By the time a formal grievance reaches a health plan’s Grievance and Appeals team, it arrives looking like an emergency. The member is angry. A response deadline is already running. A regulator may eventually read the file. What the file rarely shows is that the grievance did not begin that week, or even that month. It began on an ordinary call that did not get resolved and was never reviewed. The warning was there. No one was assigned to look for it.

Most health plans treat reducing member grievances as a downstream chore: staff the queue, meet the deadline, close the case. That posture quietly concedes the grievance as inevitable, and it is not. A grievance is the visible end of an escalation that usually runs about six weeks, and nearly all of it is recorded, in the plan’s own phone system, in the member’s own words. The signal is not missing. It is simply never listened to.

A grievance is a six week escalation, not a single event

CMS defines a grievance as a complaint about a plan’s delivery of service, and the rules let a member arrive there slowly. A Medicare Advantage enrollee has up to 60 days after the triggering event to file, and the plan then has 30 days to resolve a standard grievance, with a 14 day extension.⁵ That window is the outer edge of a story that almost always starts earlier, on the phone.

The arc is familiar to any member services team. It opens with a single call about a denied claim, a benefit change, or a stalled prior authorization, and the member hangs up with an answer that is incomplete or simply wrong. That is common. SQM Group puts first call resolution for health insurance at roughly 72 percent, so close to three in ten member calls are not settled the first time.¹ The member calls back, and calls again, and these repeat calls are the hardest to fix, because complaint calls resolve on the first contact only 47 percent of the time, the lowest rate of any call type SQM tracks.² By the fifth or sixth week the member gives up and files. Only then does the plan open a case.

The shape of it is mundane. A member is told on the phone that a drug is covered, learns at the pharmacy counter that it is not, gets a different answer from a second agent, and files after a third. Three recorded calls, one avoidable grievance, and a member now drifting toward disenrollment. Every call was captured. None was flagged.

Appeals run on the same current. A denial explained badly prompts an appeal, not just frustration. The recording of that call is the clearest account of what the member was told, and the one document the appeals file rarely contains. A single unreviewed call can feed both outcomes, a grievance about the service and an appeal against the decision, both audible weeks before either was filed.

Why the warning stays invisible

The reason is not indifference. It is arithmetic. Traditional call center quality assurance, whether member service is run internally, outsourced, or split between the two, reviews somewhere between 2 and 5 percent of calls.³ The other 95 percent, which includes nearly every repeat call in an escalating grievance, is never heard by anyone whose job is to catch problems. A sample that small will almost never contain the three or four particular calls that make up one member’s slow walk to a filing, and it is even less likely to reveal the shape of the trouble when the same benefit is being miscommunicated to hundreds of members at once.

Sample size is only half of the failure. The deeper flaw is what the sampling was built to measure. A conventional scorecard asks whether the agent greeted the member, verified identity, and read the required disclosures. It does not ask whether the member’s problem was actually solved, or whether the member hung up angrier than they picked up. A call can earn a clean score and still be a grievance in motion. The program was designed to grade etiquette, and etiquette is not the thing that turns into a filing. We have made this case before, that most grievances start as calls that never got reviewed, and nothing in the underlying mechanics has changed since.

The calls in languages other than English are the least visible of all. When a member with limited English proficiency cannot be helped without an interpreter, the quality of the resolution is harder to verify, the member is less likely to push back on an answer they do not fully understand, and the interaction is almost never scored at all. Those are precisely the calls where a small misunderstanding hardens, unseen, into a grievance.

CMS just removed your grievance scorecard

On April 2, 2026, CMS issued the Contract Year 2027 Medicare Advantage and Part D Final Rule and, with it, removed 11 measures from the Star Ratings. Four of them speak directly to this problem, and all four leave the formula beginning with the 2029 Star Ratings: Complaints about the Health and Drug Plan, Members Choosing to Leave the Plan, Plan Makes Timely Decisions about Appeals, and Reviewing Appeals Decisions.⁴ Read in a hurry, that looks like a reprieve. The measures that once turned complaints, disenrollment, and appeals handling into a Star score are going away, and the natural temptation is to slide grievance and appeals monitoring down the list of things worth watching.

