Industry Health NPS Fell to 50 and Four in Ten Claimants Hit a Problem. Put Member Experience in the Insurer Scorecard

Hansa Research's 2026 health insurance study surveyed 4,000 respondents across more than twelve health brands and put industry Net Promoter Score at 50%, down from 55% the previous year. Brand-level results ran 59% for Tata AIG, 55% for ICICI Lombard and 54% for Bajaj Allianz, reported by MediaNews4U on 20 August 2026. The same study found that 40% of policyholders who filed a claim reported at least one issue, most commonly hospital network problems, delayed payouts and difficulty with pre-authorisation. Moneylife carried it on 21 August 2026 under the headline Health Insurance Customer Experience Worsens: 4 in 10 Claimants Report Issues.

Two things are worth separating. The NPS movement is a five-point drop on a sentiment measure, and sentiment measures wobble. The second finding is harder to wave away because it is not sentiment. It counts people who filed a claim and hit a problem, and the problems named have owners and timestamps: which hospital was in network on the day of admission, how long the pre-authorisation desk took, how many days passed between file completion and money leaving the account.

The study also found that 32% of respondents hold additional policies beyond employer-provided cover. For a benefits manager that is a quiet verdict on the group programme: a third of the covered population is buying a second policy with its own money, which is what people do when they do not fully trust the first one to work at the counter.

Open a typical group health RFP evaluation sheet and you will find claim settlement ratio, premium, solvency, sometimes a claims turnaround figure quoted at national average, and a soft box marked service. The weights sit overwhelmingly on price and settlement ratio. Insurance Business made the point on 13 August 2026 in a piece titled What India's claims settlement ratio doesn't tell brokers.

Settlement ratio travels well in tenders because it is one number, published, and comparable across insurers. It is close to useless for group health because it counts claims paid over claims received across an insurer's whole book, in every line and every geography. It says nothing about how long payment took, how many times the file bounced back for another document, whether the cashless request was authorised at midnight or at noon the next day, or how many claims were settled at a fraction of the amount claimed and therefore counted as settled while the member experienced a partial denial.

Meanwhile the escalations HR actually fields are the three things the Hansa study names. A hospital the portal listed as network and the desk treated as out of network. A pre-authorisation that sat while a family waited at the admission counter. A reimbursement filed in March and paid in June. None of those appear in the sheet that chose the insurer.

The fix is a scored module in the same evaluation sheet as premium and settlement ratio, built from five measures a TPA can produce from its own systems.

1. Pre-authorisation turnaround at the client's own top ten network hospitals. Not the insurer's national average, which blends a thousand hospitals and hides the ones that matter. Take the ten hospitals where this employer's members actually admitted over the last two years, and ask for median and 90th-percentile time from cashless request received to first decision at each, split by planned and emergency admission. A national median of two hours is irrelevant if the tertiary hospital nearest your largest office runs at eleven.
2. Denial-reason mix over the last twelve months. Every declined and partially declined claim, bucketed by reason: non-disclosure, waiting period, sub-limit or room-rent proportionate deduction, non-payable consumables, exclusion, documentation. The mix tells you whether denials come from policy design (fixable at renewal) or from administration (fixable only by changing behaviour). Our note on why health claims get repudiated works through those buckets.
3. Reimbursement ageing at the 50th and 90th percentile. Average days to settle is the most misleading service number in health insurance, because a fast median and a long tail average out to something respectable while the tail holds every escalation. Ask for the distribution, measured from date of complete file receipt and separately from date of first submission. The gap between those two clocks is where query loops hide.
4. Query-loop count per claim. The number of times a claim file went back to the member for more information before a decision. One query is normal. Three is a process telling the member it does not want to pay. TPAs report this least often and it tracks most closely with what claimants describe as a bad experience.
5. A small internal NPS run on your own claimants. Only members who closed a claim in the last quarter, surveyed within two weeks of settlement, using the standard zero to ten recommendation question and one free-text field. Eighty to a hundred responses are enough to see a trend. Benchmark it against the Hansa industry figure of 50 and the relevant brand figure.

All five metrics already exist inside the TPA's claims system. The obstacle is that nobody asks for them in a comparable format, so what arrives is a slide with three averages on it. Ask for the data as a claim-level extract, not a summary. One row per claim, with policy number masked, claim type (cashless or reimbursement), hospital name and city, date and time of intimation and of each authorisation decision, each query raised with its date, date of complete documentation, date of settlement or denial, amount claimed, amount settled, and denial reason code where applicable. From that one file you can compute all five metrics yourself and never argue about someone else's definitions.

For a competitive tender, ask each bidder for the same metrics on a comparable book, meaning accounts of similar headcount, sum insured band and geographic spread. Insurers will cite confidentiality, and the workable compromise is a metric-level aggregate certified by a named officer for a book they define in writing.

Ask also, in writing, for two things a claims extract does not contain:

• The network hospital list for the client's pin codes as at the tender date, plus hospitals added and removed from it over the preceding twelve months. Network churn near your offices is a leading indicator of the problems the Hansa study puts at the top of the complaint list.
• The current status of every hospital on the client's top ten list, meaning whether each is on cashless, on cashless with a package agreement, or suspended over a payment dispute. That second question prevents the most painful escalation an HR team receives.

Scoring works best when it is coarse enough to survive imperfect data and fine enough to separate bidders. Band each metric into five levels, score one to five, then weight.

