Seen from above — the canopy everyone is standing under
The world got riskier and less protected at the same time. That should be the best news the protection industry has had in a generation. It is not — because the people who price everybody else's risk have never once priced their own — and it is now deciding how much of the market they can reach.
Two letters landed in Singapore households this year.
Same household. Same quarter. Income got less certain, protection got more expensive and thinner — in the same postbox, within weeks of each other.
If you work in insurance, you wrote the second one.
And the first one is exactly why the second one lands the way it does. The rest of this piece is about why that has to be so — and why it does not.
Whenever the world wobbles, our industry talks about "macro uncertainty" as though it were weather — something that happens to the business rather than something that arrives on the books with a claims code attached.
It is not weather. Here are five examples of what the last two years of turmoil actually turned into — each one an underserved need with a product code attached, and nowhere near the full list.
This is what gets missed when the industry reads the news as background noise: the turmoil is the demand. Fear of losing your income is income protection. Fear of the bill is health. Fear that your cover no longer keeps up is indexation, top-ups, and a conversation somebody should be having with you. The world has rarely been better at manufacturing reasons to buy what we sell.
So the numbers should be going our way.
They are not.
Every line of protection tells the same story, and it is not the story you would expect from a decade of rising anxiety.
| Line of protection | Measured as | Underserved by | Trend |
|---|---|---|---|
| Health, Asia | Premium-equivalent shortfall | US$258bn | 21% wider than 20177 |
| Mortality, Asia | Premium-equivalent shortfall | US$132bn | 35% wider than 20177 |
| Critical illness, Singapore | Share of household need uncovered | 74% of need | 91% for platform workers5 |
| Natural catastrophe, global | Value of exposure uninsured | US$424bn | ~73% of all exposure12 |
| Natural catastrophe, emerging Asia | Share of exposure uninsured | ~95% | almost none of new build covered12 |
Need went up, coverage went down — not because people stopped caring, but because of a mechanism. When money is tight, protection is the first thing a household trims: it is the only purchase where nothing visibly bad happens when you stop paying. When claims cost more, insurers price up and tighten terms — correctly, prudently, and toward less cover for the same money. Employers facing double-digit medical trend raise co-payments and narrow networks. And when work turns informal, group cover does not follow the worker out of the building.
Every one of those responses is individually rational. Together they produce something absurd: the umbrella gets smaller exactly when it starts raining.
None of it, though, is a pricing problem. Every risk in that table is well understood and competently priced.
Which leaves one obvious question. If the need is this large, this urgent and this well documented — why doesn't the industry simply sell these people something smaller, more bespoke and better fit-for-purpose?
Before diagnosing why, it is worth being precise about what. Strip out the industry vocabulary and every household in that gap is asking the same three questions — and none of them is about risk transfer.
Those three failures look unrelated. They are not. Behind them sit two capabilities, and between them they draw the boundary of an insurer's addressable market: affordability — the minimum premium at which a policy can be written and still make money — and customisability — how far a product can be reshaped for the life in front of it. Affordability first, because it decides who is even in the room.
Below a certain premium, a policy is not worth making.
Not because the risk is bad. Because of an arithmetic that has nothing to do with risk at all. Quoting, underwriting, issuing, servicing, collecting, chasing, renewing and eventually paying a policy costs roughly the same whether the premium is S$900 or S$9. The risk scales with the sum insured. The work does not.
So every insurer operates with a floor, whether or not it has ever written the number down: a minimum viable premium, below which a policy costs more to run than it will ever earn. Above the floor is a business. Below the floor is a rounding error with a compliance obligation attached. In unit-economics terms the floor is the cost-to-serve — and it decides affordability long before any actuary touches the rate.
Who lives on the wrong side of that line? Not only the households who cannot pay. Also, in far greater numbers, those who could — but do not need the package on the shelf: the platform worker whose cover should track an irregular income; the contractor who needs six months, not a whole-life commitment; the household that trimmed cover and would take back a smaller slice; the small business that buys nothing because nothing is sized for it. Some are priced out. Many more are simply not catered for. Neither group is uninsurable. They are uneconomic to serve at the size and shape they actually need — a different problem, and a far more fixable one.
This is why awareness campaigns rarely close a protection gap on their own. Six in ten consumers in emerging Asia intend to buy within the year, with the perceived price of cover among the leading barriers — not ignorance, price.6
The height of that floor is a direct product of three layers. Only the first of them even looks like money.
