Customer support

What this shelf covers

Handling customer questions at volume: shared inboxes, ticket queues, assignment, service targets, self-service content and the reporting that shows whether any of it is improving.

Products range from a light layer over mail, to platforms covering several channels with automation and workforce planning attached.

Real-time conversations with visitors sit in live chat. Published help articles belong in docs and knowledge base.

When mail stops being enough

Small teams answer support from a shared mailbox, and it works until three things start happening.

Two people reply to the same message. Something falls through because nobody owned it. Nobody can say how many requests arrived last month or what they were about.

Any of those is the signal to move. What a help desk adds is ownership, status, history and measurement, and those are worth the extra weight only when the volume justifies them.

Channels, and how many you can actually staff

• Mail, still the majority of volume for most businesses.
• Chat, for questions that block someone in the moment.
• Phone, expensive to staff and unavoidable in some markets.
• Social and messaging apps, where customers increasingly start.
• Self-service, which handles the questions nobody should have to ask.

Adding a channel adds a promise. A messaging app connected and then answered twice a week is worse than not offering it, because the expectation is immediacy.

Look for one queue covering everything rather than separate tools per channel, and for a customer history that follows the person across all of them.

Automation that respects the customer

The useful automation in support is unglamorous: routing by topic or language, priority based on customer or keyword, saved replies, and status updates that happen without anybody remembering.

Answer bots work where documentation is good and repeat questions are common. They fail when the help content is thin, and they become genuinely damaging when the route to a human is hidden. Make handover obvious and immediate.

Service targets are worth configuring only if somebody watches them. A target nobody reviews is decoration, and a breach report nobody reads is worse.

Measuring the right thing

First response and resolution times are the standard measures, and they are easy to improve in ways that help nobody.

The number worth chasing is contacts that should not have existed. A question asked four hundred times a month is a product problem, a pricing page problem or a missing article, and fixing the cause removes the work permanently rather than answering it faster.

Satisfaction scores are useful as a trend and noisy per ticket. Read the comments rather than the average, since the sentence somebody wrote is more actionable than the number they clicked.

Self-service, which pays for itself

A help centre connected to the ticket queue does two jobs at once.

It answers people before they write, and it tells you what is missing through searches that returned nothing. Suggested articles inside the contact form reduce volume measurably when the content is genuinely good and annoy people when it is not.

Keep the writing close to the queue. The team answering questions knows which article is wrong, and the shortest path from noticing that to fixing it is what keeps a help centre alive.

What it costs

Per agent per month is near universal, with automation, service targets, reporting depth and portal branding setting the tiers.

Ask about light agent seats for people who reply occasionally, since paying a full licence for an engineer who answers two tickets a week is a common waste. Then check what happens at seasonal peaks, when temporary staff need access for six weeks. The surrounding pricing habits are described in what pricing pages hide.

Staffing, and what the software cannot fix

Help desk software organises work. It does not answer tickets, and the most common disappointment in this category is buying a platform to solve a staffing problem.

Work out the arithmetic first. Tickets per week, average handling time, and the hours your promised response time actually requires. That number decides whether the answer is a tool, a hire, or reducing the volume at source.

Reducing volume is the underrated option. A confusing pricing page, an unclear error message or a missing article can each generate hundreds of contacts a month, and fixing one removes that work permanently.

When you do buy, configure less than the product allows at first. Queues, targets and automation added before the team has a rhythm produce process that gets worked around rather than used.