Cyclone Ditwah and the 88 Percent Protection Gap: Building a Risk-Financing Stack for Under-Insured East Coast Corporate Assets in 2026

When Cyclone Ditwah tracked across the Bay of Bengal and slammed into Sri Lanka in late November 2025, it caused total economic losses of roughly USD 4 billion, of which less than USD 0.5 billion was insured. The World Bank's rapid damage assessment put Sri Lanka's direct physical damage at about USD 4.1 billion, close to 4 percent of that country's GDP, with all 25 districts affected. Swiss Re ranked it among the costliest Asia-Pacific natural disasters of the year. Strip the headline down and you are left with an insured-to-economic ratio of barely 12 percent. Roughly 88 paise of every rupee of loss landed on balance sheets, governments and households, not on insurers.

That figure is not a Sri Lankan outlier. Swiss Re's analysis of India showed uninsured economic losses of around USD 33 billion across 2018 to 2022, with insured losses near USD 1.1 billion, meaning over 90 percent of catastrophe exposure went uncovered. The same Bay of Bengal cyclone tracks that hit Sri Lanka regularly graze Tamil Nadu, Andhra Pradesh, Odisha and West Bengal. The corporate assets sitting in that corridor (ports, thermal and renewable generation, refineries, warehousing clusters, auto plants) carry the same structural under-insurance.

The practitioner argument of this post is simple. For catastrophe-exposed sites on the east coast, a single indemnity property policy is no longer a defensible risk-financing answer. It needs to become one layer in a deliberately built stack.

Indemnity property and fire cover is the right instrument for most of what it does. The problem is the specific way it behaves under a wide-area catastrophe, where its weaknesses stack up at exactly the wrong moment.

The first failure is sum insured adequacy. Under-insurance is endemic in Indian commercial property books, and the average clause then bites hardest on the largest losses. A plant carried at depreciated or historic book value rather than reinstatement value can see a catastrophe claim cut proportionately right when cash is scarcest.

The second failure is exclusions and sub-limits buried in the policy wording. Flood and storm surge, the actual destructive mechanism in most Bay of Bengal cyclones, frequently sit under a sub-limit far below the declared sum insured. Many corporates discover post-event that their nat-cat sub-limit caps recovery well short of the material damage bill.

The third failure is time. Wide-area events overwhelm surveyors. Loss adjuster capacity is finite, and a regional catastrophe puts hundreds of claims into the queue at once. A business interruption claim that takes nine months to settle does not help a company meet payroll and debt service in month two.

None of this means indemnity cover is wrong. It means a corporate that relies on it as the only financing source has effectively self-funded the timing risk, the basis risk on sub-limits and the adequacy risk, without pricing any of it. A financing stack makes those decisions explicit.

A risk manager who has read three catastrophe claim files knows the policy that pays is rarely the policy that was sold. The structural answer is to stop asking one instrument to do everything.

Think of catastrophe financing the way a treasurer thinks about a capital structure, in layers ordered by cost and trigger speed rather than as a single policy.

The working layer is deliberate retention. A higher voluntary deductible on the indemnity policy lowers premium and, more importantly, signals that the corporate has its own first-loss capacity. For a group with multiple sites, that retention is better held inside a captive than left as a bare excess.

The traditional transfer layer is the indemnity property and fire policy, sized properly on reinstatement value, with the nat-cat sub-limit negotiated up to a defensible figure rather than accepted as printed.

The parametric layer sits alongside, not inside, the indemnity tower. It pays on a measured trigger (wind speed at a reference station, or storm surge or rainfall depth) regardless of assessed damage. Its job is speed and liquidity in the first weeks, covering the costs indemnity cover settles slowly or excludes.

The capital-markets layer (cat bonds and insurance-linked securities) is for the largest, rarest peak exposures, typically only relevant for very large groups or pools.

The pre-funded reserve is the floor: a ring-fenced cash or contingency reserve sized to the retained layer, so the captive or balance sheet is not improvising liquidity mid-crisis.

Ordering the layers

1. Quantify the modelled loss curve for each site (1-in-50, 1-in-100, 1-in-250 year).
2. Set the retention the balance sheet can genuinely absorb without breaching covenants.
3. Buy parametric for the early-liquidity slice the indemnity tower will not deliver fast enough.
4. Buy indemnity for the bulk repair and BI bill.
5. Pre-fund the reserve to the retained layer, and only then consider capital-markets transfer.

The stack is not about buying more cover. It is about matching each rupee of exposure to the cheapest instrument that actually pays when that loss type occurs.

Parametric insurance is the layer that has matured fastest in India, and it is the layer most under-used by corporates. IRDAI's use-and-file route has let insurers bring parametric products to market without long product-approval cycles, and public insurers including New India Assurance have launched named climate parametric offerings. State-level pilots (a parametric rainfall payout in Nagaland through SBI General, coastal community pilots) have proven that triggers fire and pay.

For a cyclone-exposed corporate, the appeal is mechanical. A parametric policy defines a measurable trigger (sustained wind speed at a named meteorological station crossing, say, a defined threshold, or storm surge height, or 24-hour rainfall depth) and pays a pre-agreed amount when the index breaches that level. There is no survey, no loss adjuster negotiation and no proof-of-loss cycle. Money can move within days, which is precisely what the indemnity tower cannot promise after a wide-area event.

