Open a support inbox that's in trouble and you'll usually find a team working hard, and working top to bottom. Oldest first, or newest first, depending on the day. Both are the same decision: letting the order things arrived decide the order they get answered. It's the most reasonable-sounding rule in support, and it is quietly building your backlog.
Adding people to an inbox worked that way buys you a few weeks. The sorting is the thing.
First in, first out is a warehouse rule
FIFO makes sense when the items are interchangeable. A pallet is a pallet. Whichever one you move first, you've moved the same amount of work.
Support messages are not interchangeable. In any hour you'll have a customer who can't complete a checkout, a thank-you, a duplicate of a thread you answered yesterday, and something only your finance lead can decide. Arrival order treats all four as the same object — and the queue charges you for it. The checkout waits behind the thank-you. By the time someone reaches it, a two-minute fix has become a refund, a chargeback, or a review.
That message didn't get more expensive because your team was slow. It got more expensive because it was in the wrong place in the line.
Every hour a blocked customer spends behind a thank-you is an hour you end up paying for twice.
What triage actually sorts on
Triage isn't reading everything twice. It's one pass, asking a short list of questions that a new agent can apply on their second day:
- Is the customer blocked? They've paid, or they're trying to. Nothing else in the inbox outranks this.
- Is it a decision rather than an answer? Then it isn't yours to hold. Summarise the thread, route it to the person who can decide, and get it out of the queue.
- Has it been asked before? Duplicates and follow-ups belong on the original thread, not as new work.
- Does it expire? A delivery window, a return period, an event date. Anything with a deadline outranks anything without one.
Four questions. If they aren't written down somewhere an agent can point at, every agent invents their own answers — and the order your queue gets worked in changes depending on who opened it that morning.
A response window is a promise about the worst case
Teams tend to report an average response time, and averages hide exactly the thing you care about. A four-hour average is what you get when most messages are answered in twenty minutes and the hard ones sit for two days. Nobody experiences your average. They experience their own ticket.
A window is a different promise: this kind of message gets answered inside this long, every time. It's harder to hold, and it's the only version a customer can feel. It also makes triage measurable — if the blocked-customer window is slipping, the sorting is wrong, not the staffing.
Rules worth settling before anyone works the queue
None of this needs a forty-page runbook. It needs decisions, written down, that an agent can act on without asking:
- The order. Which of the four questions above outranks the others when two apply at once.
- The windows. One response time per category, not one for the inbox. The blocked customer and the thank-you should not share a target.
- The escalation owner. A person and a response time, for each kind of decision. "The team" is where threads go to die.
- The tags. A short, closed list. A tag nobody agreed on is a tag nobody searches.
- The end-of-day rule. What is allowed to still be open at close, and what has to be cleared or escalated before anyone logs off.
That last one is the one people skip, and it's the one that compounds.
Yesterday's inbox is the honest number
Average response time will flatter you. The measure that tells you whether the sorting works is what's still open when the day ends — because that's what your team inherits tomorrow morning, on top of tomorrow's volume.
A queue that clears daily stays a queue. A queue that carries even a small remainder is a backlog with a slow fuse: the remainder is worked last, so it ages, and aged threads take longer to answer because the context is cold and the customer has followed up twice. The cost of a message goes up the longer it sits, which means the cheapest inbox is the one nobody inherits.
If your queue is carrying, don't start by counting heads. Take one day's messages and sort them by the four questions before you answer any of them. You'll know by lunchtime whether you have a volume problem or an order problem. If it's the order, the useful thing about that is it was never anyone's decision — which means it's still available to make.