A defensible starting weight set for a mid-size corporate programme puts member experience at roughly a quarter to a third of the total evaluation, alongside premium, policy terms and financial strength rather than replacing them. Within the module:

• Pre-authorisation turnaround at the client's own top ten hospitals: 30%
• Reimbursement ageing at the 50th and 90th percentile: 25%
• Query-loop count per claim: 20%
• Denial-reason mix and its administrative share: 15%
• Internal claimant NPS against the Hansa benchmark: 10%

The internal NPS carries the smallest weight deliberately. It is the noisiest of the five and works better as a check on the other four than as a driver: if the process metrics look good and the NPS is poor, something in communication and tone is failing that the timestamps do not capture.

Score the incumbent every quarter on the same five measures, not only at renewal. The discipline behind a quarterly insurer service scorecard applies here, and a firm already running a broker performance scorecard will recognise the mechanics. A number that appears only at renewal gets argued about at renewal. A number in every quarterly review gets managed.

Insurers report many metrics and commit to few. Confusing the two is how service schedules fill with promises nobody expects to keep.

Insurers and TPAs will generally accept hard commitments on what sits fully within their own control: time from receipt of a complete cashless request to first decision, from complete reimbursement file to payment instruction, to acknowledge an escalation from the corporate HR desk, and to issue an ID card or addition endorsement. These are clock metrics on their own process, they map onto turnaround obligations the insurer already carries under the policyholder protection framework, and a well-run TPA measures them already.

They will resist commitments on outcomes that depend on third parties or on claim merit. A denial-rate cap is the clearest example, and the buyer should resist it too, because an insurer that has agreed to a maximum denial percentage has a reason to settle weak claims and a reason to squeeze good ones once the quota is near. Hospital-side delay is another. A hospital that sends an incomplete pre-authorisation request at 2 am is not the insurer's fault, which is why the clock starts on receipt of a complete request and why the definition of complete has to be written down.

The workable middle sits in three places:

1. Committed on the client's top ten hospitals, reported everywhere else. An insurer that will not sign a national pre-authorisation turnaround will often sign one for a named list of ten hospitals, because the volume is knowable and it can put the relationship desk behind it.
2. Query loops capped by count, not by outcome. A commitment that no claim will be queried more than twice without a named officer signing off the third query is operationally cheap for the TPA and changes member experience immediately.
3. Committed at the median, reported at the percentile. Insurers will commit to a median settlement time long before a 90th percentile. Take the median as the commitment and the 90th percentile as a reported metric with a review trigger attached. Selecting the administrator itself is covered in our note on TPA selection for corporate health programmes.

A service commitment with no evidence obligation attached is a sentence in a document. Write the evidence requirement into the same schedule, with a due date and a format.

Ask for a monthly pack containing the claim-level extract for the month just closed, the five computed metrics with underlying counts shown, a line-item list of every claim that breached a committed turnaround with the reason, the current network status of the top ten hospitals including any suspension, and a note on any hospital added to or removed from the network within the client's pin codes.

Three details separate a pack that gets read from one that gets filed:

• Name the officer who signs it, with a designation and an email address, in the schedule. Packs signed by an inbox are prepared by nobody.
• Fix the definitions in an annexure. When the clock starts, what counts as a complete file, what counts as a query, how a partially settled claim is treated. Almost every argument about service metrics turns out to be an argument about definitions.
• Require the previous three months restated in every pack. This stops quiet redefinition: a metric that moved because the method changed becomes visible the moment last quarter's numbers change shape.

Twelve of these packs are also the most useful thing to hand a competing insurer at the next tender, because a bidder can price the account on real claims behaviour.

The standard drafting mistake is to attach the only consequence to termination. Nobody terminates a group health policy mid-year over pre-authorisation turnaround, because the disruption to two thousand employees is worse than the problem being solved, and both sides know it while they sign. A remedy nobody will use is not a remedy. Build a ladder instead, where each rung costs the insurer or TPA something real and no early rung requires the client to blow up its own programme.

1. Breach in one month. The pack flags it and the named officer gives a written root-cause note in the next month's pack. Cheap, but it puts the failure on a document the account team must sign.
2. Breach in two consecutive months or three in a rolling six. A joint review within fifteen working days, attended by the insurer's regional head and the TPA's operations head, not the servicing executive. The remedy is the seniority in the room and the written action plan with owners and dates that comes out of it.
3. Breach continuing through the review period. The client gains the right to require replacement of the named servicing team, to route cashless requests at the top ten hospitals through a dedicated desk at the insurer's cost, or to appoint a second TPA for reimbursement where the policy structure permits. Operational remedies bite: they change how the insurer staffs the account.
4. Sustained failure across two quarters. The renewal goes to market with the service data attached, and the schedule says so explicitly. That sentence does more work in a quarterly review than any penalty clause, because the account team is measured on retention.

Keep financial penalties modest, tied to administration fees rather than premium, and out of the first two rungs. A penalty large enough to matter to the insurer's economics invites a fight over definitions on every breach. A small one, paired with the operational remedies above, gets the behaviour without the fight.

A service level schedule works when the person who will be embarrassed by the number is in the room when it is read out. The rest of the drafting supports that one moment.

Write the review cadence into the placement slip and the broker service agreement together, so producing the pack and reviewing it sit with named parties on both sides from day one.