Optimising Self-Insured Retentions and Deductibles for Indian Corporates 2026

Every insurance programme draws a line between the loss the company keeps and the loss the company transfers to the insurer, and where that line sits, the deductible or self-insured retention (SIR), is one of the most consequential and most under-examined decisions a corporate risk buyer makes. This is a different exercise from designing exotic risk-transfer vehicles: the lever here is the attachment point on an otherwise conventional policy and the actuarial reasoning behind where to set it, not the choice of structure. Set the retention too low and the company pays the insurer to carry losses it could comfortably absorb itself, which is expensive because the insurer charges not just for the expected loss but for the cost of carrying it. Set it too high and the company exposes itself to retained losses it cannot comfortably fund. Optimising the retention is finding the level that minimises the company's total cost of risk at a level of retained volatility it can bear, and for a large corporate the saving from getting it right is material.

The core insight is that the lower layers of risk, the frequent, smaller losses, are the most expensive to insure relative to the loss they transfer, because the insurer prices them with a loading for expenses, capital and profit on top of the expected loss, and the loading is proportionally heaviest on the predictable working layer of losses. A company that retains this working layer pays the expected loss directly (which it would have paid through premium anyway) but avoids the insurer's loading on it, and uses insurance for the layer where it adds the most value: the large, infrequent, volatile losses the company cannot absorb. So the logic of retention is to keep the predictable losses the company can fund and transfer the unpredictable losses it cannot, which is the opposite of buying first-rupee cover with a token deductible.

The distinction between a deductible and an SIR matters operationally even though both retain loss. A deductible is netted from the claim: the insurer handles and settles the claim, controls the adjustment and any litigation, and pays the insured the loss above the deductible, with the deductible reimbursed back to the insurer or borne by the insured. The insurer stays in the file from the first rupee. A self-insured retention is structurally different: it sits below the insurer's cover as the insured's own primary layer, the insured handles, reserves and pays losses within the SIR itself, often through its own claims team or an appointed third-party administrator, and the excess insurer's cover attaches only once a loss pierces the SIR. The trade is control for responsibility: an SIR gives the insured authority over reserving philosophy, defence counsel and settlement strategy on its own layer, but also the duty to fund and administer it and to report into the excess insurer so the attachment is recognised. The choice between deductible and SIR, and the level at which either sits, are the two levers the retention decision works, and on long-tail liability the SIR's control over claims handling can matter as much as the premium saving.

Optimising the retention is not a one-time exercise either, because the right level changes with the company's loss experience, its financial capacity, its appetite, and the price the market charges for the lower layers. A retention that was right when the market was soft and cheap may be too low when the market hardens and the lower layers become expensive to transfer, so the retention should be revisited at each renewal against the current price of risk and the company's current capacity. The sections that follow set out how to find the right level, how the deductible types behave, what the retention does to the company's finances, and how captives and alternative risk transfer extend the same logic.

At the centre of the retention decision is a trade-off the company can quantify: as the retention rises, the premium falls, and the question is whether the premium saved is worth the additional retained loss the company takes on.

How premium responds to retention

Raising the retention removes the lower layer of loss from the insurer's cover, and because that lower layer carries the heaviest expense and capital loading, the premium falls by more than just the expected loss in the layer given up. The insurer was charging the expected loss in the layer plus a loading for handling, capital and profit, and when the company retains the layer it saves the whole charge, keeping the expected loss to fund itself but escaping the loading. This is why raising a retention from a low level often saves premium out of proportion to the loss retained: the company gives up cover on losses it would mostly have paid for through premium anyway, and saves the loading on top.

The saving is largest in the lower layers and diminishes higher up. Moving the retention from a token level to a meaningful one captures the heaviest loadings and saves the most per rupee of retention; moving it higher still saves less per rupee, because the layers being retained are less frequent and carry less loading. There is therefore a level beyond which raising the retention saves little premium while adding a lot of retained exposure, and the optimisation is to find the point where the marginal premium saved no longer justifies the marginal retained loss and volatility.

Weighing the saving against the retained risk

The premium saving is certain; the retained loss is uncertain, and the company has to weigh a certain saving against an uncertain cost. The right way to do this is not to compare the premium saving to the expected retained loss alone (which would almost always favour higher retention, since the insurer's loading guarantees the premium exceeds the expected loss), but to weigh the premium saving against the expected retained loss plus the cost of the volatility the retention adds. A company that can comfortably absorb the retained layer's variability values that volatility cheaply and should retain more; a company for which a bad loss year in the retained layer would strain its finances values the volatility dearly and should retain less. The optimisation is therefore specific to the company's financial capacity and risk appetite, not a universal answer, and two companies with the same loss profile can rationally choose different retentions because they can bear the retained volatility differently.

Optimising a retention requires knowing how much loss the company would retain at each candidate retention level, and that comes from analysing the company's loss history, which is what burning-cost and loss-modelling techniques do. Without this analysis the retention decision is a guess; with it, the company can quantify the trade-off the previous section described.

