To review every call in a casino call centre, you score recordings against your own script and criteria, and give the supervisor the exact fragment where an operator lost the deposit. A QA team listening by hand covers a small share of calls, so most coaching is guesswork. A scoring system extends the review to all of them, including calls in languages the supervisor does not speak. It should not replace the supervisor. This article describes what to look for and how to keep it honest.
What is wrong with manual call review?
Coverage. A QA team can only listen to so many calls, so most of what happens on the line is never heard by anyone responsible for quality. Coaching then depends on the few calls that were sampled, which may or may not be typical.
The second problem is language. On multi-language desks a supervisor cannot tell what went wrong in a language they do not speak. The operator may be doing something good or something harmful, and nobody who manages them can say which.
What does call scoring do?
It listens to every recording, evaluates it against your script and criteria, and produces a score with evidence. In Sellrise, call analytics scores each call per language against your script, shows the operator's weak spots with the exact recording fragment, and leaves the final verdict to a supervisor who confirms it.
That last part matters. A score without a fragment is an opinion. A score with the moment attached, where the operator skipped the offer or answered an objection poorly, is something a supervisor can verify in thirty seconds and use in a coaching conversation.
What should the criteria be?
Yours. The script and the checklist are your business, and a tool should apply them, not invent them. A sensible set is short and tied to outcomes:
- Did the operator identify the reason for the call?
- Was the right offer from your catalogue presented?
- Were objections handled according to your guidance?
- Were the compliance statements made, including anything required by your licence?
- How did the call end: deposit, promise of a deposit, or nothing?
Keep the list short enough that two supervisors would score a call in the same way. If they would not, the criteria are too vague for a machine either.
A practical way to start is to take a small set of calls your supervisors already scored by hand, run them through the same criteria, and compare. Where the two disagree, the criterion is usually the problem, not the people. Fix the wording, run it again, and only then widen the scope to every call. It is slower than switching everything on at once, and it is the reason the first scores your operators see will be ones they can accept.
Who gets called in the first place?
Scoring improves what happens on a call. It does not decide who should be called. In Sellrise the operator queue comes from the same forecasts as the rest of the system: players whose value justifies the time, with the reason and the offer shown. The queue also shows who was never dialled. For why calls should be reserved for valuable players, see how to choose messaging channels for casino retention.
A well-run call on the wrong player is still a wasted call. A forecast and a call review work well as a pair.
How do you keep a human in charge?
- Supervisors confirm or overrule every verdict that leads to action about a person.
- Operators can see the fragment and the criterion behind their score.
- Disagreements between a supervisor and the score feed back into the criteria.
- Scores are used for coaching first, and for performance management only after a period of trust.
I would be careful about using automatic scores to penalise people in the first months. A system that has not been checked against your supervisors' judgement will make mistakes, and operators will notice.
What about recordings and privacy?
Recordings are sensitive. In Sellrise the data of each client sits in a separate database on our side, not mixed with other clients, and retention and deletion terms are agreed before the start. Whatever vendor you consider, ask where recordings are stored, for how long, who can listen and how deletion works. Check also what your local rules say about call recording and notice to the caller.
How do you tell whether it helped?
As with any change, compare. Hold back a group of operators, or a period of calls, that does not receive the new coaching, and compare the outcome you care about: deposits per call, or average turnover per player contacted. Do not judge by the score itself, since an operator can learn to please the scoring and not the customer. The test design is in how to run a control group test.
Is it a separate product?
Call analytics is a module in Sellrise, priced from $500, and it can run alongside the other modules or on its own. The page on call analytics has the overview, and the pricing logic is explained in how casino CRM is priced. If you want to see it on your recordings, a pilot is the way to start.