That reading has it exactly backward. The measures are leaving the scorecard. The exposure is staying exactly where it was. CMS still runs the Complaints Tracking Module, and plans are still bound to the resolution timelines set at 42 CFR 422.125 and 422.564.⁵ The complaint and customer service questions remain on the CAHPS survey, and the survey based measures carry heavy and rising weight in the Medicare Advantage formula.⁶ A member who leaves still leaves. What actually changed is narrower and more dangerous than relief: the warning light that used to sit on the Star dashboard has gone dark. Plans that had quietly relied on those measures as their grievance scorecard now have no scorecard, and precisely the same risk underneath it.

What it takes to see it coming

Seeing a grievance coming means catching the escalation while it is still a service problem, not a case number. It rests on a reversal: stop sampling calls to grade agents, and start reading all of them to find members. A plan that analyzes every interaction can connect the several calls behind one escalation and watch the same complaint surface across the population, where a single systemic fix replaces hundreds of grievances not yet filed.

The point is not more scores but earlier ones. A member who has called three times about one decision is a retention risk however politely each call was handled, and no rubric that grades greetings will say so. A few practical tests separate a system that sees a grievance coming from one that only counts calls afterward:

• Whole population analysis, not sampling, so one member’s escalating calls and the patterns forming across the book are both visible before anyone files.

• Scoring that reads the member, not just the agent, weighing sentiment, unresolved issues, and repeat contact rather than whether the greeting was correct.

• The same rigor for calls in every language, including interpreter lines, because that is where hidden escalation collects and Section 1557 makes language access a legal duty.

• Plan ownership of the conversation data, scoring models, and trend lines, so the early warning system survives changes in telephony and staffing and can be retuned in days.

• Alerts that reach the people who can act, so member services, grievance and appeals, and compliance see a trend while there is still time to intervene.

Two ways to run grievance and appeals operations

‍

Mizzeto built Claro for exactly this gap. It analyzes the full volume of member calls, including the ones in other languages and on interpreter lines that sampled QA never reaches, and its Member Sentiment & At-Risk Identification scoring is built to surface the unresolved, escalating conversations that become grievances and appeals weeks later, all while the plan keeps ownership of the underlying data and the logic that scores it. If a grievance and appeals team is meeting every deadline and still watching volume climb, the explanation is usually sitting in calls the plan has already recorded and has never had a way to hear.

The bottom line

A formal grievance is the most expensive way a health plan can learn about a problem it could have seen six weeks earlier. The information was never missing. It was sitting, unheard, in the 95 percent of calls no one reviews. CMS has taken away the measures that once forced plans to watch their complaints and their appeals, but the price of a grievance, paid in CAHPS, in disenrollment, and in compliance exposure, has not fallen by a cent. Reducing member grievances begins with a decision to stop waiting for them to arrive.

Any plan can find out what its own calls are already saying. Mizzeto will score a sample of them and show a plan the warnings hidden inside before it commits to anything further. For the upstream work that keeps these calls from going wrong in the first place, see how payers can fix their call centers.

References

1.  SQM Group, “First Call Resolution Benchmarking by Industry” (health insurance first call resolution approx. 72 percent). sqmgroup.com

2.  SQM Group, first call resolution by call type (complaint calls resolve at 47 percent, the lowest of all call types). sqmgroup.com

3.  SQM Group, call center quality assurance benchmarking (traditional programs review roughly 2 to 5 percent of interactions). sqmgroup.com

4.  Centers for Medicare & Medicaid Services, “Contract Year 2027 Medicare Advantage and Part D Final Rule,” April 2, 2026 (removal of 11 Star Ratings measures, including Complaints about the Health and Drug Plan, Members Choosing to Leave the Plan, Plan Makes Timely Decisions about Appeals, and Reviewing Appeals Decisions, effective with the 2029 Star Ratings). cms.gov; Federal Register, April 6, 2026.

