Every customer-contact process can be measured on four rungs: did we reach the person, did they stay in the conversation, did we obtain the information the process exists for, and did the business result follow. We use this ladder for every process we run, in every sector, because it makes clear where value is created and where it is lost.

It also forces an honest conversation about responsibility. The first three rungs are largely in the hands of whoever runs the contact. The fourth, a payment, a signed contract, a patient who turns up, depends on things no contact supplier controls. We influence it, often strongly, but indirectly, and we say so in every proposal. This explainer walks through each rung, the levers behind it, the traps in measuring it, and the questions to ask any supplier who reports on it.

Why one ladder for every process#

The processes we run look very different on the surface. A bank wants a promise to pay from a customer in early delinquency. A clinic wants to know, early, that a visit will not happen so the slot can be offered to someone else. An energy company wants a meter-access window confirmed before the technician drives out. A sales team wants interested leads, not lists of numbers.

Underneath, the logic is the same. A conversation cannot happen without contact; information cannot be collected from someone who has hung up; a result cannot follow from information that was never obtained. The ladder turns that dependency into a measurement frame. It lets a head of collections and a head of patient access read the same report structure, and it stops a supplier from presenting a strong number on one rung as if it described the whole process.

The same four rungs across different processes
RungEarly collectionsAppointment confirmationLead qualificationCustomer survey
1. ReachDebtor answers in personPatient answers in personLead answers in personCustomer answers in person
2. ContinuityStays past the opening and identity checkStays past the openingStays past the brand nameStays past the introduction
3. Key informationPromise to pay with a dateConfirmation or cancellationStated interest, qualifiedFully completed survey
4. Business outcomePayment receivedAttendance, or slot refilledSigned contractUsable insight acted on

Definitions are agreed per process with the client. Rung 4 is measured in the client's own systems.

Rung one: reach#

Without reaching the customer, there is nothing else. – Dawid Wójcicki, Founder & CEO of Vee

Reach is the share of assigned cases in which a live person was actually spoken to. It sounds like the dullest rung and it is the one that most often decides the result. In our outbound processes, most dialling attempts never become a conversation: they meet no answer, a busy line, voicemail, a disconnected number or a call-screening service. The difference between a process that reaches a small minority of its base and one that reaches a substantial majority is, in our experience, usually larger than any difference in dialogue quality.

We work reach through six contact parameters. Together they form what we call a reachability strategy.

  1. Outbound number: which number the call comes from.
  2. Destination number: which of the customer's numbers is dialled.
  3. Time window: the hours and days when calls are allowed and when they work, enforced by the system per market.
  4. Channel and content: the call itself plus SMS before or after it, and what those messages say.
  5. Attempt intensity and intervals: how many attempts and how far apart, set per process and per market.
  6. Persona, voice and style: who the customer hears and how.

One decision stays outside this list on purpose: allocation, meaning which accounts are assigned, when and to whom. That belongs to the client's own analysts, who know their portfolios, risk models and capacity. We optimise how the assigned cases are reached; we do not second-guess which cases should be assigned.

What breaks reach#

The usual culprits are mundane. A dialler gives up too early because that is its default. Calls are bunched into the hours that suit the operation, not the customer. Outbound numbers are not managed at all. Contact data is stale. We also tell clients plainly when the base itself is the problem: we cannot jump over a bad base. A list of numbers that were never valid will not become reachable through a better voice.

Reach is measured on cases, not attempts#

A common reporting trick is to divide conversations by dial attempts and call the result a contact rate. It looks precise and says very little, because a process that makes more attempts per case will show a lower rate even when it reaches more people. We measure reach on the case: of the cases assigned in a period, how many reached a live person at least once. Attempts are a cost and a diagnostic tool; they are not the denominator of the first rung.

Speed belongs to this rung as well. In early collections and in lead qualification, a case reached on day one is worth more than the same case reached on day five, because debtors drift into the next bucket and leads cool by the hour. A good reach report therefore shows not only how many cases were reached, but how quickly the whole assignment was covered.

