Your numbers aren't blocked. They're labelled.

One morning your connect rate halves and nothing about your list, your reps, or your script has changed. This is the most common way teams discover caller ID reputation: after it has already cost them a week.

The important thing to understand first is that your number almost certainly has not been blocked. It has been labelled, which is a different problem with a different fix.

Who decides, and how

When you place a call, the receiving carrier does a lookup before the handset rings. That lookup hits an analytics provider — the big ones are First Orion, Hiya and TNS — which returns a reputation verdict for your number. The carrier then decides what to show: the number as normal, a caution label, a "Scam Likely" banner, or in the worst case nothing at all because the call was silently dropped.

Three things matter about this arrangement:

It is per-carrier. Your number can be clean on one network and flagged on another, because each carrier uses a different provider with different thresholds. Testing from one phone tells you almost nothing.

It is algorithmic and unpublished. No provider documents its exact thresholds, for the obvious reason that robocallers would optimise against them. Anyone who tells you they know the precise number is guessing.

It is not the same thing as STIR/SHAKEN. Attestation proves you are authorised to use the number you are displaying. That is about spoofing. You can be perfectly attested and still get labelled, because attestation answers "is this caller who they claim to be" and reputation answers "does this caller behave like a nuisance." Passing the first does not exempt you from the second.

What actually trips the flag

The signals are behavioural, and they are the ones you would expect if you think like a carrier trying to spot a robocall operation.

Call velocity. A sudden jump in volume from a number with no history is the single loudest signal. A new number that goes from zero to four hundred calls in a day looks exactly like a burner being spun up.

Short average duration. If most of your calls last under ten seconds, the pattern reads as auto-dialled spam, because that is what spam looks like: ring, pickup, immediate hangup.

Low answer rate paired with high volume. Lots of dialing, very little talking.

Abandoned calls. This one is worth dwelling on because parallel dialers can generate it structurally. If your dialer connects a human and has no rep free, that person hears silence and hangs up. To the analytics engine that is indistinguishable from a robocall checking whether a line is live — which is precisely the behaviour it exists to catch.

Complaints. Direct user reports carry disproportionate weight. A handful of people tapping "Report Spam" moves you faster than thousands of unanswered calls.

Calling outside normal hours. Dialing at times a legitimate business would not is a strong negative signal, and it compounds with everything above.

Notice that a completely legitimate team running an aggressive outbound push with no number management can trip most of these without doing anything unethical. Intent is not an input.

Checking your own reputation

Before you change anything, find out where you actually stand.

The free registries are the place to start. First Orion, Hiya and TNS each operate a portal where you can look up a number you own and, if it has been mislabelled, file a remediation request. This is genuinely worth doing and most teams have never done it.

Beyond that, register your numbers properly. In the US that means having your calling registered under the relevant industry frameworks so your traffic is identifiable as a known business rather than an anonymous source. It does not make you immune, but it gives the analytics engines something to weigh against the behavioural signals.

And test across carriers, not from your own phone. A number that looks fine on your handset may be flagged on the two networks your prospects actually use.

Resting and rotating, done properly

The operational fix is to treat phone numbers as a consumable resource with a health state rather than as fixed infrastructure.

Warm new numbers up. Do not take a brand new number to full volume on day one. Ramp it over a week or two. This is the same logic as warming a new sending domain for email, and for the same reason.

Spread volume across numbers. A pool of numbers each doing moderate volume is far more durable than one number doing all of it. There is a line here — a very large pool of very low-volume numbers is itself a recognised spam pattern — so the goal is reasonable distribution, not maximum fragmentation.

Rest a number the moment it degrades. When a number's connect rate drops materially below the rest of your pool, pull it out of rotation rather than pushing harder. A rested number can recover. A number you keep hammering while it is flagged will not.

Watch per-number connect rate, not the average. This is the part teams get wrong. Your fleet average will look merely disappointing while three individual numbers are effectively dead. The average hides exactly the signal you need.

Fix the underlying behaviour. Rotation buys time; it does not cure the cause. If your abandon rate is generating the flags, you will burn through fresh numbers as fast as you can buy them. Cap your line count so your abandon rate stays low, keep to sensible calling windows, and make sure your list is accurate enough that you are not dialing large volumes of dead numbers.

The uncomfortable part

There is no appeal process with a guaranteed outcome and no vendor can promise your numbers will never be flagged. Anyone claiming otherwise is selling something they cannot deliver.

What you can control is the behaviour that produces the flags and how quickly you notice. The teams that stay clean are not doing anything clever — they are watching per-number connect rate closely enough to catch degradation in hours instead of weeks, and they are running a line count their connect rate can actually support.