5.  42 CFR 422.564 (grievance procedures: 60 day filing window, 30 day standard resolution, 14 day extension) and 42 CFR 422.125 (resolution of complaints in the Complaints Tracking Module). ecfr.gov

6.  Analysis of the CY2027 Final Rule finding relative Star Ratings weight shifting toward survey based measures such as CAHPS for Medicare Advantage plans (e.g., Crowell & Moring; Holland & Knight client alerts, April 2026).

‍

Most health systems measure patient experience after it is over. The survey arrives weeks after the visit, asks about the physician, the nurse, and the discharge instructions, and lands as a number on a dashboard a full quarter after the care happened. By the time anyone reads it, the experience is already history.

For a large share of patients, though, the experience did not begin in the exam room. It began on the phone. It began when they called to book an appointment, asked a question about a bill, tried to reach someone in a language other than English, or waited on hold to learn whether a procedure was covered. That first call sets the tone for everything that follows, and most organizations have almost no visibility into how it went.

The call center is the front door to the health system, and also the least measured room in it. Most organizations review fewer than 5 percent of their calls, which means the interactions that decide whether a patient stays or leaves are usually the ones nobody hears. You cannot improve an experience you only sample. This article looks at why patient experience starts in the call center, what it costs when those calls go unreviewed, and what to look for in a way of measuring them that does not depend on a survey arriving months too late.

The first impression is usually a phone call

The first call is the first test of the relationship, and patients grade it against every other service they use. In a recent Harris Poll, 61 percent of Americans said they want their healthcare experience to feel more like a convenience app such as Amazon Prime or Uber, and 60 percent said they find the process of seeing a new provider frustrating.¹ For most of them, that judgment forms on the phone.

That call carries more weight than it appears to. Accenture, which has surveyed more than 21,000 consumers on healthcare experience, found that 30 percent of patients selected a new provider in 2021, and that nearly 80 percent of those who switched cited poor navigation factors as the reason, including difficulty doing business and bad experiences with administrative staff.² Navigation and administrative staff are not abstractions. They are the scheduling line, the billing line, and the front desk phone.

What this looks like in practice is familiar. A patient calls to schedule, gets transferred twice, sits on hold, never gets a clear answer about cost, and quietly books somewhere else. No survey ever captures that call, because the caller never became a patient. They simply did not come back.

The survey is a rear-view mirror

Patient experience surveys are valuable, but they share a structural limitation. HCAHPS for hospitals and CG-CAHPS for medical groups measure experience after the fact, on a sample of patients, and report it weeks or months later. They tell you what happened. They cannot tell you what is happening right now, while you can still do something about it.

A survey score also cannot explain itself. It can tell you a patient rated you a six. It cannot tell you that the patient was transferred three times trying to reschedule, that your Spanish-speaking callers are routinely routed to English-only agents, or that a billing conversation went sideways and broke trust. The reason behind the score lives in the call, not in the survey.

And the score is not a vanity metric. Through the CMS Hospital Value-Based Purchasing program, 2 percent of a participating hospital’s Medicare payments are withheld and redistributed based on performance, and the patient experience domain measured by HCAHPS accounts for 25 percent of the Total Performance Score.³ For a mid-sized hospital, that is millions of dollars tied to patient experience, a meaningful share of which is shaped before the patient ever arrives for care. That direct exposure is the hospital case. Physician groups do not sit under Hospital Value-Based Purchasing, but they answer to their own version through CG-CAHPS and value-based programs such as MIPS. The mechanism differs by setting. The underlying point does not: the experience that drives the score is built on calls, and it is rarely captured by them.

The calls that matter are the ones you never hear

Here is the part that does not get said often enough. Traditional manual call quality assurance reviews a small fraction of total calls, typically less than 5 percent. This is not a failure of the people doing the work. A human QA team, however skilled, can only listen to so many calls in a day. It is a limitation of the measurement model, and it holds true whether your contact center is in-house, outsourced, or a combination of both.

