When you are holding one chat, the question of what to do next answers itself. When you are holding five, it does not. The tab strip along the top of your console is an inventory, not a priority list: it tells you which conversations are open, in the order they arrived, and nothing at all about which one is quietly turning into a problem.
The attention ranking exists to answer the other question. It walks every open tab, pulls the current state out of the other cards on your right rail, rolls it into a single number, and sorts your chats by it. This guide explains what goes into that number, what the tiers mean, and the two places where the ranking will confidently mislead you if you do not know how it works.
The question a tab strip cannot answer
Every other card on the right rail answers a question about the chat you are looking at right now. Is the visitor getting frustrated? Have you promised anything? Are there open questions you have not come back to? Useful, all of it, but each one is scoped to a single conversation, which means the only way to compare five chats is to click into all five.
That is exactly the work agents were doing. With four or five concurrent conversations, the honest strategy most people fall back on is round-robin: answer whoever spoke last, keep the tabs moving, hope nothing is rotting in the background. It usually works, and when it fails it fails on the chat you were not looking at.
The attention card reads across tabs instead of down into one. It is the only surface in the console that compares conversations to each other.
What the score actually counts
The score is a plain sum. Nothing is normalised, nothing is averaged, and every input has a cap so that no single signal can run away with the ranking.
Waiting. The clock starts on the visitor’s last message that you have not answered yet, and stops the moment you reply. It is worth one point per thirty seconds, capped at six. Three minutes of silence saturates it, which means wait time alone can never push a chat past the middle tier.
Risk flags. An open high-severity flag — the legal, chargeback, escalation, repeat-contact and deadline categories — is worth eight points, capped at sixteen. Two of them saturate. Medium flags, covering churn, review threats and abuse, are worth three each, capped at nine.
Effort. The effort reading contributes five points at Painful, three at High, one at Moderate, and nothing below that.
Open items. Each open objection is worth two points to a cap of six. Each open knowledge gap is worth one to a cap of three. Each unanswered question is worth one to a cap of two. Each open commitment is worth one to a cap of three. Steps you have asked the visitor to take are worth half a point each to a cap of one, deliberately small, because those are nudges rather than obligations on you.
A hot lead with nowhere to go. If the buying-intent reading is Hot and there is no open commitment and no decision on the board, the chat picks up three points. That combination usually means a ready buyer waiting on someone to propose the next step.
The total is rounded to one decimal place, which is why you will occasionally see a score ending in .5 — that is the visitor-steps input, the only fractional one.
The four tiers and what they mean
The number is only there to produce a tier, and the tier is what drives both the colour of the row and the sort order.
- Hot, twelve points and up. It takes a real combination to get here: an open high-severity flag plus almost anything else, or two of them on their own.
- Urgent, seven to eleven. One high-severity flag, or a painful chat that has also been waiting.
- Attend, three to six. The ordinary busy chat. Some open items, some waiting, nothing alarming.
- Idle, zero to two. Nothing is pending on you.
Chats sort by tier first and by score within the tier. There is no tiebreak beyond that, so two chats on the same score hold whatever relative order the tab list gave them. Your currently selected chat is ranked on exactly the same terms as the rest — it is not pinned to the top and it does not float, because the point of the list is to tell you when the chat you are looking at is not the one that needs you.
Clicking a row switches you to that tab and leaves the card open, so you can work down the list without reopening it each time. On a keyboard, Ctrl+Shift+4 toggles it. On a Mac that chord is intercepted by the operating system for region screenshots before the browser ever sees it, so the shortcut silently does nothing there and you click the pill instead.
Acknowledging a risk flag changes the ranking
This is the behaviour most worth internalising, because it is the one that will move a chat when you did not expect it to.
The attention score counts open risk flags. The risk card itself has three live states: open, acknowledged, and resolved. Acknowledging a flag means you have seen it and you are dealing with it — it is emphatically not the same as resolving it, and the card keeps counting it as unresolved, as does the wrap-up check that stops you closing a chat with an unresolved high-severity flag on it.
The attention ranking does not make that distinction. The moment you acknowledge a high-severity flag, eight points leave the score. A chat sitting at thirteen in the Hot tier can drop to five and the Attend tier on a single click, without anything about the underlying situation having changed.
In practice this is usually harmless, because you acknowledge a flag at the moment you start working on it and you are already in that tab. It becomes a trap when you acknowledge several flags in a sweep to clear the visual noise, then come back twenty minutes later and read the ranking as though it were still tracking them. It is not. Treat acknowledgement as a statement that you have taken personal ownership of that flag, because as far as the ranking is concerned, you have.
What the ranking cannot see
The attention card owns almost none of its own data. Apart from the wait clock, which it measures directly, every input is read from another card on the rail, and every one of those reads is written defensively: if a card is not loaded, or does not answer, the read fails quietly and that signal contributes zero.
That is the right engineering choice — one broken card should not take down the ranking for every chat you have open. It has a consequence worth knowing. A missing input does not show up as a gap or a warning. It shows up as a lower score, which is indistinguishable from a calmer conversation.
So the ranking is a floor, not a verdict. A chat scoring nine is at least a nine. A chat scoring one is not necessarily fine; it may simply be a conversation whose problems live in signals this card could not read.
The other thing it cannot see is anything you know and have not written down. A visitor who told you on the phone yesterday that they are about to churn contributes nothing to the score, because nothing in the console knows about it. The ranking reads the conversation, not the account.
How to use it without outsourcing your judgment
Use it as an interrupt, not as a work queue. The failure mode with any ranked list is that you start at the top and work down mechanically, which is just round-robin with extra steps and a worse sense of what is going on.
The useful habit is narrower: before you send a reply in the chat you are already in, glance at the pill. If nothing above your current chat is in a higher tier, carry on. If something is, decide deliberately whether to switch. Most of the time you will finish your sentence first, and that is correct — abandoning a half-written reply to chase a ranking is how you end up with two neglected chats instead of one.
It also gives you a defensible answer to a question that is otherwise pure vibes: which of these should I drop? When you are over capacity and something has to give, the tier tells you which conversation can absorb five more minutes of silence. That is a much better basis than whoever happened to type most recently.
And when the list is telling you that three chats are Urgent at once, it is not really telling you about the chats. It is telling you that you are past your concurrency limit, which is a staffing observation rather than a triage one.
How MyLiveChat fits
The attention ranking is part of the agent console, alongside the other right-rail cards it reads from. There is nothing to switch on and nothing to configure: it appears once you have chats open, and its inputs are whatever other cards are running.
It stores nothing on our servers and nothing about the ranking leaves your browser. The score is recomputed roughly once a second from live state, which is why the list reorders while you watch it — the wait clock is always advancing on somebody.
If you are routinely holding enough concurrent chats that you need a ranking to keep track of them, the ranking is worth having, but it is also worth reading as a signal in its own right. A console that is permanently showing two or three Urgent rows is describing a capacity problem that no amount of better triage will fix.