Somebody asks what a good contact rate is. Somebody else says 15%. A third person says their floor runs 40% and everyone assumes they are either lying or working a much better list.
Both numbers can be true simultaneously, and usually are, because contact rate is not one metric. It is a family of metrics that share a name, and the disagreement is almost always about the denominator rather than about performance.
Four numbers, one name
| What is being measured | Denominator | Reads high or low |
|---|---|---|
| A human picked up | Dial attempts | Lowest of the four. Multiple attempts per record push it down mechanically |
| A human picked up | Unique records touched | Higher, and the most commonly quoted version without saying so |
| The intended person answered | Unique records touched | Lower again, and by far the most useful of the four |
| The intended person answered | Records with a working number | Highest. Flattering, and it hides the data-quality problem entirely |
The spread between the first and last rows on the same file, worked by the same team on the same day, is large enough to make any cross-company comparison meaningless. Which is why the benchmark question has no answer and the definition question does.
The one contact rate we can actually source
We build a canvassing app, so the channel we can speak to with evidence is the door rather than the phone. In 237,329 logged door outcomes across 7,241 real field days, derived from anonymised logs on 17 August 2026, someone answered at 37.0% of knocked doors. The other 63.0% reached nobody. Within that, the rate moved from 31.5% in the 9 a.m. hour to 46.2% in the 8 p.m. hour, and from 34.7% on Saturdays to 38.1% on Mondays.
That is a door-knock number and it does not transfer to a phone. A knock reaches a fixed address; a dial reaches a number that may have been ported, reassigned, screened or flagged as spam by a carrier. We are citing it for two reasons only: it is the one contact rate on this site that comes with a methodology, and its shape — a rate that swings by a third depending on the hour — is the shape the phone version has too.
We do not publish an outbound dialling benchmark because we do not run the dialers and would be guessing. The same page explains which figures we deliberately did not publish and why.
Define yours in one sentence
Write the definition down, in a document, with the version date on it. This sounds bureaucratic and it is the entire discipline.
Contact rate = records where a human answered ÷ unique records attempted, within a fixed attempt policy
Four decisions are buried in that sentence, and all four must be fixed:
- What counts as a contact. Any live human, or the intended party? Pick one. Track wrong-party as its own disposition either way, because it is the only signal that tells you whether your data is matching to the right household.
- What counts as attempted. A record you dialled once and a record you dialled nine times are not comparable. Fix the attempt policy — how many tries, over how many days, at what spread of hours — or the metric moves whenever the dialer configuration does.
- What window. A record attempted on Friday and reached on Monday belongs to one of those weeks. Decide which.
- Which records are excluded. Suppressed, scrubbed, invalid on arrival. Excluding them is defensible; excluding them silently is how a floor congratulates itself for a data problem.
Then hold it still. A definition that changes is worse than a bad definition, because it destroys your own history — the only benchmark you were ever going to have.
What actually moves it
Roughly in order of effect size, and only one of these is about how hard the floor is working.
| Factor | Direction | Notes |
|---|---|---|
| Whether the number belongs to the right person | Large | A file matched to whoever is associated with an address rather than to the owner on title produces conversations with the wrong household, which count as contacts and convert at nothing |
| List age and how many times it has been worked | Large | The decay is structural. Reachable records get reached first |
| Mobile share and line type | Large | Landline-heavy files inflate match rate and deflate contact rate |
| Caller-ID reputation and spam labelling | Large and invisible | The call is placed, nothing errors, and the phone never rings the way you think it does |
| Time of day and day of week | Moderate | Real, and the pattern is usually the opposite of the schedule the floor is on |
| Attempt policy | Moderate, and it is a definition change | More attempts per record raises contacts per record and lowers contacts per dial |
| Rep behaviour | Small | Reps affect what happens after the pickup far more than whether there is one |
The last row is the uncomfortable one, because contact rate is the metric most often used to manage reps. Almost everything above it is a property of the data and the telephony.
Diagnosing a drop
Three questions, in this order, and they take an afternoon:
- Is it falling on new segments too? Split contact rate by segment age. If fresh loads still perform and worked segments do not, the file is exhausted and this is a rotation problem, not a coaching one.
- Has the wrong-party share moved? If contacts are holding but right-party contacts are falling, the list changed — a new vendor, a new match method, or an older extract than you thought.
- Are the calls being labelled? Check your outbound numbers against carrier spam labelling. A reputation problem shows up as a rate that fell across every segment simultaneously, including brand-new ones, which no list problem does.
Those three tests separate a data problem from a telephony problem from a genuine performance problem, and the answer is a list problem more often than any floor expects. List hygiene for call centers covers the routine that prevents the first one, and outbound dialer setup covers pacing, caller ID and disposition design underneath it.
The number to manage instead
Contact rate is a diagnostic, not a goal. It can be raised by dialling fewer records more often, by loosening what counts as a contact, or by excluding the hard records from the denominator — none of which produces another sale.
The number worth managing is cost per right-party conversation: what you paid for the data, divided by the conversations with the person you meant to reach. It is immune to definitional games, it makes two vendors comparable on the same unit, and it is the only version of this arithmetic that connects to the P&L. The skip tracing cost calculator runs it on any two quotes, and skip tracing accuracy explains why the match rate on the invoice is not an input to it.
Disclosure: Scout Data sells the data at the top of this funnel, which is exactly the input the table above says matters most — so read that section knowing who wrote it. Our List Builder matches phones to the owner on title rather than to the address, returns line type on every number and replaces numbers that stop working. We do not claim a contact rate for it, for the reason this whole page is about: we would be quoting your floor’s number, not ours.
Frequently asked questions
What is a good contact rate for an outbound call center?
There is no honest single number, and the sites publishing one are quoting each other. Contact rate depends on what you divide by — dials, unique records, or reachable records — on line type, on list age, on caller-ID reputation and on the hour of day. A figure quoted without all of those is not a benchmark, it is a decoration. Measure your own, hold the definition still, and compare weeks to weeks.
Is contact rate the same as connect rate or answer rate?
Not usually, and the difference is the whole problem. Answer or connect rate normally counts any live pickup per dial attempt. Contact rate more often means reaching the person you intended, per record. Right-party contact narrows it further. Three metrics, three denominators, one vocabulary — which is how two teams compare numbers that were never measuring the same thing.
Why is our contact rate falling?
Most often because the reachable records in the file have already been reached. Every pass removes connects and leaves a residue enriched in dead numbers and people who never answer, so the rate declines as a property of the file rather than of the floor. Before coaching anyone, check the rate by segment age — if new segments still perform, the reps are fine and the list is the problem.
Does Scout Data publish an outbound contact rate benchmark?
No, and we will not, because we do not run the dialers. What we do publish is a door-knocking contact rate derived from our own canvassing logs, with a full methodology and the figures we chose not to publish listed alongside it. Different channel, same standard: if we cannot show where a number came from, we do not print it.