The question that arrives for the fortieth time
"How do I reset my portal password?" "Where is my order?" "Can I change the delivery address after dispatch?" Every support team has a short list of questions that make up a large share of its day. The agents know the answers by heart. That is exactly the problem: the answers live in their heads, in old chat threads and in a colleague's saved replies, and every customer who asks is served by a person retyping something that has been said hundreds of times.
This article is about the labour cost of that pattern. It is not about the quality of the answers, which are usually fine. It is about the fact that the business pays a trained person to deliver a known answer by hand when the same answer could be written down once and found in seconds, by the customer or by the next agent.
Why the same questions keep landing on people
- The knowledge is tacit. Experienced agents never needed documentation, so none was written. New agents learn by asking, which pulls the experienced ones off the queue.
- Documentation exists but cannot be found. A shared drive with forty similarly named files is not a knowledge base. If search fails, people ask a human instead.
- Articles go stale. A price changes, a courier is replaced, a policy is updated, and nobody owns the article. Agents learn to distrust the written version and ask a colleague.
- Customers cannot see the answers. Help content, where it exists, sits behind a login or is buried on the website, so customers pick up the phone.
- Nobody reviews what is being asked. Tickets are closed, not studied, so the top ten questions never get a permanent fix.
Atlassian describes a knowledge base as a centralised library of articles that helps people find answers without opening a ticket, and notes that information often gets stuck in email, tickets and the heads of individual service desk agents. The description is aimed at IT service desks, but the diagnosis fits any support function.
Symptoms to look for
- agents keeping personal notes or "cheat sheets" because the official ones are not trusted;
- senior staff interrupted all day by juniors asking how to handle routine cases;
- the same email text pasted, with small edits, many times a week;
- different agents giving different answers to the same question;
- long onboarding times for new support hires;
- customers saying "I already asked this last week";
- a spike in contacts after every product, price or process change.
A related internal version of this problem is covered in staff answering the same internal questions because knowledge is undocumented. Here the questions come from customers, but the cause and the cure are close relatives.
How the pain turns into money
- Handling time. Each repeat question takes agent minutes that add up across the month. Nothing is produced by that time except a reply that a published answer could have supplied.
- Interrupted experts. When a junior asks a senior, two salaries are spent on one routine answer, and the senior loses the thread of what they were doing.
- Slow onboarding. If knowledge can only be absorbed by sitting beside someone, every new hire takes longer to become productive.
- Inconsistent answers. Different replies lead to follow-ups, escalations and occasional goodwill gestures that a single correct answer would have avoided.
- Waiting customers. Queues grow when capacity is used on repeat work. Customers abandon, repeat contact or leave.
- Key-person exposure. If the one person who understands a tricky process resigns, the answer leaves with them. The wider risk is described in how to move operational knowledge out of people's heads.
Atlassian's guide to knowledge management in ITSM says the practice reduces duplicate work, shortens onboarding and resolution times, and improves consistency across teams and channels. Those are exactly the cost lines above.
An illustrative cost model
These figures are illustrative assumptions for a hypothetical customer support team of six at a Cape Town services business. Use your own ticket data.
| Assumption | Value |
|---|---|
| Customer contacts a month | 1,800 |
| Share that are repeat questions with a standard answer | 45% |
| Average handling time for a repeat question | 9 minutes |
| Loaded cost of an agent hour | R260 |
| Senior time interrupted by colleagues' questions | 4 hours a week at R420 |
| Working weeks a year | 46 |
- Repeat contacts: 1,800 × 45% = 810 a month
- Agent time: 810 × 9 minutes ÷ 60 = 121.5 hours a month
- Annual cost: 121.5 × R260 × 12 = R379,080
- Senior interruptions: 4 × R420 × 46 = R77,280
- Illustrative total: about R456,000 a year
Not every repeat question can be removed. Some customers prefer to talk to a person, and some cases need judgement. But if even a third of that volume could be answered from published content, the saving would pay for a considerable amount of knowledge-writing time.
Why it is tolerated
Answering questions feels like the job. The queue is always full, so there is never a quiet week to write articles. The cost is spread across many small replies, and no report labels them "repeat". The individuals who hold the knowledge are valued precisely because they know things, which makes documenting it feel like making oneself redundant. And the savings from a better knowledge base are indirect, so budget holders rarely see a line item to approve.
When it becomes urgent
- ticket volume grows faster than headcount can be added;
- a key agent resigns and the team discovers how much they carried;
- a product launch or price change causes a wave of identical questions;
- customers begin posting the same complaint publicly;
- a new service channel (chat, WhatsApp) multiplies the places where answers must be kept consistent.
What better looks like
A small, well-kept body of answers beats a large neglected one. The aim is a loop in which every solved question can become a findable answer.
- Find the real top ten. Tag the last month of tickets by question type. The top ten usually account for a large share of volume.
- Write the answer once, in plain language. Short steps, a screenshot where useful, a named owner and a review date.
- Make it findable by the agent first. If agents cannot find an article in under thirty seconds, customers certainly cannot. Test with real phrases customers use.
- Work knowledge into the ticket process. Atlassian's description of knowledge-centered service is a good model: agents consult the knowledge base first and update or create an article as part of resolving the issue, rather than treating documentation as a separate project.
- Publish what is safe to publish. Put the most common answers where customers can see them: on the website, in the order confirmation email, in the WhatsApp auto-reply.
- Assign owners and review dates. An article without an owner will be wrong within a year.
- Close the loop on causes. If the same question appears every week, ask whether the process, the invoice or the delivery notice is the real fault. The related issue of unclear ownership is covered in support requests without clear owners.
Staff who must search several tools to find an answer add their own delay, as explained in customers waiting while staff search several systems.
Practical next actions
- Export one month of tickets or emails and group them by question type.
- Pick the five most frequent and write a short answer for each this week.
- Ask the longest-serving agent to review each answer for accuracy.
- Store the answers in one searchable place and link it from the agent's workspace.
- Add a rule that anyone who answers a question for the third time must write or update the article.
- Re-count the repeat question share after a month.
Bottom line
When knowledge is undocumented or hard to reach, the business pays skilled people to be a search engine. The labour is small per question and large per year. Writing down the ten most common answers, keeping them current and making them easy to find costs far less than answering them again, and it makes the team less dependent on any one person's memory.
Sources and further reading
- Knowledge Base Guide: Examples, Templates & Best Practices (last verified 10 October 2026)
- Knowledge Management in ITSM: Process & Benefits - Atlassian (last verified 10 October 2026)
- What is KCS and Why Does it Matter? - Atlassian - Atlassian (last verified 10 October 2026)
Related Mellorca reading
- "Just Checking on My Order": The Labour Cost of Manual Progress Updates
- Fixing Avoidable Mistakes Instead of Serving Customers: The Cost of Rework
- Onboarding New Customers by Email: The Labour and Revenue Cost of a Slow Start