What actually fits into one interview round, and what to leave for the next

Interview scripts are almost always overloaded, and almost never out of ambition. They are overloaded because the script gets written on a page, where eight questions take half a sheet and look reasonable, rather than against a clock, where eight questions force every answer under a minute.

The question that follows from that arithmetic is not how much the tool gives you. It is how a process gets divided among its rounds. The scope of each one is a design decision: you choose what this conversation answers and what you leave to the one after it. This issue does the arithmetic behind that decision, and it is worth redoing with your own recordings as soon as you have a few.

The real budget of a round

Duration is set in the script: the minimum is five minutes, and the maximum is agreed during each client’s implementation. The numbers below come from a short round of ten minutes, because that is the tightest case and the one that exposes a design error fastest; the same subtraction works at any other length.

Ten minutes of interview are not ten minutes of assessment. Four things have to come out first, and none of them is optional.

• Consent, and the explanation of what is recorded and why. It is not a formality and it cannot be trimmed until it stops being readable; the issue on consent covers how to do it without breaking the conversation.
• The framing: what this conversation is, how long it will take, and what happens next. Without it, the person spends their first answers working it out.
• The first minute, which goes to nerves and to adjusting to the format. It exists whether or not there is a human interviewer.
• The close: when they will hear something, and through which channel. Thirty seconds that save a week of uncertainty.

Once you subtract all that, a ten-minute session leaves around seven minutes of assessable conversation. A fifteen-minute one leaves about eleven. That is the number to design against, not the one on the calendar.

What an item costs, measured in turns

A behavioural item is not a question, it is a sequence: the question, the answer, one or two follow-ups and their answers. Four to six turns, and none of them is optional if what you want is an answer containing the situation, the person’s own action and the outcome.

Telling a complete story takes between forty-five and ninety seconds when someone has it fresh and tells it well. It takes considerably longer when it has to be pulled out with follow-ups, which is the normal case. Added up, a properly explored item costs two to three minutes.

If your item resolves in forty seconds, check the question before congratulating yourself: it is probably not asking for a story. Questions that get answered quickly tend to be the ones that do not filter, which is the subject of the issue on questions that tell candidates apart.

Three competencies, four if you are lucky

Seven assessable minutes divided by two and a half per item comes to three. With someone who narrates well and a tightened script, four. That is the ceiling of this round, and it is set by something you cannot configure: how long a human being takes to recount something that happened to them. It is not the ceiling of the process, which carries as many rounds as it needs, nor of the tool. It follows from having decided that this particular conversation lasts what it lasts.

The interesting question, then, is not how many competencies to assess but which three belong here. The criterion is not importance: nearly every competency in a matrix is important, which is why it is there. The criterion is which ones separate the candidates who reach this round. A competency everybody meets equally well takes two minutes and returns a column of identical values.

The temptation to add a fourth “because it matters too” is constant, and it has to be resisted with a concrete argument: adding it does not add a competency, it takes half a minute away from the other three. Nothing is lost by keeping that fourth one out of here — it gets asked later, with enough room for the answer to arrive somewhere.

The trade-off, in numbers

The same seven assessable minutes, split four ways. The right-hand column is the one that tends to surprise.

There is a consequence worth saying out loud, because it produces unfair decisions. When a script is overloaded, nearly every candidate lands on low levels — not because of what they did at work, but because they had no time to tell it. The report appears to say something about people, and is in fact describing the format.

What moves to the next round

A short round answers a narrow question: is there evidence this person has done X? It does not answer whether they are the right person, and it is better not to ask it to try. What does not fit here does not leave the process — it changes rounds.

• Technical depth. A real case with its constraints needs time, and someone who knows how to press by instinct: that is material for a technical round, with more room per answer.
• Motivation and where the person is heading. Asked in passing, they return the rehearsed version; asked in a conversation that is not racing a clock, they stop being rehearsed.
• Fit with the team. It is not assessed from a transcript and should not be attempted: it is a conversation between people, and it belongs in the round where those people are already sitting.
• Negotiation. Terms, availability and expectations get settled once both sides are interested.

That split is not an editorial recommendation each team has to carry on its own. A role holds as many rounds as the process needs, in order, each with its name and its interviewer profile. A round orders the process, the invitation and the file; the questionnaire comes from the role, so splitting across rounds is a decision about what gets published and when. The same person goes through them without turning into three different candidates: their candidacy is one, and it accumulates whatever they answered in each. The round’s name is frozen into the interview already answered, so renaming “First round” to “Technical” in September does not rewrite the file of whoever answered it in March. The platform page lays out the whole path, and the cases by type of process show how the split usually ends up depending on what is being filled.

Who moves to the next round is decided by a person from the platform, or by the client’s system through the API. At that moment the link for that round is issued, and the team or the ATS gets it to the candidate. The cut is made by whoever answers for the hire, and it is recorded with their name and the time.

How to recover time without losing depth

There are four hidden minutes in almost every script, and none of them comes from shortening answers.

1. Take requirements out of the conversation. Licences, shifts, permits and availability are data points: they get settled in the application or in the system the company already runs, which is what syncing the role is for.
2. Do not ask again for what is already on file. If the ATS has the years of experience, asking for them again costs a minute and adds nothing. And if something was already asked in an earlier round, the answer is still there: the candidacy is the same one.
3. Put in writing whatever can be read beforehand. The notice about what is recorded and why reads better than it listens, and it leaves the conversation for what can only be conversed.
4. Cap the follow-ups at two per item. Without a cap, the first item eats the interview and the last one never gets explored.

What a single round cannot tell you

A round does not tell you whether the person will do the job well. It tells you whether there is evidence of three concrete behaviours, and where that evidence comes from. The genre promises to assess everything in one sitting; bounding the scope round by round is precisely what makes each assessment defensible, because every level rests on an answer that had room to exist. The rest is not discarded: it gets asked afterwards.

Which is why the record also has to say what was not asked in this round. A report listing three assessed competencies while staying silent about the other five invites a reader to mistake a silence for a result, when what is there is a question scheduled for later. What to keep from each interview so that stays clear is the subject of the issue on turn-level evidence.