Skip to content
AI Dialer

Opinion

The same 1,000 calls produced six different “connect rates”

  • 7 min read

By Last updated

The short answer

Connect rate has no industry-standard definition. Counting answered signalling gives one number, excluding voicemail gives a lower one, excluding short calls lower still, and right-party contact rate lower again, and changing the denominator from dials to unique contacts moves it in the other direction. The same 1,000-call log legitimately scores anywhere from 6.0% to 34.0%.

Two vendors will quote you connect rates in the same pitch deck and neither will define the term. A board deck will show it trending up while pipeline is flat and nobody in the room will ask what changed in the denominator. It is the most-quoted number in outbound calling and it is the least specified.

So here is one call log, scored every legitimate way, with the arithmetic in the open.

The log

A thousand dials against a US B2B list on a normal Tuesday. Every category below comes from something a telephony platform can actually observe: SIP response codes, answering-machine detection, call duration, and a disposition the rep sets. No estimates.

1,000 dials, broken down by what actually happened
OutcomeCallsHow it is detected
Invalid, busy, or disconnected60SIP 4xx/5xx before ringing
Rang, nobody answered620No answer signal before timeout
Answered by voicemail or a machine180Answering-machine detection
Answered by a human, hung up under 6s45Answer signal, short duration
Human conversation, wrong person or gatekeeper35Rep disposition
Human conversation with the intended contact60Rep disposition
Total1,000

Six defensible numbers from one log

Each of these is a definition somebody in this industry uses in earnest and will defend in a meeting. None of them is a lie. That is the problem.

Same day, same list, same reps
DefinitionArithmeticResult
Answered signal ÷ dials(180 + 45 + 35 + 60) ÷ 1,00032.0%
Answered ÷ deliverable dials (invalids removed)320 ÷ 94034.0%
Human answers ÷ dials (voicemail removed)140 ÷ 1,00014.0%
Conversations ÷ dials (short hang-ups removed)95 ÷ 1,0009.5%
Right-party contact ÷ dials60 ÷ 1,0006.0%
Right-party contact ÷ unique people dialed (400)60 ÷ 40015.0%
One log of 1,000 dials broken into 60 invalid numbers, 620 unanswered, 180 voicemails, 45 human hang-ups under six seconds, 35 wrong-person conversations and 60 right-party conversations, then scored six ways: 32.0% counting any answered call, 34.0% with invalid numbers removed from the denominator, 14.0% counting human answers only, 9.5% counting conversations, 6.0% counting right-party contacts, and 15.0% dividing right-party contacts by unique people dialled.
Every one of these six is a definition somebody in this industry defends in a meeting. None of them is a lie.

6.0%

Lowest honest reading of this log

34.0%

Highest honest reading of the same log

5.7×

Spread available with zero change to performance

A 5.7× improvement is available to any vendor, any quarter, without touching a single call. It is a spreadsheet change.

The two levers are the numerator and the denominator, and both are unregulated. Move voicemail into the numerator and the rate roughly doubles. Strip invalid numbers out of the denominator and it lifts again. Switch from dials to unique contacts and it lifts a third time. Do all three and you have a case study.

Which line items get quietly reclassified

Once you know the levers, the specific tricks are predictable, and they show up in benchmark reports far more often than in anyone's own dashboard.

  • Voicemail counted as a connect. Technically the call was answered, by a machine that will never call you back. This single choice is worth roughly double on most B2B lists.
  • Invalid numbers removed from the denominator. Defensible on its own terms, and it conveniently rewards a dirty list: the worse your data, the bigger the lift when you exclude it.
  • Sub-6-second calls kept in. A pickup-and-hang-up is a rejection with extra steps, and it is the most common human outcome on a cold list.
  • Unique contacts as the denominator. Reasonable for a multi-touch cadence, and it means every extra attempt at the same person improves the rate rather than diluting it.
  • Answering-machine detection turned off. No machine detection means no machine category, which means every voicemail lands in the human bucket by default.

The question that settles it in one line

Ask any vendor quoting a connect rate: what is in the numerator, what is in the denominator, and is voicemail counted? If they cannot answer in a sentence, the number is decoration. If the answer is "answered calls over dials", you now know it includes voicemail and you can halve it in your head.

Why this survives, and it is not only vendors

It would be neater if this were purely vendor dishonesty. It is not. Every layer of the reporting chain is rewarded for the same distortion, which is why the definition drifts upward on its own without anyone deciding to move it.

