KPIs aren't a report.
They're a to-do list.

GCR mostly reflects how inflated the charge master is, so I don't manage to it. NCR is the number that tells the truth. The points sitting under 100 percent aren't random, so I break that gap into denials, underpayments, and write offs, give each one an owner and a target, and work it down. When NCR gets stuck around 89, there are almost always underpayments that nobody is flagging.

Every point under 95 is rework I'm paying for twice. I trace the rejections back to the few edit failures that cause most of them, and those shift by specialty. Eligibility shows up everywhere, modifiers and bundling drive it in surgery and multispecialty, medical necessity and LCD mismatches dominate in labs, and provider taxonomy trips up behavioral health. I tune the scrubber to the actual mix so the claim never goes out dirty in the first place. Fixing a claim at this stage costs a fraction of what it costs to fix the same claim later as a denial.

This is the metric that turns claim quality into a number leadership actually feels. I model an FPRR improvement as FTE savings. On a 10,000 claim book, every ten points is roughly one and a half analysts pulled off rework and put back on appeals that recover real money. That's how I make the case for scrubbing and automation. Not "cleaner claims," but two people back.

Days in AR is a lagging symptom, so I don't chase it in the aged buckets. By then the claim is already old. I work it upstream at submission lag, clean claim rate, and authorization, and AR comes down on its own. If AR is climbing while clean claim rate looks healthy, the problem isn't the claims. It's how we're following up with payers, and I retune the worklist.

The 90+ bucket is where revenue quietly turns into a write off as filing windows close. I run a worklist scored by days to deadline and dollar value, so the team always works the claims closest to expiring before the easy ones. Nothing collectible should ever age out by accident.

Patient AR behaves nothing like payer AR, so it gets its own approach. Collect at the point of service, set up FDCPA compliant payment plans, and offer financial counseling before anything escalates. As deductibles keep climbing, the practices that let patient aging hide inside the payer numbers watch their bad debt grow without noticing. I break it out so they can see it.

Velocity is really a payer behavior metric, so I never look at it blended. A payer paying in 34 days against a 14 day contract term is a follow up escalation and a contract conversation, and it goes straight onto the payer scorecard. A blended average is dangerous, because it hides the one payer that's choking the client's cash flow.

This is the slice of AR I control completely, with no payer in the way. Every day of charge lag is a day added to AR before the payer has even seen the claim. I keep charge lag under two or three days and posting close to same day, so both the AR clock and the cash picture stay honest. It isn't glamorous, and it's free money.

This is how I catch underpayments. I line up the actual average reimbursement per CPT against the contracted rate by payer, and the gap is silent money. A $15 shortfall on a high volume office visit, or a few cents on a high volume lab test, adds up to tens of thousands a quarter, and because it never shows up as a denial, nobody fights it unless someone is watching this number. I make sure someone is.

A high overturn rate sitting next to a low appeal rate is the worst combination there is. It means we win when we fight, but we only fight half the time. I triage every denial into front end, medical necessity, or payer behavior, appeal the medical necessity ones with the clinical documentation and payer policy to back them, and push the front end ones back upstream so they stop happening. The number I watch hardest isn't the denial rate. It's how many winnable denials we're walking away from.

Touches are labor, and labor is the cost line I can actually move. When the same claim gets picked up three or four times before it resolves, that's usually missing information, a denial that should have been prevented, or a follow up that left no clear next step. I go after the highest touch reason codes first, because pulling touches per claim from 2.5 down to 1.5 on a 6,000 claim book hands me back a couple of analysts without losing a dollar of collections. It's the cleanest way I know to add capacity without adding headcount.

I'm careful with this one, because productivity numbers get weaponized fast. A low figure rarely means the team is slow. More often they're reworking the same accounts, which is what K11 exposes, or fighting a payer that keeps denying. So I read productivity next to touches per claim and first pass rate, never on its own. When I do lift it, I do it by clearing the friction out of the work, not by pushing people to dial faster.

Cost to collect keeps me honest. I can always lift collections by throwing more bodies at the problem, but that isn't running things well. I bring this down by automating the predictable work, so the cost line drops while collections hold or grow. It's the metric that proves an improvement came from real efficiency and not just more effort.

Most teams measure how much of an underpayment they recover once they've found it, and stop there. The real leak is everything they never found, which is exactly why K09 feeds straight into this. Find the per CPT variance first, then go get it. You can't recover money you never knew was short.