Rung two: continuity#

Continuity is the inverse of hang-ups. Of the people we reach, how many stay in the conversation past the opening? In outbound contact this rung is decided in seconds. The person did not ask to be called, and many decide whether to listen when they hear the company name.

The levers are the first sentence and everything that carries it: its wording and structure, the voice and the delivery. Since 2 August 2026 Article 50 of the AI Act has required people to be informed when they are interacting with an AI system. We have always introduced ourselves as AI, and the craft is to build that disclosure into an opening that still earns the next ten seconds.

Continuity also depends on what comes immediately after the opening. In collections, identity verification must happen before any detail of the debt is disclosed. That rule is enforced by the platform, not left to a script line, and the way the verification question is asked changes how many people stay to answer it. Our first outbound campaign years ago only began to convert when a synthesised voice was replaced by a recorded studio voice. The content had not changed; continuity had.

Measuring where people leave#

A single continuity rate hides the useful part. What matters is where in the conversation people hang up: during the greeting, at the brand name, at the AI disclosure, at the identity question, or when the purpose of the call becomes clear. Because every conversation is logged step by step, we can see the drop-off at each point and test a change against it. A reworded sentence goes live in every conversation at once, so the effect shows within days rather than after a retraining cycle.

Numbers alone are not enough here. Our quality team listens to selected calls every day, in every project, and especially to the unusual ones. A spike in hang-ups after a particular sentence often has a cause that only a human ear catches: a word that sounds like a sales pitch in one region, a pause that feels like a dropped line, a brand name that a customer associates with an earlier dispute.

Rung three: key information#

This is the rung most reports start with, and the one most often misread. Key information is whatever the process exists to obtain: a promise to pay with a date and an amount, a confirmation or cancellation of a visit, a qualified statement of interest, a fully completed survey. It is recorded in the conversation and written back to the client as a status the client's team can act on the same day.

Two things matter here. First, a short conversation is often a better conversation. In banking, cases ending with a promise to pay had an average handling time of 116.5 seconds (historical). The aim is to get to the commitment clearly and respectfully, not to keep the customer on the line. Second, the information has to be specific enough to be checked later. A vague “I'll try to pay soon” is not a promise; a date is.

61.9%First contact closure rateShare of cases closed with the required outcome at the first effective contact.Soft collections, insurance, historical.
43.5%First-attempt completion rateShare of cases converted at the first permitted attempt; here, the conversion is a fully completed survey.CSAT survey, energy sector, historical.

Both figures describe rung three in specific processes. They are not universal benchmarks, and they say nothing on their own about rung four.

The declaration trap#

Rung three is declarative: it records what someone said. It is easy to inflate. A process tuned to maximise promises to pay can collect promises that were never meant, and look excellent until the payments fail to arrive. This is why we never present rung three as the result, and why we design conversations that name a broken promise directly when the customer's history contains one. The test of a good third rung is whether it holds on the fourth.

Rung four: the business outcome#

The fourth rung is what the client actually buys: a payment received, a contract signed, a patient in the chair, a technician who found the door open. It is measured in the client's systems, not ours, after an agreed window, against an agreed source of truth.

We influence this rung. Reaching more of the base, faster, matters enormously in early collections, because each day without contact moves debtors closer to the next bucket and to the point where the bank must build reserves. A well-run contact process changes behaviour; some customers pay simply because their bank called, even without a full conversation. But we do not control the fourth rung, and we think any supplier that claims to is overselling.

Our own year illustrates the point. In the third quarter of 2026, banks in Poland significantly tightened their criteria for consumer credit, as the National Bank of Poland's survey published in November 2026 showed. In lead processes, confirming a customer's interest is one stage; whether the loan is granted depends on the lender's credit policy and its assessment of the applicant. The contact work could be excellent and the fourth rung could still fall, for reasons that sit entirely in the client's risk policy.

+7–13%On-time payment frequencyIncrease in the share of customers paying on time.Early-stage collections, historical, versus a human-agent comparator on the same portfolios.

