Lead callback: how fast to call, who calls and when to stop
The HBR study behind the one-hour rule, a callback sequence a small team can run, which steps an AI can take on its own, and the FCC rules on AI voice calls and opt-outs.

A lead callback is the call your team makes after someone fills in a web form, leaves a voicemail or texts after hours. It costs almost nothing and, on small teams, it is usually the first step to slip. The best-known research on it is fifteen years old, and the gap it found still comes up in what owners post about lost jobs.
What the lead callback research found
In March 2011 James Oldroyd, Kristina McElheran and David Elkington published The Short Life of Online Sales Leads in Harvard Business Review. They sent a web-generated test lead to 2,241 U.S. companies and timed the reply. 37% responded within an hour, 16% within one to 24 hours, 24% took more than 24 hours and 23% never responded at all. Among the companies that did answer within 30 days, the average response time was 42 hours.
A second study in the same article covered 1.25 million leads received by 29 B2C and 13 B2B companies. Firms that tried to contact a lead within an hour were nearly seven times as likely to qualify it as firms that tried even an hour later, and more than 60 times as likely as firms that waited 24 hours or longer. "Qualify" meant a meaningful conversation with a key decision maker, not a closed sale.
The data is old and lead sources have changed since. I still treat the direction as settled. Nothing about phones or buyer patience since 2011 suggests a slower callback works better.
Why callbacks slip
The authors pointed to leads pulled from CRM databases daily instead of continuously, sales forces focused on generating their own leads, and distribution rules based on geography and "fairness".
Small businesses have a plainer version. On X this fall, contractors and agency owners keep posting the same story: the customer called once, nobody called back, and the next company that picked up got the job. The other recurring complaint is the 9 p.m. form fill that sits in an inbox until someone opens a laptop the next morning.
In both cases the business had the lead. It landed in one tool, the person who should call worked in another, and nothing connected them.
A lead callback sequence a small team can run
This is the sequence I'd set up for a team of two to twenty people. The one-hour mark comes from the study. The other timings are my recommendations.
In the first five minutes: acknowledge
Send a text or email saying the inquiry arrived, who will call and roughly when. If the lead came in as a missed call, a missed-call text back does this job. It isn't a callback, but it tells the person someone saw them and gives them a reason to wait.
Inside the first hour: a real conversation
This is the window the research points to. During business hours, route the lead to whoever is free rather than whoever owns the territory. After hours, an AI phone agent or after-hours answering can take the call, ask the qualifying questions you wrote and book a time with a person.
The next business day: a second try with context
If the first call goes to voicemail, the second attempt should mention what the lead asked about. That only works when the person calling can see the form, the text thread and the voicemail in one place instead of three tabs.
After a set number of tries: stop and log it
Pick a fixed number of attempts and a final message, then close the lead with a reason. Open-ended "just checking in" texts irritate people, and the rules below make ignoring an opt-out a legal problem as well.
Deciding what the AI does on its own
Market Canopy lists missed-call text back, after-hours answering, lead qualification and sales follow-up among the jobs Canopy handles, and each action sits under one of three permissions: Allow, Ask or Deny. Mapping callback tasks to those settings is where I'd spend the setup time.
Callback task | Setting I'd start with | Reason |
|---|---|---|
Acknowledgment text after a missed call or form | Allow | Low risk and time-sensitive |
Qualifying questions on an inbound call | Allow | You write the script; the answers land on the customer record |
Booking a callback on a rep's calendar | Allow | Cuts out a round of phone tag |
Outbound AI voice call to a new lead | Ask | Consent rules apply |
Quoting a price or promising availability | Deny | A person should own that commitment |
Any message to a lead who replied "stop" | Deny | Revocation rules |
Start strict and loosen later. Moving a task from Ask to Allow after a few weeks of reviewing what Canopy proposed is easier than pulling it back after a bad call. The same sorting works for leasing inquiries; our post on what a property management chatbot should handle walks through that version.
The rules that apply to automated callbacks
This is not legal advice, and many states add rules on top of the federal ones. Four federal points matter for anyone automating callbacks:
AI voices count as artificial voices. In February 2024 the FCC confirmed that the TCPA's restrictions on "artificial or prerecorded voice" cover current AI technologies that generate human voices, so those calls need the called party's prior express consent. How your web form words that consent matters.
Calling hours. Federal rules bar telephone solicitations to residential subscribers before 8 a.m. or after 9 p.m., local time at the called party's location. Whether a requested callback counts as a solicitation is a question for your counsel. Keeping automated outreach inside those hours, in the lead's time zone, is the simple default.
Opt-outs. Since April 11, 2025, people can revoke consent by any reasonable method, and a text reply of "stop," "quit," "end," "revoke," "opt out," "cancel" or "unsubscribe" counts on its own. Revocations must be honored within a reasonable time not to exceed ten business days.
One piece is still on hold. The requirement that an opt-out from one type of informational message applies to all unrelated robocalls and robotexts from that caller has been waived until January 31, 2027.
What to measure each week
Track the median time from a lead arriving to its first conversation with a person or the AI phone agent. Track the share of leads that never got a conversation at all. Then look at how many closed leads have a reason logged.
The 23% no-response figure from 2011 is a fair benchmark to hold yourself against. If your share of never-contacted leads is anywhere close to it, fix that before paying for more leads.



