Switching Costs: Why Customers Stay Even When They Could Leave

What does the argument rest on before any of it is worked out?

Four things sit underneath a switching cost argument, and they are worth having in view before a single figure is touched. Two of them are numbers. The third is one of those numbers turned into a length of time. The fourth is not a number at all, and it is the best evidence there is.

The first is a cost that is entirely real, is paid by the buyer, and appears in nobody's accounts. The asymmetry is the whole thing a switching cost turns on, so it is worth stating slowly. The seller did nothing when the buyer decided not to move, so the seller records nothing. The buyer avoided the cost rather than incurring it, and an accounting system has no place to put a payment that was never made, so the buyer records nothing either. An avoided cost leaves no trace anywhere, and that single fact decides how the rest of the subject has to be read: off behaviour, and never off a statement.

The second is four published figures and nothing else. Setu Bazaar, an invented marketplace, spends Rs 6,000/- of acquisition costWhat a business spends to win one buyer who was not buying before. How that figure gets assembled is worked out under customer acquisition cost. to win one buyer. Each buyer leaves behind Rs 2,000/- of contributionWhat one buyer leaves behind across a year once every cost that moved because that buyer existed has been taken out of it. a year. Divide the first by the second and the paybackHow long it takes for what a buyer leaves behind to add up to what winning that buyer cost in the first place. The answer is a length of time, not an amount. is 3.00 years. And its retentionThe share of last year's buyers still buying this year. On Setu Bazaar this is published as an assumption rather than as something anybody went and measured. is published at 80.00 per cent, stated as an assumption everywhere it appears rather than as a measurement.

The third is what that eighty becomes when it is written as a duration. At the published 80.00 per cent retention the average buyer stays 5.00 years. The 5.00 years is one divided by the churn the rate implies, so it is the same assumption wearing a different unit. The 5.00 years is never a second fact about how buyers behave, and it should never be quoted without the eighty standing beside it, in every sentence, without exception. The same rule holds under customer loyalty, where the rate itself is examined.

The fourth is a published fact with no number in it. Anjani Stationers Private Limited, an invented stationer, has supplied the Sunrise Public School group for eleven years, and its order bookThe orders a business has in hand but has not yet delivered. A book that refills without anybody chasing it is saying something; the question is what. refills each spring without anybody having to persuade the schools again. No figure for what moving would cost those schools was ever published anywhere.

Together, the third and the fourth set a duration against a duration, and only that allows a conclusion at all. 5.00 years of average life at the published 80.00 per cent, against a 3.00 year payback, leaves 2.00 years. Both sides of that comparison are lengths of time, and lengths of time subtract cleanly, so no discounting, no forecast and no rupee amount per buyer is needed.

What is a switching cost, and who actually pays it?

A switching cost is everything a buyer would have to spend, redo or relearn in order to move to a different supplier, whether or not any part of it ever reaches an invoice. The definition has no clever second half. The idea is obvious the moment it is said. The part worth working through is who carries the cost, and where it goes afterwards.

Start with a household rather than a business. A household has banked at the same branch for years. Nothing about the account is especially good. The queue is long, the app is slow, and the branch two streets away is offering better terms to anybody who walks in. Moving means new mandatesStanding instructions telling a bank to pay a fixed amount to a named party on a fixed date, so the payment happens without anybody remembering it. for the rent, for the school fee and for the electricity. Each of those can fail separately, each failure lands on a different afternoon, and every one of them is somebody's evening spent on the telephone. Nobody in that household is loyal. Everybody in it is busy.

The seller benefits from a cost it never pays and never records. The bank did not spend anything to create those three mandates and cannot point to them anywhere in its accounts, yet they are the reason the account is still there. Every switching cost has that shape. The advantage is genuine, it is doing real work, and it lives entirely on the other side of the relationship.

Naming who pays is the same discipline required under network effects, where an effect had to name a side before anybody was allowed to call it an advantage. A cost of moving has to name a payer, and the payer is the buyer, every single time. A claim that a business has switching costs, made without saying who would be paying them, cannot be checked.

What are the five things a buyer would actually have to do?

Five, and all five can be named rather than gestured at. A business choosing between suppliers, the question set out under the value network, faces the identical five from the opposite side of the invoice. Nothing is added to the list and nothing is dropped from it. Only the person doing the work changes.

Run them over something small enough to picture. A school is thinking of changing the printer who supplies its registers, and the printer down the road has quoted less.

Not one of those five appears on an invoice, and all five are paid in somebody's time. Time is why they get left out of the argument so reliably. A quote from the printer down the road is a number on paper that anybody can put into a comparison. Two afternoons of the school clerk's attention is not, so it silently gets counted as zero, and a decision that looked obvious on the paper turns out never to happen.

