By the Onsites AI team · Last updated · 5-minute read
The queue is honest for exactly one week. Then it fills with everything — billing questions next to feature requests next to a shipment complaint — and the only way to find anything is scrolling. Tags, priorities and views are the three tools that stop this, and they cost nothing but discipline. Done well, you get reporting that tells you what to fix, a queue that routes itself, and a triage ritual that takes five minutes. Done badly, you get a second layer of chaos. This guide is the difference.
The failure pattern is universal: someone creates "urgent", "important", "question" and "other", and within a month there are thirty tags, four spellings of "billing", and nobody trusts the reports. Tags are worth exactly as much as the discipline behind them, so constrain yourself from day one.
Five to eight tags, named for what customers write about — not what your org chart looks like. For an e-commerce desk: "order-status", "returns", "shipping", "product-question", "billing". For a services desk: "quote-request", "invoice", "active-project", "bug", "account". Each tag must pass two tests: a new hire could guess what belongs in it without asking, and a weekly report by tag would make someone actually do something. "Misc" fails both — if it becomes your biggest tag, the vocabulary is wrong, not the customers.
The habit that powers everything: tag when you close, not when the conversation arrives. Threads wander — a "delivery" conversation turns into an invoice dispute halfway through — and the close-time tag records what the conversation turned out to be, which is what you want to count. The volume-by-tag report is the single most honest picture of your support operation, and it costs five seconds per conversation. If your desk supports sub-tags or nested labels, do not, at this size: flat lists survive; trees decay.
Everything marked high priority is no priority. Priorities only work when the definition is written, short, and slightly cold-blooded:
Two rules keep the system honest. First, sales leads are not high priority — they get a "sales" or "quote-request" tag and a fast reply through your SLA, not the emergency lane; if everything hot is "high", escalation stops meaning anything. Second, priority is set by the impact rules above, never by the tone of the message — and any teammate may upgrade a conversation to high without asking permission, because the downside of over-escalating is minutes, while the downside of under-escalating is a churned customer nobody saw.
Views are saved filters, and they turn a flat list into a workspace. Start with four:
Resist the urge to build fifteen views on day one. Views are answers to questions you have actually asked twice; each one you add after the first four has to earn its place by being opened daily. A view nobody opens is a configuration nobody maintains — delete it and keep the workspace small. The auto-assignment guide shows how these views connect to routing rules once the team grows past three people.
Three conventions that keep the vocabulary healthy without anyone thinking about it. Singular, lowercase, no spaces ("refund", not "Refunds & Money Back"), because consistency is what makes reports sortable. One word for one thing — if two tags overlap ("invoice" and "billing"), merge them now; the distinction that matters will re-emerge on its own, usually as a view rather than a tag. And no tags for states: "waiting", "later" and "blocked" are what the status of a conversation is for; a tag that describes the queue's mood rather than the customer's topic pollutes every report it touches. When the vocabulary needs a change — and it will, around month two — rename the old tags rather than adding new ones beside them, so history stays countable in one column instead of two.
Fifteen minutes, morning and end of day. Morning: open "High priority, open" and clear it; open "Unclaimed" and claim what you will do today. End of day: everything still open has a note saying what it waits on and when it will next be touched; every closed conversation has its tag. That is the whole ceremony — no dashboards, no meetings — and it produces the two numbers that matter on any desk: unclaimed age and high-priority residence time. Both live in your queue views, and both are visible for free.
The compounding return arrives in week four, when the volume-by-tag report is large enough to act on. "Shipping" is your biggest tag and rising? That is a policy or carrier problem, not an inbox problem — no amount of reply speed fixes a warehouse that misses cut-off times. "Billing" is second? Write the billing explainer once and let saved replies carry it. "Quotes" cluster around one product line? That is a pricing-page rewrite. Tags turn support from a cost center that answers mail into the earliest-warning system in the company — the customers tell you what is breaking weeks before the sales numbers do.
Close the loop on your own data too. Free-tier cloud workspaces are not backed up server-side, so the weekly export is your history of record; and if customers ever demand data residency, the same desk runs self-hosted on your own servers. The tags travel with you either way — they live in the exports, they power the reports, and they are the reason a queue you set up in twenty minutes can still answer strategic questions a year later.
How many tags should a help desk have?
Five to eight, named for what customers write about — order status, returns, shipping, billing, product questions — with no nested categories. If "misc" becomes the largest tag, the vocabulary is wrong and should be rebuilt from a month of real conversations.
When should a ticket be tagged — at open or at close?
At close. Topics wander mid-thread, and the close-time tag records what the conversation turned out to be, which is what volume reports should count.
What does "high priority" mean?
The customer is blocked or money is wrong — a double charge, a locked account, an outage on a live channel. Sales leads and angry tone are not high priority; impact is, and it is defined in writing.
What is a "view" in a help desk?
A saved filter on the queue — for example "unclaimed, oldest first" or "high priority, open" — that turns a flat list into a working list. Start with four: Mine, Unclaimed, Waiting on customer, and High priority open.
Do tags help beyond organization?
Yes — tag volumes are the most honest picture of what customers struggle with, an early-warning system for product, policy and logistics problems, and the basis for every report the desk produces.