In outbound contact, the most expensive second is the one right after the customer picks up. Reaching someone takes schedules, numbers, retries and infrastructure; losing them takes one sentence, and very often it is the sentence that says who is calling.
We call this rung of performance continuity: the share of answered calls in which the person stays on the line long enough for a conversation to begin. It is the inverse of the hang-up rate after the greeting, and in our experience it is the most underestimated number in outbound work. This piece explains why it matters, what actually sits in that first sentence, and what a buyer should ask about it.
Where continuity sits on the ladder#
Every process we run, whether collections, lead qualification, appointment confirmation or surveys, is measured on the same four rungs. Each rung only exists if the one beneath it has been reached.
- Reach. Did we get through to the customer at all? Without reaching the customer, there is nothing else.
- Continuity. Once they answered, did they stay, or did they hang up on hearing the opening?
- Key information. Did the conversation produce what it exists for: a promise to pay (PTP), a declared interest, a confirmed or cancelled visit, a fully completed survey?
- Business effect. Did the payment arrive, the contract get signed, the patient attend? We influence this rung only indirectly, and we say so.
Reach gets most of the attention because it is visible in dialler statistics, and the business effect gets the attention of the board. Continuity sits between them and is easy to miss, because a call that lasted four seconds still shows up as “connected”. A campaign can have excellent reach and still produce almost nothing, simply because a large part of the people it reaches leave before the second sentence.
Clients often assume outbound is simpler than inbound. Someone who calls you will cope even with an IVR. Someone you call may not even want to pick up, let alone keep talking. – Dawid Wójcicki, Founder & CEO of Vee
The name carries the customer’s expectations#
Some bases have a high hang-up rate that has nothing to do with the quality of the conversation. Someone hears the name of a particular bank or a television operator and puts the phone down before anything else is said. That is not the caller’s fault and not the dialogue designer’s fault; it is what the customer expects that call to be about.
The expectation differs by situation. A customer in early delinquency who hears their bank’s name knows, or suspects, why it is calling, and avoidance is a natural first reflex. A subscriber who hears a telecoms brand often assumes a sales pitch. A patient who hears the name of their clinic usually has a reason to listen, because the call is about something already arranged. The same opening technique will behave very differently across these three, which is why we never transfer a “good opening” from one process to another without testing it again.
This has a commercial side we learned early. In bases where many people hang up on hearing the brand, an operator still has to schedule every attempt, rotate numbers, keep the infrastructure ready and try again, yet very few minutes result. Per-minute billing makes such bases unattractive to serve and quietly rewards whoever avoids them. It is one of the reasons new agreements are no longer billed by the minute.
What is actually in the first sentence#
The opening of a regulated outbound call is not a free creative space. Several elements must be there, and some must not be there yet. The design work is in the order, the wording and the delivery.
| Component | Fixed or designed | The design question |
|---|---|---|
| Who is calling (the organisation) | Fixed: never hidden | Where in the sentence it falls and what comes with it |
| That the caller is an AI | Fixed: always disclosed, at the start | Plain wording that informs without sounding like a disclaimer |
| Whether we have the right person | Fixed in collections: verification first | How to ask without revealing anything to a third party |
| Why we are calling | Designed | How specific to be before verification, and how much benefit to signal |
| Voice, pace and tone | Designed | Studio recording or synthesis, energy level, pauses |
| Moment of the call | Designed | Time window and day for this segment of the base |
Fixed elements follow the client’s rules, local law and AI Act Article 50; designed elements are tested.
Two of these constraints shape everything else. First, Vee always introduces itself as AI. Since 2 August 2026 that is also a legal transparency obligation under Article 50 of the AI Act, but it was our practice long before. Second, in collections, debt details are never disclosed before identity is verified, and this is enforced by the platform rather than left to a line in a script. If someone other than the customer answers, the call ends without disclosure and a retry is scheduled.