The problem is what that small sample misses. The calls that drive complaints, switching, and low scores are outliers by definition, and a random sample of a few percent is structurally poor at finding outliers. The frustrated caller, the dropped handoff, the limited English patient who never got a qualified interpreter: these are precisely the calls that fall outside the sample. The 95 percent nobody reviews is exactly where the risk lives.

When you can only see a sliver, you end up managing to averages. Average handle time, abandonment rate, and service level tell you the center is busy. They do not tell you whether patients felt heard, whether a financial conversation damaged trust, or whether a caller is one bad interaction away from leaving. The warning signs of a patient about to switch are audible long before they ever surface in a survey: repeat calls about the same unresolved issue, rising frustration in a caller’s tone, questions that never get a straight answer. Caught in the moment, those are coachable and fixable. Caught in a survey a quarter later, they are already lost revenue and a lower score.

What to look for in a way to measure patient calls

The goal is not to survey harder. It is to actually hear what happens on your calls, all of them, in a way you can act on. When evaluating how to do that, whether your contact center is run in-house, through a partner, or as a hybrid, look for a few things.

• Full coverage, not a sample. The system should review 100 percent of calls, not the small share a manual team can reach. Outliers only become visible at full coverage.

• Experience, not just compliance. It should score patient sentiment and the quality of agent communication, not only whether a script was followed. Compliance checklists miss the human signal that actually moves HCAHPS.

• Every language, automatically. Non-English calls should be reviewed the same way English calls are. Section 1557 of the Affordable Care Act requires meaningful access for patients with limited English proficiency,⁴ so a tool that cannot read those calls leaves both an experience gap and a compliance gap.

• Real time, not retrospective. It should surface issues while you can still intervene, not weeks later. A survey is a rear-view mirror. Live call intelligence is a windshield.

• Owned by your organization. You should own the data and the real-time intelligence so your teams can act on it directly and continuously, rather than at quarterly reporting intervals.

• Built to fit your stack. It should connect to the systems you already run, without a disruptive replacement project.

The difference between the two models is the difference between knowing your score and knowing why you earned it.

This is the gap Claro by Mizzeto was built to close. Claro is an AI-powered contact center intelligence platform that audits 100 percent of patient and provider calls, including calls in languages other than English, scoring patient sentiment, agent communication, and compliance in real time rather than on a sample weeks after the fact, and it connects to the systems you already use. The result is the ability to see and coach to what is actually happening on every call, in every language, instead of inferring it from a fraction of them. You can read more about how this reshapes day-to-day operations in our overview of modern call center operations.

The bottom line

The visit is not where patient experience begins. For a growing share of patients, it begins on the phone, and it often ends there too, before anyone in a white coat is involved. Health systems have spent years measuring experience after the fact and running the contact center to averages, while the moments that decide loyalty, reputation, and a meaningful slice of value-based reimbursement play out on calls nobody hears.

The organizations that move first to hear every call, in every language, as it happens will hold a real advantage: stronger HCAHPS performance, better patient retention, and operational decisions grounded in what patients actually experienced rather than what a delayed survey implied. Patient experience starts on the phone. The only question is whether you can hear it.

References

1. Medical Economics, “69% of patients would switch providers for better services” (The Harris Poll), 2026. medicaleconomics.com

2. American Hospital Association, “Why Patients Leave: 4 Nonnegotiable Consumer Expectations” (Accenture consumer research), 2023. aha.org

3. Centers for Medicare and Medicaid Services, “Hospital Value-Based Purchasing.” cms.gov

4. HCAHPS and Hospital VBP, Patient Experience of Care domain weighting. hcahpsonline.org

5. U.S. Department of Health and Human Services, Office for Civil Rights, Section 1557 non discrimination final rule, 2024. hhs.gov

‍

How call intelligence catches member complaints before they reach your grievance and appeals team

Every health plan has a grievance and appeals operation. Staff, timelines, case management workflows, regulatory tracking. For Medicare Advantage plans, CMS audits it. For Medicaid plans, states audit it. The compliance burden is real. The operational cost is real.