So the height of the floor — and with it the size of the protection gap — is not really a statement about customers at all. It is a direct product of those three layers: how the business is structured, what its systems allow, and how long its people take to decide. Lower any one and the floor drops; lower all three and it drops a long way. What that buys is not a cheaper product but flexibility in what a product can cost — cover that adjusts monthly, cover that attaches to a gig rather than a job, cover that costs S$9 and still makes money. Nothing about the risk changes. The addressable market does.
Where the floor sits is therefore the first thing that decides an insurer's market reach. There is a second, and it is audited even less often.
The floor decides who an insurer can afford to serve. The mould decides what it can offer them — and it is cast by the same three layers.
A product that takes eighteen months to develop, ships in one shape and is repriced once a year is not really a product. It is a mould — and not only of the product. The underwriting rules, servicing path, claims workflow, partner integration and reporting are each hard-wired to the product they were built for, so reshaping the product means reshaping the operation by hand. Customisability runs through both: what can be built, and how it can be run. Every customer who does not fit the mould is either turned away or sold something that does not fit — and the second is more common, and more corrosive, than the first.
The industry standardised for a reason. For most of its history affordability and customisability pulled against each other: every variant was a build, every build had a cost, and one shape for everyone was the only way to keep the floor low. The trade-off was real — but it was a consequence of how the engine was built, not of insurance itself. An engine built as one piece has to be rebuilt to make a different shape. An engine designed and built as modules — cover, rules, price, eligibility, channel, servicing path, claims path, each a component that can be recombined — makes a new shape by assembly, and the hundredth shape costs what the first one did. Every other industry has a name for this: mass customisation, and it was always a design achievement rather than a spending one.16 Insurance is the last industry to reach it, and it is reachable now, because a policy is rules all the way down. What such an engine looks like, and who is building it, is the subject of Part 2.
Every other consumer business now builds and prices dynamically — watch, learn, change the offer, watch again. Insurance holds the richest behavioural data of any industry and uses almost none of it to shape what it sells. Four levers, one result: a product sized, priced and assembled to the customer in front of it.
Singapore has already run part of this experiment and published the result. In 2025, online direct channels produced 10.0% of all new life policies sold here — and 1.2% of weighted new business premium.15 The figure says two things. Customers will buy protection unassisted, at scale, when it is simple and quick to reach: one policy in ten is now bought that way. And what they buy is small, because small and simple is all the online shelf carries — the one place the industry has, almost by accident, lowered its floor. What it does not yet show is whether those buyers would take a larger or better-shaped policy if one were offered, because none has been. The ratio is not a verdict on the customer. It is an inventory report on the shelf — and very probably the tip of a much larger market.
All four levers need the same underlying capability: not a new product or channel, but the ability to see what is happening, decide, and change the product quickly — exactly what the three layers of the floor prevent. None of it means pricing free of oversight; a configurable product still files, still passes a fairness test, still reserves against its tail. What changes is the cost and time of each variant, not the rules it must meet.
Affordability and customisability: efficiency in serving a client, and effectiveness. Market reach is the product of the two. Built as one piece, they trade off. Built as modules, they do not. Two engines, one bearing — and nobody has measured the bearing.
The usual consolation is that every incumbent industry is slow. The record says otherwise.
Every business selling to the public has the same fixed-cost problem: somebody must take the order, handle the money, service the account and sort it out when something goes wrong — and that costs roughly the same whether the transaction is large or small. Every one of them had a floor, and every one of them sold a standard package. One by one they lowered the floor and broke the mould — and none did so without knowing what one unit of their own work cost.
| Industry | Affordabilitythe floor moved | Customisabilitythe mould broke | Unit costtracked |
|---|---|---|---|
| Telecoms | Monthly contract → prepaid top-up, per-second billing | One plan for all → plans assembled from the handset | Cost per subscriber |
| Retail | Minimum order → one low-value item, next day | Fixed range → offers personalised to the buyer | Contribution per order |
| Banking | Minimum balance → no-minimum, near-free transfers | Branch product set → credit sized and priced to the customer's data | Cost per account |
| Airlines | A route worth flying → seats priced to live demand | One fare → unbundled fares, every seat priced differently | Cost per seat-km |
| Insurance | A premium worth writing — unchanged | Annual products in fixed shapes — largely unchanged | Expense ratio — a share of premium, not a cost |
Telecoms is the closest analogue, because both levers moved at once. Prepaid lowered the floor: a customer who could not commit to a contract could buy ten dollars of airtime and be a subscriber that afternoon. Dynamic plans broke the mould: data, minutes and add-ons assembled from the handset, priced to usage rather than a tariff card. Cost per subscriber fell while the number of people who could be a subscriber rose — one programme of work, and market reach was the result. The operators that did it were incumbents, on billing systems every bit as old as a policy administration system. The analogy holds for the operating model, not the risk model — an operator carries no tail, an insurer must — but the floor and the mould were never about the tail.