So the opening cannot say “we are calling about your overdue instalment” to whoever picks up. It has to name the organisation, disclose the AI, establish that it is talking to the right person, and give that person a reason to continue, all in a few seconds and all in a tone that does not sound like a trap.
The variables that move continuity#
Within those constraints there is a lot to work with. These are the areas we work on, always with the client’s approval of any content change.
- Order. Greeting, name of the organisation, disclosure, identity question and purpose can be sequenced in several legitimate ways. The order changes what the listener is deciding about when they decide whether to stay.
- Wording. Small differences in wording are heard differently, and we test them rather than assume.
- Voice. In cold contact, recorded studio voices often hold people better than synthesis. We learned that on our first outbound campaign, when conversion only appeared after synthesis was replaced by a recorded lector.
- Tone and pace. An opening that sounds rushed signals a script; one that sounds hesitant invites an interruption. The first two seconds of a voice are judged before the words are.
- Moment. Answer rates and willingness to talk vary by time of day within the same base. Inside one base there are islands of people who cannot talk during working hours and others who prefer exactly then.
- Recognition before the call. A text message sent ahead of the attempt and the outbound number used both change how the call is received before a word is spoken.
Because Vee employs no call agents, a change to the opening applies to every live conversation at once. That makes continuity unusually testable: there is no retraining, no drift between shifts, and no agent who quietly improvises a better opening that nobody records. Where the client agrees, we run A/B tests and compare continuity on the same buckets and portfolios, so that the base itself is not the hidden variable.
Measuring it honestly#
A single hang-up rate for a campaign tells you very little. To act on continuity you need to know where in the opening people leave, and that requires dialogue-level data rather than call-level totals.
We separate the technical status of an attempt (no answer, voicemail, busy, dropped) from its dialogue status (what was said, and how far the conversation got). A hang-up during the organisation’s name is a different finding from a hang-up after the disclosure, after the identity question or after the purpose. Each points to a different fix, and only the first is mostly about the base rather than the dialogue.
Then we listen. Every day, in every project, people at Vee listen to selected calls, especially unusual ones. Short calls are among the most informative: you hear whether the person hung up on the brand, on the voice, on a pause that went on too long, or because they were driving. Automated measures tell you that continuity dropped; listening tells you why.
Finally, we compare like with like. Warm leads, cold leads, recovered leads, repeat customers and one-off customers all behave differently when they hear the same opening. A continuity figure is only meaningful against a comparator on the same kind of base. Moving a campaign onto a friendlier base will improve the number without improving anything you control.
What we will not promise#
We cannot jump over a bad base. If a large share of a list hangs up on hearing a brand, no wording will turn that into a high continuity rate, and a supplier who promises otherwise is either planning to hide the brand or measuring something else. What a well-designed opening does is stop losing the people who would have stayed: the ones who left because the sentence was too long, the voice too flat, or the purpose too vague.
We also keep continuity in its place. It is a means to the third and fourth rungs, not an end. An opening that keeps people on the line by delaying the point of the call will raise continuity and lower the share of conversations that produce a promise, an answer or a decision. We judge every opening change by what happens further down the ladder.
What to ask a supplier#
- Do you report continuity separately from reach, and can you show where in the opening people leave?
- How do you disclose the AI and the organisation, and who approves that wording?
- In collections, what happens if someone other than the customer answers, and is that enforced by the platform or by the script?
- When you test an opening, what is the comparator, and how do you keep the base constant?
- Who listens to short calls, how often, and what changed last month because of it?
- If a better opening is found, how quickly does it reach every live conversation?
Continuity rarely appears in a demo, because a demo is a conversation with someone who wants to listen. In production, the first sentence meets people who did not ask to be called. Designing for them, without hiding who is calling, is the part of customer service that technology alone does not do for you, and it is the part we spend much of our listening time on.
Follow The First Second
Occasional issues on the mechanics and economics of AI-powered customer contact. No noise, no vendor news.
Follow via RSS