What most plans have not built is anything upstream of it.

A formal grievance is not where the problem starts. It is where the problem ends up after several earlier opportunities to fix it were missed. A member's prior authorization status was communicated incorrectly. A benefit change was explained poorly and the member hung up confused. An agent gave a formulary answer that turned out to be wrong. None of those calls showed up in the QA report, because the QA program reviewed 2 to 5 percent of call volume^[1] and these particular calls were not in the sample. Weeks later, the member called back. Then again. Then filed.

This is the upstream problem that most health plan grievance operations are not resourced to see. The tools needed to catch it simply have not been part of how call center quality has historically worked.

What grievances are actually made of

CMS defines a grievance as a complaint about a plan's delivery of service. That definition covers a wide range of situations: an agent who was dismissive, a hold time that felt unreasonable, a coverage question that went unanswered. The operational category that generates the most preventable grievances is simpler than the regulatory definition suggests.

Most formal grievances start as an unresolved phone call.

SQM Group's benchmarking data puts the healthcare insurance call center first call resolution rate at approximately 72 percent^[2], meaning roughly 28 percent of member calls do not get resolved on first contact. Some of those members call back once. Some call back twice. When the calls involve a coverage dispute, a prior authorization denial, or a benefit question with real financial consequences for the member, the escalation path eventually leads to a formal grievance.

The more telling finding is about complaint calls specifically. The FCR rate for interactions where a member is already expressing dissatisfaction is only 47 percent.^[3] Less than half of the calls most likely to become a grievance get resolved on the first contact. Plans are sending a continuous stream of unresolved member issues toward the back end operations that cost the most to run.

There is a second finding worth flagging here. SQM's research estimates that approximately 14 percent of callers describe their call as a complaint call. Most contact centers believe that number is under 5 percent.^[4] Complaint volume is substantially underreported. Plans are not just missing the resolution. They are missing the signal that a problem exists at all.

The enforcement environment has changed

In January 2026, California's Department of Managed Health Care fined Anthem Blue Cross $15 million for what it described as longstanding and widespread deficiencies in handling member grievances spanning more than 15 years.^[5] The action included a requirement for an independent auditor to oversee grievance system corrections for up to four years. This was not the first enforcement action for the same pattern. Prior survey findings had not produced sustained correction.

The pattern regulators are reacting to is not primarily about whether a plan has a compliant grievance process on paper. It is about whether grievances are actually resolved. A well documented compliance framework that leaves a significant share of complaints unresolved is still a regulatory liability.

CMS also removed the Complaints about the Health Plan and Complaints about the Drug Plan measures from the Star Ratings formula, effective with the 2029 Star Ratings.^[6] The instinctive response to that change is to treat complaint monitoring as less important. That is the wrong conclusion. CMS retains these as display measures, complaint trends are widely regarded in payer operations as leading indicators of CAHPS deterioration, and enforcement authority over grievance handling is completely independent of Star Ratings. The scorecard accountability signal is gone. The underlying compliance exposure is not.

What the call reveals that the grievance does not

By the time a formal grievance reaches the G&A team, the plan already has a documentation obligation, a response deadline, and a case that may be reviewed by a regulator. What the plan usually does not have is insight into the original interaction.

Not because the call was not recorded. Most plans record calls. But because the call was not analyzed. Nobody reviewed what the agent communicated about the prior authorization decision, whether the member left with an accurate understanding of their coverage, or whether the issue raised on that call had appeared on 200 other interactions in the previous 30 days.

That missing analysis is the upstream gap. A grievance intake form tells the plan what the member is upset about now. The original call tells the plan what actually happened, who was responsible, whether it was an agent accuracy issue or a systemic script failure, and whether the same pattern is playing out across hundreds of calls currently in queue.

When plans close that gap, the economics look different. Processing a formal grievance costs real money in staff time, documentation, and in some cases regulatory engagement. Catching the underlying issue in the call costs a fraction of that. The intervention point is upstream, and it is the cheaper one.