That is how we prefer to report the fourth rung: as an increment against a comparator, not as an absolute rate. A process that collects from 70% of a bucket may be adding very little if most of those customers would have paid anyway. The honest number is the difference, measured on the same buckets and portfolios, over the same period.

Defining the fourth rung before launch#

Because the fourth rung lives in the client's systems, it has to be defined before the first call, not argued about after the first invoice. For every process we agree the following in writing:

  • What counts as success: a payment of any amount, a payment of the promised amount, a cured account, an attended visit.
  • When it must happen, and the measurement window after the contact.
  • Source of truth: which system and which report decide.
  • Attribution: how cases also worked by other channels or teams are treated.
  • Disputes: how disagreements over data are raised and settled.

This list is less exciting than a dialogue demo, and far more predictive of whether a project will still be running in three years.

How the rungs interact#

The rungs multiply. The share of assigned cases that end in a business result is the product of the rate at each rung, so a weakness low on the ladder caps everything above it. A dialogue that converts brilliantly cannot rescue a process that reaches too few people. This is also why improvements at the bottom are often the cheapest to find: a better retry clock or a reworded first sentence affects every case that passes through.

The rungs can also pull against each other, and a good operator manages the trade-offs openly.

  • Reach versus tolerance. More attempts raise reach up to a point; beyond it they irritate customers and damage continuity on later calls. Market rules and the client's own limits set the ceiling.
  • Continuity versus disclosure. Information that must be given, such as the AI disclosure or a data-protection notice, costs seconds. The answer is better wording, never omission.
  • Information versus outcome. Pressure for more promises can produce weaker ones. We would rather report fewer promises that turn into payments.
  • Speed versus depth. Covering the whole assignment quickly matters in collections and lead processes; sometimes that means shorter conversations with a clear next step rather than longer ones that try to do everything.

Each rung also has its own denominator. Reach is measured on assigned cases, continuity on reached cases, information on continued conversations, outcome on whatever base the client agrees. Mixing denominators is the most common way to make a report look better than it is.

How the ladder runs in practice#

Every case we handle carries four kinds of status: technical (what happened on each dial attempt), dialogue (what was said in the conversation), processing (where the case is in its life cycle) and business (the client's own version of the truth, in its own language). The first three rungs are built from our statuses; the fourth comes from the client's data. Business statuses are defined with each client and are not comparable across projects, which is one more reason we never publish cross-client league tables.

Reports go to the client daily, at case level, so its team can act on them. Then the ladder becomes the agenda of the weekly meeting: which rung moved, why, what we will change next, and what we need from the client. A typical sequence starts at the bottom. In the first days of a controlled pilot on a sample of the database, the work is mostly on reach and continuity. Once those hold, attention moves to the quality of the information collected, and only then to the fourth rung, once the measurement window has closed.

The ladder also shapes how we prefer to be paid. A per-minute price rewards none of these rungs. We prefer billing linked to the work of running the process and to the results the client cares about, with the model for each process set out in the agreement. That aligns our incentive with the ladder rather than with talk time.

What to ask a supplier who reports on your contact#

  1. Which rung is this number on, and what is its denominator?
  2. How do you define reach, and does voicemail count as contact?
  3. What happens to continuity after the AI disclosure and the identity check?
  4. Is the reported success a declaration or a confirmed result in our systems?
  5. What is the comparator for the business outcome, and is it on the same buckets and portfolios?
  6. Which parameters do you control, and which decisions stay with our analysts?
  7. Which of these rungs is your fee linked to?

A supplier who can answer all seven without hesitating understands that customer contact is a chain, not a conversation. A supplier who answers only the third has probably been selling you a voicebot. We built the ladder because we needed it to run our own processes honestly, and we share it because it makes every conversation about results shorter and more useful.

Follow The First Second

Occasional issues on the mechanics and economics of AI-powered customer contact. No noise, no vendor news.

Follow via RSS