The part worth copying is not the cost cutting. None of them treated affordability and customisability as a choice: the system that knows a parcel's cost to the cent is the system that knows what to offer next. Operating leverage in its purest form — lower cost per unit, and more units.
Insurance measures itself differently. An expense ratio falls when premiums rise, so it can improve in a year when the work got worse. It tells you how heavy the business is; it cannot tell you whether the next policy is worth writing. A weighing scale being used as a ruler.
I went looking for the other number — a published cost to serve one policy for one year, tracked over time, the way retail tracks cost per order. It does not exist. Not a gap in the research. That is the finding.
That peak was built on rate increases, a mild catastrophe year and investment income on high yields — the rate cycle and the bond market, not the operating model. Premium growth is forecast at 1.3% for 2026.13 The rent is running out, and nothing underneath has changed.
The industry knows. Three-quarters of insurers are targeting a 10% cost reduction by 2030 and a third are targeting more than 20%. A quarter say they have been highly successful at hitting cost goals. About a third say their cost objectives are even well defined.14
Meanwhile the customer, who has been shopping in all four of those other industries, has learned to keep score in days.
Nobody formed that expectation by comparing insurers with each other. Which leaves one question, and it is the one nobody in the building can answer.
Our industry prices things nobody else will touch.
A life. A container ship in a war zone. A satellite launch. A pianist's hands. A cyber-extortion event using a technique nobody has invented yet. We quote it to three decimal places, reserve against it, reinsure it, and stress it against a hundred-year event.
There is exactly one risk the industry has never priced. Its own.
Not credit risk, not market risk — those have models, committees and capital held against them. I mean a whole category the industry has simply never turned the instrument on: the risks sitting on its own side of the table.
The test is simple to run. Ask the actuarial team for the probability that a 42-year-old non-smoker dies within the year: an answer in seconds, three decimal places, with a confidence interval. Then ask what it costs, administratively, to keep that same person's policy in force for twelve months — servicing, collection, changes, correspondence and renewal, fully loaded. A pause. A range. An offer to come back to you.
There are three such risks, and between them they explain why the floor sits where it does and why the mould has never been broken.
Every insurer has an underwriting policy: hundreds of pages, committee-approved, defining which risks it will not accept. Every insurer also has a second policy — one that decides which people it cannot afford to serve at all. Never drafted, never approved, and nobody can tell you what it says. It is not a document. It is those three unpriced risks, compounding quietly.
The fair objection: none of this is hidden. Anyone who has spent a year inside an insurer could list all three from memory — and a blind spot everybody can see is not much of a blind spot.
It survives because the system rarely asks for it. No regulator requires it, no statement discloses it, no auditor tests it. Where somebody inside the building has asked — and somebody usually has — the answer had nowhere to go: no standard, no benchmark, no committee that owns it. Which is why, when I went looking for a published figure earlier, there wasn't one.
And the reason that actually holds it in place: every individual in that room is measured on something else entirely. Not an accounting problem, and no longer a technology problem — something harder to put in a report. Part 2 is about what is arguably the most important root cause of all of it — and about the one engine that all five of the people behind it could say yes to.
Every risk looks like a risk from a distance. The one you are standing inside just looks like the way things are done.
Step back far enough to put a number on it, and it stops being how things are done. It becomes something you can move.
We price the whole world to three decimal places. The three numbers that decide who we can protect — what our own work costs, the market we cannot reach, the price of waiting — we have never priced at all.
Insured against everything but this.
Answer for your own organisation. Nothing is sent anywhere — this runs in your browser and we never see it.
We price strangers to three decimal places. Our own cost-to-serve, not at all.
The obvious conclusion is that this is an engineering problem. The engine was built as one piece; somebody should rebuild it in modules.
The industry has drawn that conclusion before. It is what every core-replacement programme of the last twenty years set out to do, and a good number of them are still running. I have spent my career inside large institutions across several industries, and the programmes that stalled rarely stalled on the engineering. They stalled in the room where the decision was taken — five wide-awake people, each with an excellent reason, none of them wrong. Between them they have been carrying these three unpriced risks for twenty years. Unreserved, uncapped, and personally.
Part 2 is about those five, and about the one engine all five could say yes to — modular without having to be rebuilt, because it sits beside what is already there. It closes with a single page that puts these three risks in front of all five at once. Fair warning: one of the five is you.
This piece was written by human intelligence. For now.