The call types most likely to generate a formal grievance follow a recognizable pattern. Each one shares the same underlying failure: an interaction that ended without resolution, with no mechanism to catch it before the member decided to escalate.

The language access dimension

Non-English member calls carry a disproportionate share of the upstream grievance risk. When a limited English proficient member reaches an agent who cannot serve them without an interpreter, resolution quality is harder to verify, the member is less likely to push back on an unclear answer, and the interaction is almost never scored in a traditional QA program.

The CY2027 Final Rule removed the Call Center Foreign Language Interpreter and TTY Availability measure from Star Ratings, effective with the 2029 Star Ratings.^[7] Plans may read that as reduced pressure on language access. The old measure rewarded having an interpreter line available. The bar has shifted. LEP member dissatisfaction with service quality now surfaces in overall CAHPS scores, and those scores carry direct financial consequences for Medicare Advantage plans. It shows up 12 to 18 months after the interaction.

Plans that do not score non-English calls with the same rigor as English calls have a quality gap that will surface in member satisfaction surveys and, for Medicare Advantage plans, directly in CAHPS scores. It will just arrive on a delay.

What to look for in a solution

If you are evaluating call intelligence for this purpose, these are the capabilities that actually matter:

100 percent call coverage, not sampling. The interactions that generate formal grievances are rarely in a 2 to 5 percent random sample. Full coverage is what makes upstream intervention possible.

Root cause analytics, not just QA scores. An agent compliance score tells you whether the script was followed. It does not tell you why members are calling back about the same issue or whether a benefit change was communicated incorrectly across an entire team. Systemic failure identification is what separates a grievance prevention tool from a QA scorecard.

Linkage between call data and G&A intake. If a formal grievance arrives and the underlying call is retrievable, analyzable, and traceable to a root cause, that is a materially different case than one built from member description alone. The connection between call intelligence and grievance case management is where the prevention value is.

Native language quality monitoring. Non-English calls should be scored in the language they were conducted, not translated after the fact and then graded. Post hoc translation loses the nuances that determine whether a member actually understood the answer they received.

Plan ownership of the underlying data. If call recordings, transcripts, and scoring models belong to a vendor rather than the plan, the plan cannot connect call intelligence to its own member satisfaction data, grievance analytics, and retention tracking. That intelligence needs to stay with the plan.

How Claro by Mizzeto approaches this

Claro by Mizzeto was purpose built for health plans and analyzes 100 percent of member interactions across languages, scoring every call against six dimensions: CMS guidelines compliance, HIPAA compliance, resolution rate, CMS accuracy, member sentiment, and communication and empathy. Automated post call translation of non-English interactions is the enabling capability that makes scoring in any language possible. Call data stays with the plan. For related context, see Are Health Plans Really Listening to All Their Members? and our guide to how payers can fix their call centers.

The bottom line

Grievance volume is a lagging indicator. By the time a formal complaint reaches the G&A team, the call that caused it has already happened, the member's patience has already worn out, and the cheap intervention window has already closed. The plans that reduce preventable grievances are not the ones that build bigger G&A teams. They are the ones that get visibility into every call before the member decides to escalate.

The data is already in the call recordings. The question is whether the plan is actually looking at it.

References

1. SQM Group. Published guidance on manual QA sampling rates, noting that most contact centers review 2 to 5 percent of call volume. https://www.sqmgroup.com/software  

2. SQM Group. Call Center FCR Benchmark Results by Industry. Healthcare insurance industry first call resolution rate cited at approximately 72 percent. https://www.sqmgroup.com/resources/library/blog/call-center-fcr-benchmark-2024-results-by-industry  

3. SQM Group. Top 10 Reasons for Repeat Call Complaints. FCR rate for complaint calls cited at 47 percent. https://www.sqmgroup.com/resources/library/blog/customer-complaint-calls  

4. SQM Group. Top 10 Reasons for Repeat Call Complaints. Discrepancy between actual complaint call volume (approximately 14 percent) and the typical contact center estimate (under 5 percent). https://www.sqmgroup.com/resources/library/blog/customer-complaint-calls  

5. California Department of Managed Health Care. Press release: DMHC fines Anthem Blue Cross $15 million for longstanding and widespread failures with member complaints. January 30, 2026. https://www.dmhc.ca.gov/Resources/Newsroom/PressReleases/January30,2026.aspx  

6. Centers for Medicare and Medicaid Services. Contract Year 2027 Medicare Advantage and Part D Final Rule, April 2026. Complaints about the Health Plan and Complaints about the Drug Plan measures removed from Star Ratings, effective with the 2029 Star Ratings.

