ConKairos
Action-based customer value management
Customer value is not a state you predict. It is a trajectory your actions change. We measure the change, action by action: the sale, the service added or dropped years later, and the reaction when it happens.
One measure, three moments. Each of them is already in your systems.
The sale
How a service is sold decides who stays.
Sales call analysis and coaching
Call recordings sit in one system. Cancellations sit in another. Nothing joins them, so the calls that produce loyal customers and the ones that produce a cancellation on the first invoice sound equally acceptable to everyone reviewing them.
Both ends are already in your systems.
What runs between them is not.
Our own research on hundreds of telco sales calls joined the two. Where in the call retention is decided, and what decides it, is identifiable. Two things carry it: what is argued about the service, and when in the call it is argued. Neither works alone, and there is a point in a telesales call after which further selling stops adding retention.
We research what that point costs in your own calls, in the same unit as everything else on this page: days of remaining relationship.
Those findings do not have to stay in a report. We work alongside your analytics and ML teams on what to extract from call data and how to label it, so that ML models train on the elements that actually move retention rather than on proxies for them. Once the signals are defined, ML scoring can run close enough to the call to give agents feedback while the behaviour is still correctable.
Your own calls scored on two scales, timing and argument quality, with the phrasing that keeps customers in your market and the phrasing that costs you them. Then coaching and script guidance, and support while it takes hold.
Qualitative research on call transcripts, quantified with configurational analysis (fsQCA)
The portfolio change
How a service sits on the core decides for how long.
Causal analysis of service effects on customer lifetime in multiservice environments
A service either buys you time with a customer or it does not. We put a number on how much, by comparing customers who were alike in every observable way right up to the moment the service changed, and then following both groups forward against a stated horizon. The unit is days of remaining lifetime. What the estimate does not cover is stated alongside it.
Causal effects in multiservice portfolios: customers who take or drop a service are compared with customers who were alike at that same moment. The result is restricted mean survival time (RMST) over a fixed horizon, reported by direction of change, by service pair and by how long services were held together. RMST translates directly into business terms: days of customer lifetime gained or lost.
Four properties of the effect that a single average number cannot show.
Asymmetric
The edge does not run at the same rate in both directions. What a service gives is not the mirror image of what taking it away costs.
Losing a service can cost up to four times what gaining it ever added.
Heterogeneous
One service, many futures. The same event means different things depending on what it is paired with.
Some service pairs reverse: for them, dropping the add-on is the better outcome.
History-aware
Where a customer is today does not say where they are going. How they got there does.
Customers in the same service state can differ by months of remaining relationship, depending only on the path that led there.
Time-bound
The damage from a removal does not arrive evenly. Some of it keeps accruing for years, but the window for acting is measured in days, not reporting cycles.
Close to half of the first year's damage lands in the first three weeks.
Yours will differ. That is the point of measuring it.
The reaction
What you do when it moves, and when.
History-aware causal input to next best action
A next best action (NBA) engine ranks what to offer next by how likely the customer is to take it, based on where the customer is today. It does not know how they got there, what an action does to a customer already holding three other services, or whether now is the moment.
The effects above make that ranking history-aware: broken out by service pair, by direction and by how long the services were held together. The window gives it a clock.
Knowing what a customer's move costs you is half of it. The other half is what your reaction earns back. That work is ongoing. Stay tuned.
How we work
Scope
What your data will and will not carry: which services, which transitions, which horizon is actually reachable. This phase ends in writing, and it is where most of the disappointment gets removed.
Effects
The estimates themselves: what each action does, with the design that produced it and the uncertainty around it. On the call side, what carries the outcome and where in the call it happens.
Rules
What follows for the business, at the level of detail the effects support: by service pair and by how long services have been held together, or by argument and timing in the call. Where the direction is unresolved, it is marked unresolved.
Re-estimation
The measurement is re-run as your portfolio and scripts change. New services and new scripts fold into the same design rather than starting a new project.
An analysis project, or guidance alongside your team. The difference is how far along this sequence we go with you.
ConKairos is Hannu Peltonen, doctoral researcher at the University of Helsinki, Finland, working on B2C subscription business since 2022. His research covers both ends of the customer relationship: short-term churn and the quality of the sales call, and the long-term lifecycle effects of the service portfolio. Recent work applies causal estimates to next best action automation. Background in global B2B2C business development for value-added services.