Taking an assessment rather than buying one? This page is written for employers. Here is the page for candidates.

Customer success and account management · Mid level

How to assess a Customer Success Manager

Two failures, both specific to this seat. First, the interview selects on warmth, and warmth here is not merely insufficient — it is the trait behind the dominant failure mode. The CSM who is liked by every contact and whose accounts churn anyway is a familiar figure, and their interview performance is excellent, because the qualities that make a customer enjoy a meeting are the qualities that make a hiring panel enjoy one. What actually separates the top quartile is a willingness to be temporarily unpopular at month two, and no interview question distinguishes someone who will do that from someone who says they will. Second, and more concretely: nobody gives the candidate data. Triage is half the job — deciding which handful of a large portfolio needs attention this week — and it is done on usage exports, ticket histories and renewal dates, not on feeling. A candidate who reads a decline in one product area against a rise in support contacts and calls the right person about the right thing is doing something the candidate who calls whoever emailed last cannot do, and no current step tells them apart.

Before anything else, this role has a definition problem that should be settled before an assessment is written, because it determines what is being assessed at all. Two companies advertising for a Customer Success Manager are often hiring two different jobs: one is inbound support with a nicer title and a list of named accounts, the other is a commercial owner expected to defend and grow revenue. Candidates arrive with experience of whichever version they last did, interviewers assess against whichever version is in their head, and the mismatch surfaces a quarter later. Writing a concrete scenario forces the question early, and for many buyers that clarification is worth more than the ranking that follows it.

Taking the commercial version: the CSM's week is a portfolio problem. A portfolio of several dozen accounts, a health score of some kind, a queue of support escalations that are not theirs but land on them anyway, a handful of scheduled reviews and — if they are good — several unscheduled calls they decided to make. The unscheduled calls are where the value is, and choosing them is a data exercise. A drop in weekly active users concentrated in one team, a spike in tickets about a single workflow, an admin who has stopped logging in, a champion whose email now bounces: each is a specific signal pointing at a specific person and a specific opening sentence. The median CSM instead runs a monthly check-in with everyone, which consumes the whole week and surfaces nothing, because a customer who is quietly unhappy does not volunteer it during a friendly catch-up.

The second separator is the ability to say no well. Customers ask for features that will not be built, and the median CSM manages the moment by being vague — "that's definitely on the roadmap conversation" — which converts a small disappointment now into a large betrayal at renewal. The strong version says the thing is not coming, says why, and then does the harder work of finding what the customer was actually trying to achieve and whether any part of it is solvable today. That exchange takes ninety seconds and is the clearest single observable difference between a CSM who retains revenue and one who postpones churn.

The third is internal courage: raising a risk to the AE who sold the account, or to a product lead, before it is undeniable. Health scores go amber and sit there because escalating means an uncomfortable internal conversation, and the entire economics of the function — that fourteen-point retention spread between the top and bottom quartile of comparably priced companies — is made of those postponed conversations.

What the job actually needs

• reading usage and ticket data to choose who to call and why
• opening with a specific observation rather than a check-in
• refusing a roadmap request without retreating into vagueness
• raising a risk internally before anyone asks
• converting adoption into a documented business outcome

How people fail in this seat

• mistakes warmth for account health
• runs a check-in with no agenda and no ask
• promises a feature to end an awkward moment
• lets an amber health score sit because raising it means a difficult conversation
• becomes the customer's advocate against their own company

What most employers do instead

A competency interview that rewards warmth and a well-told account anecdote, sometimes a presentation on "how you would handle an at-risk account", product knowledge questions, and references.

The assessment

About 40 minutes end to end.

The systems it runs in

A customer success platform sitting over product usage data, which is where this seat begins its day. t1's twelve-account export is not a spreadsheet handed to the candidate in the abstract — it is the portfolio view that surface produces, with a health score per account, a usage trend, ticket volume, last login of the named administrator, renewal date and contract value on each row. Gainsight's own documentation describes the Customer 360 as carrying a health scorecard built from qualitative and quantitative signals alongside open calls to action, and the fixture is built to that shape deliberately, because the decoys only work inside it: an account can carry a red score and be safe, and the account that matters can sit at green with a flat usage line composed of one team rising and another collapsing. The candidate drills into the account, opens the person rather than the logo, and afterwards writes t3's internal note onto that account's timeline with the ask raised as a task or call to action carrying a named owner and a due date — which is the difference between flagging a risk and filing one the AE has to respond to.