The honest part of the pitch is basis risk. The trigger may fire when the actual site damage is modest, or worse, the site can be wrecked while the reference station reads just below the threshold. That gap is the cost of speed. You manage it, you do not eliminate it.

• Tighten station selection. Use a reference station genuinely correlated to the site's exposure, not the nearest convenient one.
• Use a multi-trigger or stepped payout so a near-miss on one index does not zero the recovery.
• Size parametric to the liquidity gap, not the full loss. It funds the first weeks; indemnity funds the repair.
• Document the basis-risk decision so the board understands parametric is a complement, not a replacement.

Used this way, parametric stops being a novelty line and becomes the working-capital bridge that keeps a cyclone-hit operation solvent while the indemnity claim grinds through.

Most large Indian corporates already self-insure catastrophe risk, they just do it badly, as an unpriced residual sitting under sub-limits and deductibles. A captive turns that accidental retention into a managed, accounted line.

The mechanics matter for the east coast case. A captive (onshore through GIFT City, or via established Bermuda or Singapore structures) holds the retained working layer, charges the operating entities an actuarially set premium, and accumulates reserves over good years to fund the bad one. That accumulation is the point. Instead of absorbing a 1-in-50 cyclone loss as a sudden hit to a single year's P&L, the group has been pre-funding it through the captive across every clean year.

The captive also becomes the buyer of the layers above it. It can purchase reinsurance or fronted parametric cover on terms that reflect the group's actual loss experience, rather than the broad-brush rate the open market applies to a coastal manufacturing risk. For a conglomerate with sites across Chennai, Vizag, Paradip and Haldia, that aggregation and self-rating is where the economics turn.

The pre-funded reserve is what most stacks skip, and it is what separates a real captive from a deductible with a board. Size the reserve to the retained layer's modelled loss, ring-fence it, and treat any drawdown as a financing event rather than a discretionary spend. A captive that holds risk but never funds against it has simply relabelled the exposure without retiring any of it.

GIFT City has lowered the friction for onshore Indian captives, and fronting arrangements let a captive sit behind a licensed insurer where direct placement is impractical. The decision is no longer whether the group can run a captive, but whether it can justify continuing to retain catastrophe risk informally when a structured alternative is available.

The 2026 market gives brokers room to push this agenda. Property rates have softened in much of the commercial book, which means capacity is available and underwriters are willing to talk structure rather than just price. Use that window deliberately.

On wordings, three points carry most of the value:

• Move the sum insured basis to reinstatement value and document it, so the average clause cannot gut a catastrophe claim.
• Negotiate the nat-cat (storm, flood, surge) sub-limit up to a figure backed by your modelled loss curve, and get it stated as a named peril limit, not buried.
• Pin down the business interruption indemnity period and the consequential loss definition. A 12-month indemnity period is often too short for a port or plant facing a long catastrophe rebuild and supply-chain recovery.

On pricing, present the stack as a total cost of risk, not a stack of separate premiums. A higher retained layer plus parametric plus a tighter indemnity tower can land at a lower all-in cost than a single over-broad policy carrying expensive low-frequency capacity it rarely uses.

On placement, sequence it. Bind the indemnity tower and the retention first, then place parametric to fill the early-liquidity gap, then size the reserve. Running them in parallel without an agreed loss curve produces overlap and double-paid capacity.

The broker who walks in with a modelled loss curve and a layered structure is selling risk engineering. The one who walks in with a renewal quote is selling a commodity, and will be priced like one.

A financing stack is only as good as its behaviour on claims day, so the governance has to be built before the season, not after the warning.

Start with trigger documentation for the parametric layer. The reference station, the index source, the threshold and the payout schedule must be agreed and on file. The whole value of parametric is that there is nothing to argue about post-event, so any ambiguity in the trigger definition destroys the speed advantage you paid for.

For the indemnity layer, pre-loss valuations and a current asset register decide how fast a surveyor can work. The single biggest avoidable delay in a catastrophe claim is the absence of agreed values and documentation. Sites that have done pre-loss surveys settle materially faster, because the loss adjuster is verifying a known position rather than building one from scratch.

Governance also means deciding, in advance, how the layers interact:

• Which layer funds the first 30 days of cash needs (parametric, then captive reserve).
• How the captive accounts for a drawdown and rebuilds the reserve afterward.
• Who has authority to declare a financing event and release reserve funds.
• How parametric proceeds reconcile against the eventual indemnity settlement, so the group is neither double-recovered nor short.

The board-level case is straightforward once framed correctly. Ditwah showed that 88 percent of a catastrophe loss can fall outside insurance. A financing stack is the mechanism that decides, deliberately and in advance, which parts of that exposure the company transfers, which it pre-funds, and which it knowingly retains. That is a defensible risk-management decision. Discovering the answer through an uninsured loss is not.

The east coast corporate that builds this stack in the 2026 soft market is buying structure while it is cheap. The one that waits will negotiate it after the next cyclone has already repriced the entire coast.