Technical customer success fundamentals

Support is reactive and ticket-shaped: a customer has a problem, raises it, and the interaction ends when the problem is resolved. Customer success is accountable for whether the customer gets the business result they bought, which nobody raises a ticket about. That difference sets everything else. Support is measured on resolution time and satisfaction per interaction, customer success on retention, adoption and expansion across a contract term. It also means a customer success manager can be doing badly while every ticket is closed on time, because the account is quietly not using the thing it paid for. The cost of the model is that the outcome depends heavily on decisions inside the customer's organisation that you can influence and cannot control.

Candidates quote net retention as a single figure without being able to say what went into it, which an interviewer tests by asking for the arithmetic.

Cohort: the 10 accounts you had at the start of the year
Starting annual recurring revenue          1,000,000

During the year:
  2 accounts churned entirely                -140,000
  1 account reduced its licence count         -30,000
  4 accounts expanded seats or modules       +220,000
                                            ---------
Ending ARR from that same cohort              1,050,000

Gross retention = 1,000,000 - 140,000 - 30,000 = 830,000
                  830,000 / 1,000,000 = 83%

Net retention   = 1,050,000 / 1,000,000 = 105%

Two numbers, one cohort, and the difference between them is the whole story. Gross retention counts only losses and is therefore capped at one hundred per cent; it answers "how much of what we had did we keep". Net retention adds expansion back in and can exceed one hundred; it answers "did this cohort grow".

The trap is reading the net figure alone. At 105% net and 83% gross this business is losing roughly a sixth of its revenue base every year and covering it with growth inside the surviving accounts. That works while the expansion holds and collapses the moment it does not, because expansion is concentrated in a handful of accounts while the losses are spread across many. An interviewer asking about this is checking whether you understand that a healthy net figure can conceal a retention problem that will surface a year later.

Two more details matter. New customers acquired during the year are excluded, otherwise you are measuring sales rather than retention. And the cohort must be fixed at the start of the period, or a shrinking denominator flatters the result.

Logo churn counts customers lost, revenue churn counts money lost, and the two diverge whenever your customers are different sizes. A business can lose a quarter of its logos and almost none of its revenue if the departures are small accounts, or lose one logo and a tenth of its revenue if that logo was an enterprise. Reporting only one of them is how a board gets a misleading picture. The pair is diagnostic when read together: high logo churn with low revenue churn usually points at a weak self-serve or small-business segment, while the reverse points at a concentration risk in a few large accounts and a real problem in enterprise delivery.

Time-to-first-value is the interval between a contract starting and the customer getting a result they would notice if you took it away. The definition matters more than the measurement, because the temptation is to declare value at a milestone you control — provisioning done, training delivered, integration connected — none of which is value to the customer. The customer decides, and the useful test is whether a named person inside their organisation could describe the benefit to their own manager without mentioning your product's features. Defining it credibly forces you to know what the account bought for, which is why interviewers ask about it.

A health score compresses several signals about an account into one comparable number so that a team carrying dozens of accounts can decide where to spend its week. That is its only real purpose, and it is worth saying so, because the common misuse is treating the score as a prediction. It is a triage tool: a red account is an instruction to go and look, not a conclusion. The cost is that any weighted composite can be gamed and can average away the one signal that mattered — an account can score well overall while its executive sponsor has gone silent, which is the single most predictive signal there is.

Interviewers ask because candidates apply for one and describe another, and because the same title means different things at different companies.

Role                     Owns                  Measured on        Technical depth
-----------------------  --------------------  -----------------  -----------------
Customer success         Adoption and the      Retention,         Enough to be
manager                  customer's outcome    adoption, health   credible, not
                                                                  to debug

Account manager          The commercial        Renewal and        Often little
                         relationship          expansion revenue

Technical account        The technical         Escalation         Real, including
manager                  relationship and      outcomes and       logs and APIs
                         escalations           technical adoption

Solutions or             Getting it live       Time to go-live,   Deep, hands-on
implementation engineer   and integrated       integration        build work
                                               quality

The rows overlap in practice and the boundary is drawn differently at every company, which is why the first question to ask in an interview is which of these the role actually is. A CSM job that carries a revenue quota is an account manager with a friendlier title, and a technical account manager role with no log access is a CSM. The distinction changes how you should prepare: a genuinely technical role will put you in front of an engineer who asks how you would investigate a failing integration, while a relationship role will spend the loop on stakeholder scenarios and commercial judgement.

Read down the measured-on column rather than across the rows, because that column is what actually decides behaviour. Anyone measured on revenue will eventually prioritise the conversation that moves revenue, and anyone measured on adoption will keep working an account that has already decided to leave. When one person holds two of these columns at once, which small companies do routinely, the tension is not resolved by the org chart — it is resolved by whichever number their manager asks about first.

Seniority moves along a different axis. A strategic or enterprise CSM carries fewer accounts with deeper executive relationships and is expected to influence the customer's own roadmap; a scaled or pooled CSM carries hundreds and works through programmes, campaigns and in-product guidance rather than named relationships.

A business review is a scheduled conversation with the people who fund the contract, about whether it is delivering. What makes it worthless is treating it as a usage report: a deck of adoption charts tells a sponsor what their own staff did, which they can get elsewhere and do not care about. What makes it valuable is arriving with a claim about their business, evidence for it, and a decision you want from them. The best ones end with the customer taking an action inside their own organisation. It is also the only reliable mechanism for discovering that your sponsor has changed priorities before the renewal tells you.

Land and expand is a go-to-market pattern where the first contract is deliberately small — one team, one use case, one region — and growth comes from repeating it elsewhere inside the same customer. It shifts the hard work from selling to proving, which is why it puts customer success on the critical path for revenue rather than beside it. The practical consequence is that the first deployment is a reference asset, not just a deployment: it needs a result someone will vouch for internally. The cost is that a badly chosen landing team can poison the whole account, because the next team's first question is what the first team thought.