• Gainsight
• Totango
• ChurnZero
• Catalyst
• Vitally
• Planhat
• Pendo

Any customer success platform with a per-account health view, a usage or engagement signal, a timeline, and a task or CTA object that can be assigned to someone else. Teams without a dedicated platform run the same job from a CRM account object plus a product-analytics dashboard, and the fixture supports that split-screen version unchanged. What cannot be swapped away is the composition problem: the usage data has to arrive broken down by team or seat, because an aggregate number cannot hide a collapse inside it, and a fixture that hands the candidate only totals has removed the thing being measured. Rebuilt against the buyer's own scorecard measures and their own definition of an active user where they can provide a sandbox.

What the candidate actually does

The mark scheme

Each criterion is scored 1 to 5 against written anchors, and every score is reported with the excerpt that earned it. A criterion marked floored is reported as a finding rather than averaged into the total. The first is open; open any other to read its anchors in full.

How it is scored

Weighted mean of the six criteria, 1-5 against the anchors, each attached to the excerpt that earned it. Two rules for this role. The triage answer is reported in full as written, not only as a score, because the reasoning behind the account choice is the most transferable evidence in the session and hiring managers read it faster than they read a number. And the roadmap-refusal exchange is extracted as a standalone excerpt for every candidate, because it lasts around ninety seconds, it is the clearest observable difference between retaining revenue and postponing churn, and it is easy to lose inside a fifteen-minute transcript.

Integrity

• the competitor demo and the internal pressure are never in the brief, and the sponsor discloses them only under a specific prompt
• per-candidate randomisation of the portfolio so the masked collapse sits in a different team, product and account
• the call is run against whichever account the candidate selected, so a shared answer to the triage does not transfer
• keystroke and paste timing on the internal note
• a follow-up question asking which figure in the export supports the confidence level stated in the note

The log describes what happened. It does not produce a cheating verdict — the follow-up conversation is the control, because a statistical accusation is not something we would ask a reviewer to defend.

What you receive

• the triage answer with the candidate's stated reasoning and their opening sentence
• full call transcript, with the point at which the hidden pressure became discoverable marked
• the internal note as written
• per-criterion score with the excerpt that earned it

Who decides

Recommended, with named checks. A reviewer reads the roadmap-refusal excerpt and the internal note for every shortlisted candidate, not only outliers, since those two artefacts carry most of the decision. Three things need a person. A candidate may pick a defensible account other than the intended one — the recently renewed red account is a reasonable choice if their reasoning is about the next renewal cycle rather than this one — and the anchors under-reward it. A sponsor who discloses the competitor demo early makes the rest of the call easier in a way the score does not show, so the reviewer checks when disclosure happened. And the internal note is where a reviewer should apply their own knowledge of the deploying company's culture, because the right level of directness in that note is genuinely company-specific and the anchors describe a default, not a universal. Overrides are written with a reason. Nobody is rejected on the composite alone.

What this does not measure

This design does not score warmth, and that is the point of it. The role page identifies warmth as the trait the existing interview selects for and the trait behind the dominant failure mode in the seat, so no criterion above rewards rapport, likeability, or how pleasant the call was. It equally does not score accent, dialect, fluency, pace or manner of speech, directly or through any composite descriptor. "Executive presence" and its neighbours do not appear and must not be added by a deployer, because they are the standard route by which speech patterns and class markers are scored under a professional name. Each anchor points at an event in a transcript — was the evidence put in front of the sponsor, was the request refused in terms, was a date named. The data task requires arithmetic and attention, not spreadsheet tooling or analytical training, so it does not proxy for education. The written note is judged on whether the two named readers could act on it; idiom and phrasing are not scored and non-native writing is not a deduction. Product knowledge is supplied in the brief so that prior experience of a particular vendor's stack confers no advantage. What forty minutes cannot see is a renewal year. This role's performance is a portfolio worked over twelve months, and the behaviour that matters most — raising a signal at month two that nobody asked about — is defined partly by its timing, which no session reproduces. What is observed here is the quality of one triage decision, one decisive conversation, and one internal escalation. It gives no evidence about consistency across dozens of accounts, about whether the candidate keeps looking at data in a quiet month, about relationship durability, or about whether they sustain internal candour after the first time it costs them something. It should rank a shortlist on evidence-led triage and candour, and should be back-tested against gross retention and the lead time between a risk being raised and a renewal date, at two and four quarters. The live-call format may disadvantage candidates with speech, hearing or anxiety-related disabilities. A text-based run of the same conversation, with the same hidden facts and release conditions, must be offered on request, scored on identical anchors, and left unmarked on the result. Extra time on the data task and the note must be available without disclosure.

