Reduce repeat questions: finding the cause with a chatbot

The same question comes back every day

The support team answers the same questions every day: «how many days is delivery?», «can I pay on arrival?», «can I return it?». After a chatbot is set up the questions do not go down — the bot simply answers them now. That is a step forward, but it leaves the real question open: why do customers keep asking the same thing?

The cause of a repeated question is not always «it's not in the FAQ». Sometimes the answer exists but cannot be found; sometimes it exists but is unclear; sometimes the question itself is a symptom of a product or process problem. This article covers four causes of repeated questions, an intent map, answer design, what the chatbot can do here and measurement. The choice between an FAQ page and chat is a separate topic; here the subject is diagnosis.

Four causes of a repeated question

1. 1The information does not existThe answer is written nowhere — neither on the site nor in the knowledge base. The easiest cause to fix.
2. 2It exists but cannot be foundThe answer is in the FAQ's fourth section, in the middle of a long paragraph. The customer does not search; they write.
3. 3It exists but is unclear«Delivery 1–5 working days». The customer does not know whether their city is 1 or 5, so they ask anyway.
4. 4The question is a symptom of a process problemA rise in «where is my order?» often means delivery is running late. Logistics, not the FAQ, solves this one.

An intent map

For diagnosis, repeated questions are grouped by intent. A simple table is enough:

• Intent: what the customer actually wants to know — «delivery time»
• Variants: how customers ask it — «how many days to Ganja?», «will it come tomorrow?»
• Volume: how many times a week
• Channel: where it is asked most — Instagram, WhatsApp, the website
• Cause: which of the four causes
• Owner of the fix: knowledge base, website, product team, logistics

The map is built from two weeks of exported conversations. In Vexvon, the Q&A export comes as an Excel file — sorting it into intents takes an hour or two.

A fix for each cause

1. 1Missing → write itA new knowledge base entry, in the customer's words, with an owner.
2. 2Not found → move itBring the answer to where the question is asked: the product page, checkout, the order confirmation message. A chatbot helps most with this cause, because it brings the answer to the channel where the question was asked.
3. 3Unclear → make it preciseA list by city instead of «1–5 days». The bot can ask the customer's city and give a specific figure — but the figure has to be in the knowledge base.
4. 4Symptom of a process → fix the processTake the rise in the question, with numbers, to the team responsible. Here the bot is a measuring tool, not the solution.

Answer design: an answer that is given once

An answer that reduces repeats has three properties:

• Complete: it anticipates the next question. «Delivery is 1 day in Baku and 2–3 days in the regions; the courier calls before arriving» — «will the courier call?» is already answered.
• Specific: about the customer's case, not the general rule
• Short: one to three sentences; if a list is needed, no more than three points
• With a next step: «Would you like to check your order status?»

Add the same answer to the order confirmation message too. If the customer gets the answer before asking, the question never arises.

Example: diagnosing one intent

1. 1ObservationOver two weeks the intent «can I pay on arrival?» came in 180 times, mostly from Instagram.
2. 2CauseThe information is on the website, but not in the Instagram posts or profile. Cause: «not found».
3. 3FixOne sentence is added to the profile and under campaign posts; when the bot answers this question, it offers the order step in the same message.
4. 4Checking the resultThe intent's volume over the next two weeks is compared. If it has not fallen, the cause is something else.

Answering before the question arises

The best answer is one the customer never needs to ask for. Most top intents arise at a specific point in the customer journey, and the answer can be placed at that point.

1. 1Product pageSize, material, delivery time — where «in stock?» and «when?» arise.
2. 2CheckoutPayment methods, the delivery fee, the return rule — doubts that appear at the last step.
3. 3Order confirmation messageThe estimated date, that the courier will call, how to check the status — half of «where is my order?» is closed here.
4. 4Social media profileOne or two sentences on address, hours, payment and delivery — for the simplest questions arriving from Instagram.

What the chatbot does here

A chatbot has two roles with repeated questions, and the second is less well known.

1. 1It answersA repeated question gets an answer in seconds, on the customer's channel, at any hour. Agent time is freed.
2. 2It measuresEvery conversation is data about an intent. Which question, how often, on which channel, at what hour — a ready-made list for the product and process teams.

Drawing wider conclusions from conversations — objections, lost intents, product signals — is a topic of its own; flow by hour and channel appears in the analytics report.

A 30-minute weekly review

• The top 10 intents compared with last week
• A newly appearing intent — a campaign, a new product or a sign of a problem
• Questions the bot could not answer — the list of knowledge base fixes
• Intents that are process symptoms — a short note sent to their owner
• The result of last week's fixes

The head of support, the knowledge owner and, when needed, someone from product or logistics attend. Thirty minutes is enough on one condition: every decision gets an owner and a date.

Measurement

• Weekly volume of the top 10 intents — the trend
• The bot's full-answer rate for each intent
• The same customer asking the same intent again within 7 days
• The change in an intent's volume after a fix
• Intents that are symptoms of a process — these go to the relevant team in a separate report

Common mistakes

• Treating every repeated question as «a gap in the FAQ»
• Counting the bot's answer as success rather than the question falling
• Building the intent map once and never updating it
• Keeping process-symptom questions inside support instead of passing them to their owner

Limits

Some questions will never go down: every new customer wants to know the same things, and that is normal. The aim is not zero questions but less unnecessary repetition — the same customer asking the same thing, a question caused by an unclear answer. A chatbot does not fix a process problem; it makes it visible.

Conclusion: a repeated question is a diagnosis

Repeated questions are a support burden, but they are also information about the company: where information is missing, hard to find or unclear, and where a process is not working. A chatbot answers these questions and measures them at the same time. With an intent map and an owner for each cause, repetition falls, and the bot answers the rest in seconds.

Which questions an FAQ page answers better than chat is covered in chatbot vs FAQ page; answer ownership is in knowledge management. The rest of the section is in this category. To build your intent map together, ask for a demo.

Frequently asked questions

1. 1Does a chatbot reduce repeated questions?Not directly — it answers them. To reduce them, the cause has to be found and fixed; the bot supplies the measurement for that.
2. 2How often should the intent map be updated?A monthly review is enough; weekly during a campaign or a product launch.
3. 3Which intents should we start with?The five with the highest volume. They usually make up most of the flow.