Notice also what is not on the list. Money already spent with the current printer is spent either way and changes nothing about the choice in front of anybody, so it does not appear anywhere. A switching cost is always about what would have to happen next, never about what has already gone.

Why does a buyer stay with a supplier it could leave?

Because of when the two amounts arrive, and not because of how anybody feels. Timing is the part of the subject that gets dressed up in language about relationships, and the plain version is better.

The cost of moving is paid at the moment of moving. All of it, in one block, before anything good has happened. The saving from moving arrives afterwards, spread thin across months, and only if the new supplier turns out to be as good as it looked from the outside. One of those is certain and immediate. The other is uncertain and deferred. A buyer facing a certain cost now and an uncertain saving later stays, and nothing about that is sentiment.

Here is the uncomfortable half. A great deal of writing about business treats a customer who stays as a customer who is pleased, and those are two different states that look completely identical from outside. The order arrives either way. The account stays open either way. Somebody reading the relationship from a distance cannot tell contentment from inertia, and quite often neither can the seller.

Picture a shopkeeper who knows perfectly well that the wholesaler two lanes away sells the same stock for less. The shopkeeper has even worked out what moving would take: one morning to open the account and get the credit agreed, and a second to learn which days that wholesaler actually delivers and how short deliveries get sorted out. Two mornings, both certain. Against that, a saving that starts small, depends on the new wholesaler being as reliable as the old one, and can be taken away by a single bad month. The shopkeeper stays, and would be able to explain exactly why if anybody asked.

What does a switching cost let a business do?

Pricing power describes what a charge change does arithmetically and then stops. Why any business might be able to raise its charge without losing buyers, or without losing the volume that goes with them, is a separate question, and four things decide it: the name the business carries, what it costs a buyer to move, how scarce the thing is, and rivalry. Of those four, the second is the switching cost: what it costs a buyer to move.

Here is the answer, and it is a bound rather than a licence. For the buyer, moving would come to more than staying, so a business whose buyers face a cost to move can raise what it charges by less than that cost without losing them. The whole of the mechanism is in that one sentence. The mechanism is not about goodwill, it is not about the strength of anybody's name, and it does not require the buyer to be happy about it.

The obvious next step has to be refused before it can be taken. Nobody has measured how large any switching cost is, so nobody can say how much any charge could rise. A bound nobody has measured is a direction and not an amount, and treating it as an amount is the commonest error made with this idea. A switching cost shows which way the room runs, not how big it is.

Whether a position allows a rise is one question. Whether a business ought to use it is another, and a switching cost answers only the first. A cost of moving says nothing about whether any charge is too low, or whether stickiness is a thing to go out and build.

The other three of the four are covered elsewhere. A name, and what it does before anybody prices anything, is taken up under Brand Equity: What a Brand Does Before Anyone Prices It. The rest of the field, and what it is doing, is taken up under Competitive Rivalry: How Intensity Shapes Industry Returns. Scarcity sits with Industry Types: How Sectors Behave Differently. And the arithmetic of a charge change itself, what a rise of a given size does to volume and to a total, belongs to Pricing Power: The Ability to Raise Price Without Losing Volume.

What do the years past payback have to do with any of this?

Past payback, stickiness stops being a description of a relationship and starts deciding whether a business gets anything out of it.

Setu Bazaar pays Rs 6,000/- to win a buyer, and the buyer leaves behind Rs 2,000/- a year. Do the division rather than reading about it: Rs 6,000/- divided by Rs 2,000/- a year is 3.00 years. The first 3.00 years repay a bill that was already paid, and earn nothing at all. A buyer who leaves at the end of the third year has cost exactly what it returned, and the business is back where it started, having done three years of work.

At the published 80.00 per cent retention the average buyer stays 5.00 years. Take the payback away from the life and 2.00 years are left over. Everything a buyer is worth to this business sits in the years after the third, and a switching cost is what decides whether those years happen at all.

The entire return lives past the third year, and the third year is a long way away, so a business can be excellent at winning buyers and still get nothing for it. Whatever holds a buyer through years four and five is doing all of the earning, and none of it is visible anywhere in the accounts.

Two cautions, both of which matter more than they look. The first is that the 5.00 years is not a discovery. At the published 80.00 per cent retention it is that rate restated in a different unit, so the eighty has to travel with it in every sentence. The second is that the measurement of retention itself belongs elsewhere. How that rate gets analysed, what a group of buyers taken together shows year by year, and what it costs a business simply to stand still are all set out separately under Customer Loyalty: Retention as an Economic Asset.

Does a long relationship measure a switching cost?

No. Careful people stop being careful at exactly this point, so getting it right is worth more than everything above it.