The first design decision for this role is not about tasks. It is that the scenario has to state which version of the job is being assessed, because the role page is explicit that two companies advertising for a Customer Success Manager are frequently hiring two different jobs — a support function with named accounts, or a commercial owner of revenue. This design assesses the second. The brief says so in the candidate's instructions, which has a useful side effect: buyers who cannot say which version they are hiring discover it while configuring the assessment rather than a quarter after the hire.

The strongest observable signal in this seat is the triage decision, and it is worth being precise about why. Most of what a CSM does is invisible to any observer, including their manager, because it consists of deciding what not to do. A portfolio of several dozen accounts cannot be worked evenly; the value of the role is concentrated in the handful of unscheduled calls the CSM chose to make. That choice is made on a usage export, a ticket history and a renewal date, and it is fully reproducible in twelve minutes. It is also the one part of the job with a right answer that a hiring manager can check. Everything else in this design is downstream of it: the live call runs against whichever account the candidate picked, so a candidate who triaged badly spends fifteen minutes talking to the wrong person, which is exactly what the failure costs in the job.

The export is built with two decoys because the characteristic error is not laziness, it is responsiveness. The loudest account generates the most tickets and the most email, and calling it feels like work; its tickets are about a feature its users are using constantly, which is a health signal rather than a risk signal. The red-scored account renewed a month ago and cannot churn this quarter. The account that matters is flat, and flat is what a busy CSM skips. Reading the composition of that flat number — one team up, one team gone — is the single behaviour that separates the top quartile from the median on evidence rather than on manner.

The call then tests whether the reading survives contact. The sponsor says everything is fine, which is what a quietly unhappy customer says to someone they like, and the design gives the candidate an unambiguous piece of contradicting evidence to use. The cheap path is to accept the reassurance and have a warm conversation; the role page names that exact conversation as the reason a monthly check-in cadence consumes a week and surfaces nothing. The expensive path is to put the evidence on the table without accusation and then tolerate a silence.

Midway, the roadmap request arrives, and this is where the design spends a fifth of its weight on ninety seconds. Vagueness here — the roadmap conversation, let me raise it with product — is the most common single behaviour in the seat and converts a small disappointment now into a large betrayal at renewal. A flat no is better and still not good. What the top anchor describes is a refusal followed by the recovery of the underlying need, and it is observable in a way almost nothing else in customer success is.

The internal note closes the loop on the third separator the role page names: the willingness to raise a risk to the person who sold the account. The brief deliberately tells the candidate that this account executive reacted badly to a risk flag last quarter, because the whole question is whether the risk survives knowing that. A note that softens into an update, or that turns into a complaint about how the account was sold, both fail for different reasons, and both are readable in a hundred words.

Sources

Every figure on this page is traceable. Where a claim could not be sourced it is stated qualitatively instead.

1. SaaS Capital, What Is a Good Retention Rate for a Private SaaS Company (2025 survey of private B2B SaaS companies above $1M ARR): $25,000-$50,000 ACV band median NRR 102 percent, top quartile 111 percent, bottom quartile 97 percent; median NRR rises with ACV band, https://www.saas-capital.com/blog-posts/what-is-a-good-retention-rate-for-a-private-saas-company/
2. US Bureau of Labor Statistics, 2018 Standard Occupational Classification Definitions (contains no 'customer success' or 'account manager' occupation), https://www.bls.gov/soc/2018/soc_2018_definitions.pdf