Everyone in the chain benefits from the loosest definition
WhoWhat they want the number to showWhat they lose by tightening it
The vendorTheir dialer improves connect ratesThe headline claim in the deck
The agencyThe campaign is workingRenewal leverage
The SDR managerThe team is productiveAn uncomfortable pipeline conversation
The repActivity targets metTheir number against quota
The CMOA clean slideNothing, which is why they never ask

A metric nobody in the reporting chain wants to be lower is not a metric. It is a mood.

This is also why connect rate drifts up over a year while meetings booked stays flat. Nobody falsified anything. The definition simply relaxed a little at each hand-off, in the direction everyone preferred.

What to measure instead

You do not fix this by finding the one true connect rate. You fix it by measuring things whose definition cannot slide, and by writing your own definitions down before you start looking at the numbers.

  1. 1

    Write the definitions down first, in one place

    One sentence each for dial, answer, conversation, and right-party contact, with voicemail explicitly in or out. Put it in the same document as the dashboard. A definition that lives only in a query gets changed by whoever edits the query.
  2. 2

    Report right-party contacts per hour worked

    The denominator is hours, which nobody can reclassify, and the numerator is a human conversation with the person you meant to reach. It moves only when the list, the timing, or the dialing improves.
  3. 3

    Track cost per right-party conversation

    Total spend (licences, minutes, numbers, data) over right-party contacts. It catches the failure mode where a dialing mode lifts volume and destroys quality, which every rate-based metric hides.
  4. 4

    Break every rate out by number

    One damaged number can drag a whole team's aggregate down while every other number is fine. Per-number answer rate is where deliverability problems actually show up first; see number reputation management.
  5. 5

    Keep a fixed control segment

    Hold a slice of the list on unchanged settings. When a rate moves, the control tells you whether the change came from the campaign or from the market, or from the reporting.

The honest benchmark

There is no credible industry connect-rate benchmark, because there is no shared definition to benchmark against. Your own number, defined in writing and tracked over time on a fixed definition, is worth more than every published figure combined. The full metrics reference has the formulas.

The uncomfortable part

If you tighten your definitions this quarter, your connect rate will fall. It will fall a lot, plausibly from something in the thirties to something around six. Nothing about your team's performance will have changed.

That drop is the most valuable reporting change most outbound teams can make, and it is the one that is hardest to survive politically, because the number that goes down is the number someone already put on a slide. Which is precisely why it has not happened yet at most companies, and why the ones that do it stop confusing a busy dialer with a working one.

Frequently asked questions

What is the difference between connect rate and contact rate?
Connect rate usually counts any answered call, which on most platforms includes voicemail. Contact rate counts calls where a human conversation took place. Right-party contact rate counts only conversations with the person you intended to reach. On the same 1,000-call log those three come out at 32.0%, 9.5%, and 6.0% respectively; none of them wrong, all of them different.
What is a good connect rate for cold calling?
The question cannot be answered without a definition, which is why published benchmarks are unusable. Counting only right-party contacts against total dials, mid single digits is normal on a cold B2B list. Any figure above 25% almost certainly counts voicemail as a connect.
Does voicemail count as a connect?
It depends entirely on the platform, and most count it because the call was technically answered. Ask explicitly. Whether voicemail is in or out of the numerator is typically the single largest factor in the number you are shown, worth roughly a doubling on a normal B2B list.
Which outbound calling metric is hardest to game?
Right-party contacts per hour worked, and cost per right-party conversation. Both have denominators (hours and money) that cannot be quietly reclassified, and both fall when a change lifts activity at the expense of quality.
Why did our connect rate improve while meetings booked stayed flat?
Almost always a definition change rather than a performance change: answering-machine detection was switched off, invalid numbers were dropped from the denominator, or the denominator moved from dials to unique contacts. Check the query behind the dashboard before crediting the campaign.

Sources

  1. Combating Spoofed Robocalls with Caller ID AuthenticationFederal Communications CommissionThe STIR/SHAKEN framework, the attestation levels carriers sign calls with, and the mandate requiring providers to authenticate caller ID.
  2. 47 U.S.C. § 227 — Restrictions on the use of telephone equipmentCornell Legal Information InstituteThe Telephone Consumer Protection Act itself — the consent requirements, calling-hours limits, and private right of action.
  3. ITU-T Recommendation E.164 — The international public telecommunication numbering planInternational Telecommunication UnionThe international number format, the 15-digit maximum, and how country codes and national numbers compose.

See it working: call tracking

Call tracking assigns unique phone numbers to marketing sources, or swaps the number on your website per visitor, so every inbound call is attributed to the campaign, keyword, or ad that produced it. It closes the loop between ad spend and actual phone conversations.

  • No subscription
  • Numbers in 100+ countries
  • Compliance built in