chatleadr Docs
Open Chatleadr

How to Answer Product Questions From Your Catalogue

A product catalogue is written for a page layout rather than for a reader, which is not the form a bot answers well from. This guide covers restructuring products for the knowledge base, and where to use a workflow instead.

Time: an afternoon for the knowledge base, longer if you build the live lookup. You need: Starter or above.

1. Decide Which Questions Belong in the Knowledge Base#

Two kinds of product question, and they need different answers.

QuestionAnswer with
"What sizes does the patio set come in?"Knowledge base
"Is it weatherproof?"Knowledge base
"How much is the 6-seater?"Knowledge base, if prices are stable
"Do you have it in stock?"A workflow with an API call
"Where is my order?"A workflow with an API call

Anything that changes hourly does not belong in a document: a knowledge base is compiled, so a stock level written into it will be wrong.

2. Write Product Documents the Bot Can Read#

Write one document per product family rather than one catalogue file. For each product, cover the fields customers ask about:

text
6-Seater Patio Set (PW-6000)

Price: $749 including VAT
Seats: 6
Table: 1.8m x 0.9m, powder-coated aluminium frame, tempered glass top
Chairs: 6 stackable, weatherproof rattan weave, removable cushions
Colours: charcoal, stone, olive
Weatherproof: yes, UV-stable. Cushions should be stored indoors in winter.
Assembly: required, roughly 45 minutes, tools included
Delivery: 5-7 working days nationally
Warranty: 2 years on the frame, 1 year on cushions

Three things make that answerable:

  • Labelled fields. A question about warranty finds a line about warranty.
  • The product code. Customers quote codes, and the assistant can match them.
  • Plain units. "1.8m x 0.9m" answers "will it fit on my patio?".

3. Split the Catalogue to Fit the Document Limit#

Each plan caps the knowledge documents one bot may hold, which decides how far the catalogue can be split. See plans and limits.

  • Under 30 products. Group them into as many category documents as your plan allows, largest category first.
  • Hundreds of products. Keep the catalogue out of the knowledge base. Load the buying guide, the specification vocabulary and the policies, and answer specifics with an API lookup, covered in step 5.

4. Add the Questions the Catalogue Does Not Answer#

Customers also ask things no specification sheet covers:

  • "Will this fit through a standard door?"
  • "Can I leave it outside all year in Cape Town?"
  • "What is the difference between the 4-seater and the 6-seater?"

Write those as a short questions document, in the words customers use.

5. Build a Workflow for Anything Live#

For stock, pricing that changes, or order status, build a workflow instead.

NodeConfiguration
TriggerAlready on the canvas.
API RequestGET your stock or catalogue endpoint, with the product code as a parameter.
DecisionDid the API return a result?
Message (true)The live answer.
Message (false)Say the figure could not be checked, and offer to have someone confirm.

The description has to be tight, or this workflow starts catching every product question:

text
Checks live stock for a specific product using its product code or name.

USE WHEN the visitor asks whether a specific item is in stock, available now,
or how many are left.
DO NOT USE for questions about specifications, dimensions, prices or delivery
times. Answer those from the knowledge base.

6. Connect the Product Pages to the Chat#

On a product page the visitor already has the context, and a button can carry it into the conversation with the product name already in the question. The copyable button, and the one-line guard that waits for the widget to load, are on web controls.

Word the message the way a visitor would ask it. It goes through normal workflow matching, so "I have a question about the 6-Seater Patio Set" reaches the stock workflow and "product info" gives it less to match on.

7. Test With Real Phrasing#

Customers rarely use product names. Test the phrasings they type:

  • "whats the biggest patio set you have"
  • "will the 6 seater fit through a normal door"
  • "is the grey one available"
  • "how long till i get it if im in bloem"

Each one either answers correctly from a document, triggers the stock workflow, or reveals a gap to write into the knowledge base.

Common Questions#

Can Chatleadr Import My Product Feed Automatically?#

Not as a feed. Point the scraper at your product pages, or upload exported documents.

How Many Products Can a Chatbot Handle?#

As many as fit your plan's document allowance once grouped into category documents, largest category first. See plans and limits. Beyond that, keep the buying guidance in the knowledge base and answer specifics with an API lookup.

Why Does the Bot Give the Wrong Price?#

Two documents disagree, or one is a scraped page that has since changed. Prices belong in exactly one place, and if they change often they belong in an API call rather than a document. See my bot gives the wrong answer.

Should I Put Stock Levels in the Knowledge Base?#

No. The knowledge base is compiled, so a stock number written into it is a snapshot that goes stale immediately. Use an API Request node, which reads the number at the moment the visitor asks.

Can the Bot Recommend a Product?#

Yes, from what the knowledge base contains. Write a comparison document stating what each product suits and how the options differ, because the assistant can only reason with the distinctions you have written down.