DIDfleet: The Caller-ID Fleet That Runs Itself

Carriers grade every outbound number continuously. A number that dials too hard, from the wrong place, for too long starts getting rejected as spam, and the leads behind it stop hearing a ring. Most floors manage this by hand, monthly, from a spreadsheet. DIDfleet is the product we built to manage it every fifteen minutes, from the dialer's own logs, with no human in the loop. This is what it did on the first floor it ran on: a high-volume outbound call center on VICIdial, from August 2026.

Calls were being lost before a phone ever rang

The audit came first, read-only, from thirty days of dial logs. Three things were costing this floor conversations that had nothing to do with the customer:

• 12.8% of calls were being declined by carriers on a shared area-code pool. That is one dial in eight rejected before it rings.
• 30,170 leads had been retired by a single carrier error. One transient failure and the dialer permanently stopped calling them. 13,410 of those people had connected on an earlier call.
• 3,601 fresh leads were loaded with a bad value in the caller-ID field one morning. Dialed like that, every one of them fails on its first call.

None of this showed up in a dashboard. All of it was in the logs.

What DIDfleet does

DIDfleet sits beside the dialer. It reads the dial log and the lead tables, and writes one field: the number each lead dials from. Everything else follows from that.

Every lead gets a clean local number. Chosen from the lead's ZIP, not just its area code: same rate center first, then metro, then state, from numbers with a healthy live record.

Every number carries what it can prove. Caps per number come from its measured dial rate and how seasoned it is, counted against leads that can actually dial today. New and rested numbers climb a warm-up ladder, 10, 25, 45 and then 75 dials a day, before they take full load.

The fleet regrades itself every fifteen minutes. Numbers that run hot on carrier rejections are benched, and whole purchased ranges that fail as a unit are benched with them. Their leads move to a healthy local neighbour inside the same cycle, with the old value saved first.

Leads come home. A lead that spilled onto a distant number is re-homed to a local one when a slot opens, in a capped batch each night.

Leads stop dying for the wrong reasons. A lead retired by a carrier error is brought back nightly, on evidence, with a revert file for every run. A bad caller-ID value in a fresh load is caught within minutes, before the list is switched on.

Same leads, same lists. More calls reach the phone.

The before and after here are carrier responses read from the dial log, not survey figures. Same lists, same floor, same dialer.

Connects rose 9.7 points, from 80.4% to 90.1%. Carrier declines fell 63%, from 12.8% to 4.7%. The before period is the smaller sample, 2,639 dials against 49,369; steady state since has run 88.9% connect and 5.4% declined across 89,327 dials.

The cleanest test was an accident. One morning the fleet was switched off by mistake. Delivery fell four points before lunch. Switched back on, it recovered within the hour.

Where the appointments came from

More calls ringing and more leads alive is where the appointments come from. In the first full week on the fleet the floor booked 450 appointments against 400 the week before, up 12.5%, with every single day at or above the same day the prior week. It did that on a quarter fewer dials.

We also show the other lever. Fresh leads answer about five times as often as worn ones, and DIDfleet's nightly reports put lead wear next to number health. When the numbers say the problem is the list and not the fleet, the report says so, and we say not to buy more numbers.

Leads stop dying for reasons that have nothing to do with the customer

• 23,703 leads restored. A transient carrier error marks a reachable lead dead on most dialers. The fleet brings them back nightly, on evidence, with a revert file for every run.
• 4 minutes to fix a bad load of 3,601 leads. The wrong value in the caller-ID field, caught and replaced with real local numbers before the lists were switched on. None was dialed.
• 100% of active leads carry a valid number. 114,028 of 114,028 across two campaigns on the day this was written. A lead cannot fall through to a toll-free fallback.

Operated, not just installed

The fleet is watched by resident loops that name what they see. A compliance watcher flags a do-not-call, booked or transferred lead that becomes dialable again within five minutes, with the status change that did it. A pacing watcher gives the dial level, the safe ceiling and who changed it when the floor is pushed past what its agents can absorb. New purchased ranges are graded as a unit while they season. Each territory is tracked against its own healthy floor, so one bad market cannot hide inside a good average. An hourly briefing puts it in six lines.

Purchasing is computed, not guessed: per area code, per campaign, from demand against surviving clean supply, with vendor dispute files carrying the evidence when a range arrives pre-burned. No paid reputation lookups are needed; in a sealed test, grading from carrier responses alone agreed with a paid lookup on 8 of 10 bad numbers.

Every write saves the previous value first. One file stops every write. It starts in shadow mode, computing every decision and writing nothing until its numbers match yours; the latest install went from upload to a verified shadow run in under an hour.

Reputation management is not a purchase. It is an operating system. This one runs itself, and it is available as a product at didfleet.com.