Burning-cost analysis

Burning cost is the foundational technique: it takes the company's historical losses over a period (typically five or more years) and calculates what those losses would have cost the company at a given retention level, by applying the proposed retention to each historical loss and summing what the company would have retained. Run across a range of candidate retentions, the burning-cost analysis shows how the company's retained loss would have behaved at each level, which is the empirical basis for choosing the retention. A retention at which the historical retained losses are stable and affordable is a candidate; one at which a single historical year's retained losses spike to an uncomfortable level is too high for the company's capacity.

The historical losses have to be adjusted before they are used, because raw history misstates the future. Losses are adjusted for inflation (a loss from five years ago would cost more today), for changes in the company's exposure (if the company has grown, its loss frequency would scale), and for any known changes in the risk (a major loss-prevention investment, a change in operations). The adjusted loss history is a better predictor of the future retained loss than the raw history, and the quality of the adjustment is part of the quality of the analysis.

Modelling the volatility, not just the average

Burning cost gives the expected retained loss at each retention, but the retention decision turns on the volatility as much as the average, so the analysis has to capture the range of outcomes, not just the central estimate. Loss modelling extends the burning-cost view by fitting frequency and severity distributions to the loss history and simulating many possible years, producing a distribution of the retained loss at each retention level: not just the expected retained loss but the range, including the bad-year outcomes the company has to be able to fund. This is what tells the company what a 1-in-10 or 1-in-20 bad year in the retained layer would cost it, which is the number that tests whether the retention is within the company's capacity.

The modelling has to handle the company's loss profile honestly. A company with frequent small losses has a predictable working layer that models well and can be retained with confidence; a company whose losses are infrequent but large has a volatile profile where a single loss dominates and the retention has to be set with more caution, because the small-numbers problem makes the history a poor guide. The modelling should reflect this, giving a tighter retained-loss estimate where the data supports it and a wider one where the losses are sparse and volatile, so the company does not over-rely on a thin history.

The output of the loss modelling and burning-cost analysis is the input to the optimisation: for each candidate retention, the expected retained loss and the range of outcomes, which the company sets against the premium saving at that retention to choose the level that minimises its total cost of risk within the volatility it can bear. This is the analytical core of the retention decision, and a company that makes the decision without it is choosing its retention blind.

The retention is not only a level but a form, and the type of deductible or retention shapes how the retained loss behaves and how it lands on the company's cash flow and balance sheet. Choosing the type is part of optimising the retention.

The deductible types

Several deductible structures are used, and they behave differently:

1. A flat deductible is a fixed amount per loss: the company bears the first fixed sum of each claim and the insurer pays the rest. It is simple and predictable per claim, but the total retained loss depends on the number of claims, so a year with many claims accumulates many deductibles.
2. A percentage deductible is a percentage of the loss or of the sum insured, common on catastrophe perils, so the retained amount scales with the loss, which means a large loss carries a large retained portion. The company's retained exposure on a big loss is therefore larger than under a flat deductible.
3. A franchise deductible is a threshold below which nothing is paid but above which the whole loss is paid: losses under the franchise are fully retained, losses over it are fully covered. It differs from an ordinary deductible in that, above the threshold, the deductible is not netted, so a franchise behaves quite differently around the threshold and the company retains all of the small losses entirely.
4. An aggregate deductible caps the total retained loss across the year: the company retains losses up to an annual aggregate, above which the insurer pays, which limits the company's total retained exposure even in a bad year and is a way to retain the working layer while protecting against an accumulation of retained losses. The aggregate is what makes a high per-loss retention tolerable, because it caps the downside of a bad frequency year.

The choice among these shapes the retained-loss profile. A company retaining a working layer often combines a per-loss retention with an aggregate cap, so it retains each working loss but its total retained loss is bounded, converting an open-ended frequency exposure into a capped one. Matching the deductible type to the company's loss profile and capacity is part of the optimisation.

Cash-flow and balance-sheet impact

Retained loss has to be funded, and how the company funds it is a cash-flow and balance-sheet question. The premium saved by raising the retention is a certain reduction in outgo; the retained loss is an uncertain future outgo funded as losses occur, so raising the retention shifts the company from a fixed premium cost to a variable retained-loss cost, cheaper on average but lumpier. The company has to be able to fund the retained losses when they fall, including a bad year, which means having the liquidity or the reserves to absorb them.

The company can fund the retention by paying losses from operating cash flow (simple but exposed to the loss volatility), holding a reserve for retained losses (smoothing the impact but tying up funds), or formalising the funding through a captive (the next section). The accounting and tax treatment depends on the structure, and the company should take the finance and tax view, because an informal reserve for future retained losses may not be deductible in the way a formal insurance premium is, which affects the after-tax economics of retaining versus transferring.

An optimised retention is not a permanent setting, because the price the market charges for the lower layers moves with the insurance cycle, and the retention that minimises the total cost of risk in a soft market is not the one that does so in a hard market. A company that sets its retention once and leaves it misses the saving that the cycle makes available, and carries a retention mismatched to the current price of transfer.

