The setup
An employee drops spousal coverage effective July 1. The change is keyed in the ben-admin system on June 27, which is on time by any reasonable standard. The EDI file goes out that night.
The carrier generated its July bill on June 24. That bill shows the employee at EE+SP:
| July bill | Should bill (EE) | Carrier billed (EE+SP) | Variance |
|---|---|---|---|
| July | $320.09 | $687.59 | $367.50 |
A line-by-line reconciliation flags it immediately. Census says EE, bill says EE+SP, the rate difference is $367.50. Every number on that line is correct, and calling it an error today is still wrong.
Why it isn't an error yet
The change reached the carrier three days after the July bill was produced. No carrier process could have put it on that bill. Carrier billing systems handle this the same way: next cycle, bill the current month at the new tier and generate a retroactive adjustment for the month that printed before the change arrived.
So the July line is a timing exception: a known, expected variance with a specific bill where it should resolve. It turns into a discrepancy only if that bill doesn't resolve it.
Why disputing too early backfires
Send the July variance to the carrier and the answer is predictable: the change was received after the cycle cutoff and will adjust on the next bill. The carrier is right. Now your team has spent a dispute and some credibility on something that was never wrong.
Do that a few times and two things happen. The carrier's service team learns that your disputes are often premature. Your own analysts learn that a variance on a recent change usually sorts itself out. Both lessons are mostly true, which is what makes them dangerous. They are exactly how the cases that don't sort themselves out get through.
August: four possible outcomes
When the August bill arrives, the timing exception resolves one of four ways:
| Outcome | August current month | July credit on August | Still owed to you | Status |
|---|---|---|---|---|
| A — Clean | EE, $320.09 | −$367.50 | $0.00 | Closed |
| B — Credit missing | EE, $320.09 | none | $367.50 | Discrepancy |
| C — Partial credit | EE, $320.09 | −$183.75 | $183.75 | Discrepancy |
| D — Never applied | EE+SP, $687.59 | none | $735.00 | Discrepancy, aged from July |
Outcome D is the easy one to catch: the current month is still wrong, and any check of the August bill finds it. Outcome C, here exactly half the credit, usually means the carrier applied the change from the wrong date or prorated a change that should have been a full month.
Outcome B is the one to worry about. The August bill is correct: right member, right tier, right rate. Every check of the current bill passes. The only evidence of the problem is something that isn't on the bill, a July credit that never arrived. A reconciliation that only checks the current month against the census can't see it, and one that already wrote July off as "probably timing" has stopped looking.
The three inputs that tell them apart
Classifying a variance correctly takes three pieces of information that rarely sit in the same place:
- The last-change date from the ben-admin system: when the change was keyed, not just its effective date. Without it, a timing exception and a carrier error look identical on the bill.
- The carrier's cycle cutoff for each bill. Many invoices print a run or statement date. Where they don't, the cutoff can be inferred from which changes did and didn't make it onto the bill, and once established it's stable per carrier.
- The next bill, checked for both halves of the expected resolution: the corrected current month and the retro credit for every month that printed early.
| Change keyed | This bill | Classification | Action |
|---|---|---|---|
| Before cutoff | Wrong | Discrepancy | Dispute now |
| After cutoff | Wrong | Timing exception | Predict the resolution; check next bill |
| After cutoff | Next bill fully resolved | Closed | None |
| After cutoff | Next bill partly or not resolved | Discrepancy | Dispute the remaining amount, aged from the first affected month |
The retro clock doesn't wait for the classification
A timing exception still uses up retro window. Carrier retro caps run from the coverage effective date, not from the date someone noticed the variance.
In the example above, the change was keyed four days before it took effect, so the window is barely touched by the time August arrives. Now change one fact: the same EE+SP → EE change was effective May 1 and keyed on June 27. It's still a timing exception on the July bill, because it arrived after cutoff. But by the time it can be checked on August's bill, May is already about 90 days back. On a 90-day contract, part of the recovery is at risk before anyone has done anything wrong. (The full mechanics are in the retro window guide.)
That's why late keying and cycle cutoffs need to be tracked together. Each one on its own looks harmless. Combined, they can push a correctable error out of the window.
What to do with this
- Record the cycle cutoff for every carrier. It's the single most useful piece of reference data for separating noise from findings.
- Carry the last-change date into the reconciliation, not just the effective date.
- Treat timing exceptions as open items with an expected resolution: which bill, which amount, which months. Close them only when that resolution appears in full.
- When a timing exception misses, age it from the first affected month, not the month it was discovered, so the dispute covers every month.