Anjani Stationers has supplied the Sunrise Public School group for eleven years, and its order book refills each spring without anybody having to persuade the schools all over again. Competitive rivalry names that relationship once, in words, with no figure attached, and leaves open what actually holds a buyer.

Eleven years quantifies nothing. Say exactly what it fails to say. The failure is specific rather than general. Eleven years does not say how large the cost of moving would be for those schools. Eleven years does not say what they would have saved by going elsewhere. And eleven years does not say whether anybody ever went and looked, the possibility most readings quietly rule out without noticing.

Eleven years does say one modest thing, and it is worth having. Whatever the cost of moving is, it has been larger than the reason to move, every year so far. Evidence of exactly the right kind, and silent about size, is the finding to carry away.

The same ruling holds on the other side of the invoice. Under the value network, the length of time an arrangement has been running describes the relationship rather than the dependency. Whether a business depends on a supplier is decided by what happens on the morning the arrangement stops, not by how many years it has run. A long customer relationship reads like an achievement in a way a long supplier relationship never quite does, so the pull is stronger on the buyer's side, and the ruling carries across unchanged.

A household version makes it obvious. A household has used the same chemist for a decade. A decade says something genuine about the chemist, about the walk, about how the household spends its Sundays. Nobody in that household has ever asked what a different chemist would have charged, so the decade says nothing whatever about that.

Is a buyer who stays the same thing as a buyer worth having?

The same customer settles this one. The Sunrise Public School group stayed eleven years, and the Sunrise Public School group stopped paying. Both of those are true at the same time, about the same relationship, in the same year.

A switching cost holds a buyer in place and says nothing whatever about the quality of the account. Holding and quality are two entirely separate properties, and nothing about a strong one implies anything about the other. A buyer can be extremely hard to dislodge and extremely difficult to collect from, and the very thing that makes the first true tends to make the second harder to act on.

The two get merged so constantly for honest reasons. Both are read off the same relationship. Both arrive in the same conversation. Both feel like signs of a healthy business, and neither one has an obvious label attached saying which question it answers. But one of them is about whether the buyer comes back and the other is about whether the money arrives, and no amount of the first ever produces the second.

A single buyer being a large share of a business, and how slowly that buyer pays, are set out separately under How to Analyse Customer Concentration and Dependence. Why buyers stay and what it means to depend on one of them are different questions, so neither figure belongs in a switching cost argument.

The everyday version is a tenant of eleven years who is three months behind on the rent. The length of the tenancy is exactly the reason nobody has done anything about it. The uncomfortable part is that the tenancy holds the tenant in place and makes the arrears harder to raise at the same time.

Where does a switching cost show up, and where does it not?

Nowhere. Not in a statement, not on a line, not in a note at the back. A switching cost is read off behaviour and never off a statement, and an avoided cost leaves no trace anywhere. So the behaviour that counts has to be named exactly.

Be precise about the behaviour. Vagueness at this point is exactly what lets a stickiness claim be attached to any business at all. Three things count. The reorder that arrives without a negotiation attached to it. The arrangement nobody has re-tenderedPutting a supply arrangement back out for other suppliers to quote against. A buyer that never re-tenders is not comparing anybody with anybody. in years, so nobody has priced the alternative even once. And the buyer who complains and does not move. A complaint that leads nowhere separates being pleased from being held, so the third is the strongest of the three.

The rule applies to Anjani Stationers as much as to anybody else. No rival's buyers are published, so not one of those three behaviours can be compared with a rival's buyers. Each one is measured against what the same buyer did in earlier years, and against nothing else at all. Naming the comparison is the discipline any claim of an advantage has to meet.

Nobody has published the size of what any buyer's move would cost, so there is no range to test a figure against, and the arithmetic that survives, one division and one subtraction, is small enough to be redone on the back of an envelope.

What has to be asked before believing a stickiness claim?

Four questions, asked in this order. A lender sizing up whether a borrower's order book would still be there next year, an analyst reading a note that calls a business protected, an investor deciding whether a run of repeat orders means anything, and a household choosing whether to change its bank all need the same four answers, and the order matters because the first two decide whether the last two can be asked honestly.

The third question is the one people skip, and a claim resting on duration alone has answered only the easiest of the four. Duration requires nothing more than a start date, so it is the easiest evidence in the world to find, and it is the least informative for exactly the same reason.

The same test applies to Anjani Stationers. The eleven years is a duration, so on the third question it scores badly. The order book that refills each spring without anybody having to persuade the schools describes what the buyer did rather than how long the buyer has been there, and that makes it a different kind of evidence altogether. Either kind standing alone would not be enough to say anything with, so both are needed.

What goes wrong when a duration is read as a measurement?

Something specific, and it is committed by people doing the job properly rather than by people cutting corners.