How the cycle shifts the optimal retention

The insurance market moves between soft phases, where capacity is plentiful and the price of cover is low, and hard phases, where capacity tightens and prices rise, and the lower layers of a programme are where the price moves most visibly. In a soft market the working layer is cheap to transfer, the insurer's loading on it is competed down, and the case for retaining it weakens, so a company can rationally carry a lower retention and let the insurer take more of the predictable loss at a price close to its expected cost. In a hard market the lower layers become expensive, the loading on them rises, and the case for retaining them strengthens, so the company should raise its retention to keep the predictable loss it can fund and pay the insurer only for the volatile excess that has become dear.

This is why the retention should rise into a hard market and can fall into a soft one, and why a company that reviews its retention at each renewal against the current price of the lower layers captures a saving that a static retention leaves on the table. The discipline is to test, at each renewal, what the insurer is charging to take the layer the company could retain, and to retain more when that charge is high and less when it is low, so the retention tracks the price of transfer rather than sitting fixed while the price moves around it.

Governing the retention decision

Because the retention is a financial decision with balance-sheet consequences, it should be governed rather than left to the broker to fill in on the renewal schedule. The company's finance function owns the capacity to fund the retained loss and the appetite for the volatility, and it should set the retention strategy: how much retained volatility the company is willing to carry, the funding the company holds against the retained loss, and the limits on how high any line's retention can go. The risk and insurance function then optimises within that strategy, line by line and renewal by renewal, bringing the burning-cost and modelling analysis and the current market pricing to the decision.

The retention decision should also be revisited after a material change in the company's circumstances, not only at renewal: a change in the company's financial capacity (which changes how much volatility it can bear), a significant change in its loss experience (which changes the modelled retained loss), or a major change in its operations or footprint (which changes the exposure). A retention that was optimal for the company a few years ago can be wrong for the company today, and the review keeps it matched to the company as it is, not as it was when the retention was last set.

An optimised retention is rarely a single number applied across the whole programme. The lines differ in their loss profile, their volatility, the price the market charges for the lower layers, and the company's capacity to retain them, so the retention should be set line by line, and then the retained layer has to be funded in a way that survives a bad year.

Setting retentions line by line

The optimal retention on a high-frequency, low-severity line (own-damage motor, or property attritional losses) differs from that on a low-frequency, high-severity line (liability, or catastrophe property), because the working layer the company can comfortably retain is large on the first and small on the second. The attachment point where the insurer's cover begins should therefore be set per line, against that line's own burning cost and modelled severity distribution, not as a uniform percentage of sum insured pasted across the slip.

The principles for setting retentions by line follow from the earlier analysis:

1. Retain more on predictable, high-frequency lines where the working layer is stable, the insurer's expense and risk loading on the lower layer is heaviest, and the company can fund the retained losses from the predictable run of claims.
2. Retain less on volatile, low-frequency lines where a single loss dominates, the thin history is a poor guide, and the retained loss in a bad year could be large relative to the company's capacity.
3. Set each line's retention against that line's burning cost and modelled severity, not a uniform figure, so the attachment point reflects the line's actual loss behaviour rather than a convenient round number.
4. Cap the programme with an aggregate or stop-loss across retained lines so the total retained loss is bounded even if several lines have a bad year together, which protects the balance sheet from a correlated retained-loss spike and keeps the worst case fundable.
5. Reset each line's retention at renewal against the current market price for the lower layers, raising it when the market hardens and the lower layers become expensive to transfer, and lowering it when the market softens and cheap transfer is available.

Funding the retained layer

A chosen retention is only real if the company can fund it when losses fall. The simplest funding is paying retained losses from operating cash flow, which is cheap but exposes the year's earnings to the loss volatility. A step up is a formal retained-loss reserve or self-insurance fund that smooths the impact across years, though it ties up capital and may not earn the tax treatment a premium does. A company retaining a substantial, stable book may eventually formalise the funding through an owned vehicle, but that is a separate decision driven by scale and permanence, and it does not change the retention arithmetic set out above: the vehicle organises the funding, it does not lower the expected retained loss. The point for retention optimisation is narrower. Choose the attachment point per line from the burning-cost and severity analysis, cap the aggregate so a bad year is survivable, and put a funding mechanism behind the retained layer that can carry the modelled bad year, not just the average. Get those three right and the retention lowers the company's total cost of risk (premium plus retained losses plus the cost of funding and volatility) at a level of retained variability the company has consciously chosen to bear.

Setting the retention well by line, then funding it deliberately, turns the retention from a default the broker fills in on the renewal schedule into a quantified risk-financing decision. The right per-line attachment point, the right aggregate cap, and a funding mechanism sized to the bad year are the whole of the optimisation.

Making these decisions well rests on understanding precisely how each policy's deductible, attachment and self-insured-retention provisions are worded and how the cover responds above them. Sarvada gives commercial insurance brokers and corporate risk teams structured, searchable access to insurer policy wordings, so the people optimising a retention can see exactly how each insurer frames its flat, percentage, franchise and aggregate deductibles, how the SIR layer is defined and who handles claims within it, and how the cover attaches above the retained layer across lines and across the market. Request Access to ground your retention and deductible optimisation in the real terms of the covers that sit above the retention.