7. Centers for Medicare and Medicaid Services. Contract Year 2027 Medicare Advantage and Part D Final Rule, April 2026. Call Center Foreign Language Interpreter and TTY Availability measure removed from Star Ratings, effective with the 2029 Star Ratings.

‍

The honest answer for most health plans is no. Most audit only 2 to 5 percent of their member service call volume,¹ whether the calls are handled in-house, outsourced, or run on a hybrid model. The other 95 percent, including the calls that drive complaint patterns, CAHPS results for Medicare Advantage plans, NCQA accreditation outcomes, and grievance filings, is operationally invisible. This is not because plans are negligent. It is because manual quality assurance, at any scale a health plan would consider economically viable, was never going to review more than a fraction of total volume.

The model was defensible when human review of every call was infeasible. A quality assurance team, almost always internal to the plan, listens to a random sample of recordings, grades each against a rubric covering greeting, identity verification, hold etiquette, closing, and compliance disclosures, and rolls the scores up into a monthly quality report. That constraint has now changed, and the math of what the sampled program does not see has gotten more expensive.

Why a 2 percent sample cannot tell you what you need to know

Three things break the sampling model in the real operating environment of a health plan. First, the sample is too small to surface the outliers that drive dissatisfaction: the prior authorization status miscommunication, the Spanish-speaking caller transferred three times, the appeals question a non-clinical agent improvised on. At 2 percent of a million calls a year, you are reviewing 20,000 calls, and the handful that mattered are statistically invisible.

Second, the scoring rubric measures script compliance, not member outcomes. An agent can hit every checkbox on a quality assurance form and leave the member with the wrong answer. Conversely, an agent can solve a complex eligibility question with empathy and skip half the script. The first call scores well, the second scores poorly, and member experience is the inverse.

Third, and most consequentially, the quality assurance score is structurally disconnected from the outcomes the plan actually cares about. A plan can run a compliance program for three years and still watch scores drift downward, because the program is measuring what is easy to grade, not what determines plan performance.

The financial stakes vary by line of business but point in the same direction. For Medicare Advantage plans, CAHPS measures were quadruple-weighted in the 2023 through 2025 Star Ratings and remain double-weighted from the 2026 Star Ratings forward,² and QBP eligibility is binary at the four-star threshold. A half-star drop from 4.0 to 3.5 eliminates the entire 5 percent Quality Bonus Payment on the benchmark, several million dollars in annual revenue exposure on a mid-sized MA plan. For Medicaid managed care plans, the same call patterns drive access audit findings and capitation negotiations. For commercial and ACA plans, they drive NCQA accreditation outcomes and employer client retention.

What changes when you measure every call

Multilingual speech-to-text, conversation analysis, and intent classification now run reliably at full call volume across the languages CMS requires plans to support for LEP populations. When Member Experience covers 100 percent of member calls rather than a sample:

• You see the patterns that drive member experience scores before the next survey cycle comes back. Repeated transfers on prior authorization inquiries. Unresolved questions about formulary changes. Disconnects on appeals calls.

• You can connect call content to outcomes. A member who called three times in a month about the same issue is not a satisfied member, regardless of how each individual call scored. The pattern is only visible across the full population.

• You can detect compliance risk while it is still recoverable. A miscommunication about coverage that surfaces in week one of a transition can be corrected before it becomes a grievance, an appeal, or a CMS audit finding.

• You can act on multilingual experience as a measured outcome rather than a checkbox. Surveys rarely capture whether a Spanish-speaking member actually understood the answer they got. Conversation-level analysis can.

