A text message works best around a conversation, not instead of it. Sent before a call, it can raise the chance that the customer answers; sent after it, it can turn a spoken agreement into something the customer acts on. Sent in place of the call, it usually reaches fewer people, produces fewer replies and loses the information that only a conversation can surface.
We run SMS as part of the same process as the calls, on the same case and the same schedule, in early collections, appointment confirmation and other outbound processes. That has given us a clear view of what a text does well, what it does badly and when it starts to compete with the conversation it was supposed to support.
What a text can and cannot do#
SMS is cheap to send, arrives silently and can be read at any time. Those are real strengths. But a text cannot verify who is reading it. It cannot ask a follow-up question, notice hesitation, negotiate a date or hear that the customer has lost their job. Compared with a call, it also has a smaller reach and a lower response: many texts are never read with attention, and fewer still are answered.
The deeper difference is in what the customer gives you. A declaration made in a conversation, a date the customer chose and said out loud, translates into real behaviour better than a tap on a link or a one-word reply. That is our experience across processes, and it is why we treat the conversation as the main channel and the text as its complement.
| Call | SMS | |
|---|---|---|
| Reach and response | Higher, with persistence | Lower; easy to ignore |
| Identity verification | Yes, before any sensitive detail | No; goes to a number, not a person |
| Two-way negotiation | Yes | Very limited |
| Detecting uncertainty or hardship | Yes | Rarely |
| Leaves a written record for the customer | No | Yes |
| Works at any hour for the reader | No; within calling windows | Yes |
Qualitative comparison from Vee's operating experience.
Before the call: making the unknown number less unknown#
People increasingly ignore calls from numbers they do not recognise, and handsets help them do it. A short text sent ahead of the attempt can tell the customer who will call and roughly why, so the call that follows is less of a stranger. In our collections processes, the pre-call SMS exists to warm the customer and increase answer rates. That is its only job.
In regulated processes, the pre-call text has a hard constraint. It goes to a phone number, and we do not yet know who holds the phone. In early collections, debt details are never disclosed before identity is verified, and that rule is enforced by the platform, not left to wording. So a pre-call text must be neutral: who is contacting the customer and how to reach them, not how much is owed or for how long. A text that discloses the debt to whoever reads it is not a reach tactic; it is a data protection incident.
Timing matters as much as wording, and the interval between the text and the call is one of the parameters we test per process rather than assume.
After the call: turning a promise into an action#
Once a conversation has taken place and the customer has been verified, a text becomes far more useful. In early collections, a post-call SMS confirms the commitment that was made: including the date the customer chose. In appointment processes, it confirms the visit or the new slot after a reschedule. In other outcomes it follows up on what was agreed, for example the next step after a request.
This is where SMS earns its place most clearly. The promise was made in the customer's own words; the text gives them a written record of it at the moment they need to act. It also helps the business side. A promise to pay is a declaration, not a payment, and anything that raises the share of kept promises matters on the fourth rung of the ladder, where the real result is measured in the client's systems.
Instead of the call: where the text cannibalises the conversation#
The risk appears when a text tries to do the conversation's job. If the pre-call message contains everything, a payment link, an amount, a confirm button, some customers will act on it and many will simply not pick up the call that follows, because they think they already know what it is about. The ones who do not act are now less likely to talk to you.
Sometimes that trade is acceptable. If the only goal is a payment and the text produces it, the result is real; contact has value even when no conversation takes place, and some customers pay simply because their bank got in touch. But in most processes the conversation produces more than the headline result. It reveals a wrong number, a customer in financial difficulty, a visit that will not happen even though the patient replied “OK”, a broken promise that needs a different approach next time. A text-only channel loses all of that, and the client's analysts lose it with them.
We call it appointment confirmation. Internally we call it cancellation, because that is the business goal: learning early that a visit will not happen. – Dawid Wójcicki, Founder & CEO of Vee
Healthcare shows the problem clearly. Many clinics confirm visits by text alone, or by staff who have no time to try a second or third time. A text reply confirms that someone read the message. A conversation is much more likely to surface the patient who is not really coming, early enough for the slot to be offered to someone else.
As a fallback: when calls have not got through#
There is a third role, and it is a legitimate one. When repeated attempts have not reached the customer, a text is the right fallback. In appointment confirmation, calls are planned around the deadline for refilling the slot, and a text follows when they fail. The text does not replace the conversation in that case; it replaces silence.
One contact, not two channels#
From the customer's side, a text and a call from the same company are one contact. That has practical consequences. A text should not arrive minutes after the customer has already agreed something on the phone and contradict it, or ask them to do what they have just done. A call should not follow a text so closely that it feels like pressure. And the text should sound like the conversation: the same name, the same register, the same reason for getting in touch.
It also matters for limits. Calling windows and maximum contact attempts are configured per market and enforced by our platform, and texts belong in the same contact plan as calls. A customer who receives a call, a text, another call and another text within one afternoon has been over-contacted, whatever each channel's own log says. Running SMS on the same case, the same schedule and the same history as the calls is the simplest way to prevent that.
The same logic travels across sectors. In energy and utilities, where field visits for readings, meter replacements and service work are paid whether or not anyone opens the door, a call confirms the visit and a text after it gives the customer the date and time in writing. In insurance renewals, a text can make the call recognisable without trying to sell the policy in 160 characters. In each case the conversation does the work that needs judgement, and the text does the work that needs a written record.
Measuring SMS honestly#
Delivery and read receipts tell you that a text arrived. They say nothing about whether it helped. The only fair test is a comparison on the same base: cases that received the text against comparable cases that did not, measured on the rungs that matter. Did the pre-call text raise reach without lowering continuity? Did the post-call text raise the share of kept promises, or attendance, in the client's own records? We run these comparisons as live A/B tests in production, because the answer differs by sector, market, segment and timing.
Cannibalisation has to be measured explicitly, because it hides well. A content-heavy pre-call text can lift payments in the short run while lowering the share of cases that end in a conversation, and with it the statuses your analysts depend on: promises with dates, hardship declarations, wrong numbers, requests for a human. So we look at both sides of the comparison: the result the text was meant to produce, and the information the conversation would otherwise have returned. If the second falls faster than the first rises, the text is doing the conversation's job badly.
What to ask#
- Is SMS run on the same case and schedule as the calls, or by a separate team with separate data?
- What exactly may a pre-call text say before the customer's identity is verified?
- Which outcomes trigger a post-call text, and what does it confirm?
- When does a text become the fallback, and after how many failed attempts?
- How do you test whether a text helped, and on which rung is the effect measured?
A text message is a small thing. Placed well, it makes the conversation more likely and its result more durable. Placed badly, it quietly takes the conversation's place. We design both channels together because, for the customer, they are one contact.
Follow The First Second
Occasional issues on the mechanics and economics of AI-powered customer contact. No noise, no vendor news.
Follow via RSS