A VIP host has perhaps thirty useful conversations in a day and a list of hundreds of accounts. Which thirty is the entire job. The obvious ordering — biggest accounts first — feels correct every morning and is wrong expensively: it spends the day on the largest account whether or not anything is happening to it, while the account that quietly stopped playing last week sits four screens down because it was never the biggest.
Value at risk is an account's projected value multiplied by its probability of churning. A large account in perfect health scores low, because there is nothing to do about it; a medium account that is slipping scores high, because that is where the host's hour changes the outcome.
It has to be a product rather than a weighted sum. A sum lets a sufficiently large account outrank everything at any risk level, which reproduces exactly the failure the metric exists to fix. With a product, an account at zero risk contributes nothing to the queue no matter its size — correctly, since a healthy account is not work.
"Who should I call today" and "who should be on a host desk at all" are not the same question. Candidacy for the desk is ranked by worth — a large, healthy, growing account belongs to a host precisely because it is large and healthy, and it will score near zero on value at risk. Using the queue's ordering to decide who gets hosted would leave the best accounts unmanaged on the grounds that nothing is currently wrong with them. So there are two lists, and every row on both carries a reason: a host told to call somebody and not why will invent one.
Value at risk is an account's projected value multiplied by its probability of churning. It answers what a host stands to lose today rather than what an account is worth in general, which is why it puts a medium-sized account that is slipping above a large one in perfect health.
Because the ranking barely changes from day to day, so the host returns to the same largest accounts whether or not anything has happened to them, while accounts where a call would still change the outcome sit below the fold. A value-sorted list tells you who matters; it does not tell you what to do today.
Ludora's default cap is seventy-five, on the basis that it is roughly the number of accounts one person can hold and still know who they are speaking to. The exact figure depends on the market and the segment; what matters more is that the cap is enforced when accounts are assigned, because the pressure is always to add one more and the resulting shallowness is invisible until the whole desk is thin.
Less often than an unconstrained queue would suggest. Without a cooldown, the highest-scoring accounts surface every single day and get called until they leave. A cooldown of around a week keeps a recently contacted account out of the queue regardless of its score, and counting those accounts separately lets a host tell a genuinely quiet caseload from one they have just worked through.
A tier ladder is a public, rule-based statement of status that players can see and work towards. A host desk is an internal allocation of scarce human attention. They overlap but do not match: a top-tier player may not need a host, and a mid-ladder player on a strong trajectory may warrant one.