So the value that signal created belongs to the lender that read it rather than to the person it described, and somebody assessed well on one may have to be assessed from nothing at the next lender.

Put it in ordinary terms. A street vendor who has traded from the same corner for eleven years is known to everybody on that street. Move the cart two kilometres and none of that reputation moves with it. The reputation lived in the memory of people who stayed behind. A credit record is the opposite arrangement: it lives somewhere that is not any one lender, so it goes wherever the person goes. Portability is the whole of the difference, and it costs something real to whoever is on the wrong side of it.

What is any of this actually worth to the lender?

A fair question is why a lender should care which record it reads, and the honest way to answer it is with the lender's own accounts rather than with a claim about people. The worked figures below belong entirely to the lender.

Rukmini Finance Limited, invented, lends and takes no deposits. Rukmini Finance carries assets under managementThe total size of the loan book the lender is running, measured at a stated date. Here it is the base every ratio in this block is struck on. of Rs 18,000 crore for the stated year, funded by Rs 14,400 crore of borrowings and Rs 3,600 crore of net worthWhat is left of a business after everything it owes is subtracted from everything it holds. The part of the funding that belongs to the shareholders rather than to a lender., and those two funding lines add to the Rs 18,000 crore exactly. The book earns 14.50 per cent a year, or Rs 2,610 crore, and the borrowings cost 8.50 per cent a year, or Rs 1,224 crore. The remainder after the funding is paid for is Rs 1,386 crore, or 7.70 per cent of assets under management for the year.

Now the two lines that matter here. Running the place costs Rs 540 crore, or 3.00 per cent of assets under management. The credit costThe charge a lender books in a stated year against loans it does not expect to be repaid in full. The charge is a line in the lender's own accounts for that year, and nothing else. for the year is Rs 396 crore, or 2.20 per cent of the same base. Surviving all three is profit before tax of Rs 450 crore, or 2.50 per cent of that same Rs 18,000 crore for that same year.

Put those last two readings beside each other on the one base and the point lands: the credit cost line reads 2.20 per cent of assets under management and the whole of the profit before tax reads 2.50 per cent of it, so the credit line is 88.00 per cent of the size of the profit line. That is why the quality of a lender's information is a live question for the lender at all. The question is not a matter of principle. The credit line is the second largest number in that year's account.

Does that hold at a lender built the other way round?

A claim built on one institution is a claim about one institution, so it is worth checking at a second. Suvarna Commercial Bank Limited, invented, is a bank and takes deposits, so it is built quite differently. Its provisionsThe charge a lender books in a year against advances it does not expect to recover in full. The same kind of line a finance company records as its credit cost. for the year come to Rs 1,800 crore against profit before tax of Rs 3,000 crore, so the credit line there is 60.00 per cent of the size of the profit line. Different institution, different construction, and the same qualitative fact: the credit line is a large piece of the profit line at both.

Now watch what happens when the base moves. Profit before tax at Suvarna Commercial Bank is Rs 3,000 crore against total assets of Rs 2,40,000 crore, and that reads 1.25 per cent. At Rukmini Finance the same line reads 2.50 per cent of assets under management. On that base Rukmini Finance is the higher of the two.

Change nothing except the base. Measure the same profit line against net interest incomeWhat the lending produced in a year after what the funding cost is taken out of it. Settled earlier in this material and used here rather than explained again. instead. Suvarna Commercial Bank's Rs 3,000 crore against its Rs 7,440 crore reads 40.32 per cent. Rukmini Finance's Rs 450 crore against its Rs 1,386 crore reads 32.47 per cent. On the asset base Rukmini Finance is ahead and on the income base Suvarna Commercial Bank is ahead, and neither lender changed by a rupee between those two sentences. The ranking was never a property of the lenders. The ranking belonged to the denominator. A ratio has to carry its base in the same sentence as its value for exactly that reason.

Where does this material stop, and what does that make each figure?

Six things about this comparison cannot be reached from the figures above, and naming each one and classifying it is more useful than filling it. Each row below says whether what can still be said is a ceiling, a floor, or simply unknowable, and gives the reason the classification holds.

Notice that two of those six absences still yield something usable. Classifying an absence does that; listing it does not. A ceiling and a floor are both real information. Unknowable is also real information, and it is the answer to four of the six.

Where does each one stop, and why are they not rivals?

Here is the closing pair, and both halves are limits rather than criticisms. A credit record cannot describe somebody who has not borrowed. The record does not describe them badly. The event that writes an entry never happened, so there is nothing there. An alternative record can very often describe that same person in detail, and it cannot be checked, corrected or carried by the person it describes.

Read those two sentences together and the whole ranking question dissolves. Each one stops exactly where the other keeps going. Traditional and alternative credit data are not two answers to one question. The two are records made for two purposes, and a lender reads both for precisely that reason: neither one covers the ground the other covers.

Which leaves something better than a winner: a test that can be run on any record anybody produces, in one question. Ask what the record was made for. Everything else follows from the answer, including which of the two kinds of information the record is, who is holding it, and whether the person it describes could ever find out that it exists.