The relationship between surveys and conversation analysis is not a replacement. It is a stack. Member experience surveys tell the plan what the member experienced and how they felt about it. Call content tells the plan why. Surveys are the outcome layer that determines QBP eligibility for MA plans and accreditation status across other lines. Conversations are the operational layer that surfaces the specific behaviors, scripts, transfers, and miscommunications driving the outcome.

Six tests for a defensible call center Member Experience program

Member Services leaders evaluating how to upgrade their Member Experience program in 2026 can apply six tests. Each isolates a structural property that determines whether the program will actually measure member experience or merely produce a number. Most arrangements on the market today, including in-house teams running on legacy infrastructure and vendor-managed programs alike, will fail at least three of these tests.

CMS is already doing this. The question is whether you are.

Every year, CMS places test calls into the prospective beneficiary and current enrollee call centers of every Medicare Advantage and Part D plan in the country. The Accuracy and Accessibility Study measures whether plan representatives provide correct benefit information, whether qualified foreign-language interpreters are reachable for LEP callers, and whether TTY services function for deaf and hard-of-hearing members. Results feed directly into the Customer Service domain of the Star Ratings calculation, and low performance can trigger compliance action.³

CMS is, in other words, already running an independent Member Experience program on every MA plan's member-facing call center, because plan-administered quality scores alone have not been considered sufficient as a regulatory measurement substitute. But the agency's sample is tiny relative to total volume, and the protocol only measures what its test scenarios cover. Real-world failure modes that fall outside the protocol, including the prior authorization status miscommunication, the formulary question handled by an undertrained agent, and the appeals call that ends in a disconnect, are not in the data CMS sees.

What Mizzeto's Multilingual Member Experience Solution does

Mizzeto's Multilingual Member Experience Solution is built for health plans, not retrofitted from a generic CX platform, and is designed to pass all six tests above by default. Concretely:

• 100 percent analysis of member interactions, with AI-driven transcription and scoring that runs at full call volume rather than a sample.

• Native multilingual analysis across English, Spanish, Mandarin, Vietnamese, Tagalog, and the other languages CMS requires plans to support for LEP populations.

• Payer-configurable scoring rubrics the plan can update in days, with Member Experience scores linked directly to CAHPS measures, grievance volumes, appeal overturns, and disenrollment patterns.

• Plan ownership of the underlying conversation data, scoring models, and historical trend lines. Whether call handling, telephony, or transcription providers change, the member experience monitoring infrastructure stays with the plan.

• Real-time alerting and direct conversation-level access for Member Services and Compliance teams, so when a grievance trend, or audit finding requires explanation, the evidence is already in the plan's possession.

The bottom line

Member Experience measurement for health plans is at an inflection point. The sampling model that defined the last twenty years cannot survive the combination of heavily weighted member experience measures in the MA Star Ratings formula, parallel quality reporting requirements for Medicaid and ACA plans, and the simple fact that the technology to measure every call exists and is being deployed by competitors. If CMS itself has concluded plan-administered scores need an independent measurement layer, the rational response is to install one at scale, on the 95 percent of calls CMS does not sample, before CMS surfaces a finding the plan has to explain. Apply the six tests above to your current arrangement, and see how Mizzeto's Multilingual Member Experience Solution measures against them.

References

• 1. SQM Group. mySQM Auto QA published guidance on manual QA sampling rates. www.sqmgroup.com/software

• 2. Centers for Medicare and Medicaid Services. 2026 Medicare Part C and D Star Ratings Technical Notes. www.cms.gov/files/document/2026-star-ratings-technical-notes.pdf. CMS Medicare Advantage and Part D Final Rule reducing CAHPS and administrative measure weights from 4x to 2x effective 2026 Star Ratings.

• 3. Centers for Medicare and Medicaid Services. Part C and Part D Call Center Monitoring, Timeliness and Accuracy and Accessibility Studies. Annual guidance memoranda and HPMS performance reports. www.cms.gov

‍