Individual Project

Conversational UX, Chatbot Dialog


Mar.'20 - May.'20


Botsociety, Botmock, Mural, Lucidchart


Rebecca Evanhoe

A chatbot helps remind people of key steps and solve some major types of problems in furniture assembly.


Understand the problems and challenges

How many people prefer to assemble IKEA furniture on their own?

How people rate the user experience of IKEA guide books?

What people most care about when they are looking for assistance?


Goal: to get an overall knowledge of my product's potential scope and design opportunities.

29 Responses


  • 76% of participants assemble the IKEA furniture by themselves.
  • 56% of participants give a rating 4 (very hard) for the difficulty of assembling furniture.
  • 88% of participants most care about images in guide books.
  • The survey shows that most people are still not satisfied with IKEA guide books and have met some difficulties in the past assembly experience.


    Goal: to learn more details about IKEA guide book’s drawbacks and improvement possibilities.

    1 in person + 2 remote


    User behavior in furniture assembly
  • All users will browse through the entire guide books first to get a general sense of the process, and then follow up each step carefully.
  • Two users think the whole assembly process will be more efficient if they can remember multiple steps at once.

  • Pain points of IKEA guide books

  • Complicated steps are hard to remember.
  • No written instructions. Interpreting the diagrams can be difficult.
  • Sometimes gadgets in guide books are not distinguishable.
  • Mirror problem: hard to specify the orientation of certain parts.

  • Expectations toward an assistant

  • Including videos or a 3D animated process.
  • Tips on how to propose things up when doing by self.
  • Knowing which angle should be working from.
  • Having concise and easy to understand wording guidance.

    This chatbot is a friendly and helpful furniture assembly assistant that provides efficient extra instruction besides IKEA guide books. It has a low level of personification, which could serve as a more intuitive tool using conversational user experience design compared with image-only interaction.


  • Efficient: make the furniture assembly easier and help users save time.
  • Accurate: give the correct and understandable instruction for each step.
  • Always available: users can get access to the chatbot whenever they have problems.
  • Patient: the chatbot should be detailed, not to skip too much process.
  • Direct: provide clear guidance instead of wordy sentences.
  • Low Personification

    Efficiency is the fundamental feature of the chatbot. This IKEA assistant should give users information quickly and correctly. So a none-human-like character will be appropriate.

    Power Dynamics

    The chatbot should be activated all the time when users assemble their furniture. It should always monitor the step that the user is currently doing and give prompt, useful, and appropriate feedback.


  • Serious: give users a sense of professional and reliable.
  • Casual: the chatbot is used in home, which is the personal space.
  • Respectful: fully understand what the user needs.
  • Matter-of-fact: factual information matters most for this chatbot.
  • Character Traits

  • Helpful: ask users whether they have questions proactively in the confusing steps.
  • Encouraging: give users clues and inspire them to continue even if the process is dull.
  • Tolerant: let users feel it is always here no matter how many problems they meet.
  • Omnipotent: should have all-round skills and reply informatively.

    Slots, Intents, and Utterances

    The stickers in green color were what I added after the usability testing.



    After brainstorming some sample scrips, I began to write dialogs for the chatbot. Due to the time limitation, I focused the sample conversations on two features:

    Help understand the image and find out the correct screws

    Help check the direction of the board in the previous step


    Click here to interact with the prototype


    2 synchronously interview + 3 unsynchronously questionnaire


    Things that still lack in current design

  • It would be better to have more customized procedures instead of always choosing from different tabs.
  • Since the instruction pictures are printed in the book, if users are stuck to the step, seeing the same image isn't going to help.
  • "visually videos would be better than back and forth text."

    Pain points of the current design

  • Sometimes when there are multiple elements, it's hard to tell which ones are for the pictures.
  • There are some steps where users want to see what it looks like with the physical product vs. pictures on a screen.
  • Recommendations

  • Include zoomed-in or real photos of which elements users are supposed to use.
  • The user experience process could be: getting the product, scanning QR code, launching step by step video of a real furniture.
  • Provide access for users to tap to skip steps and make sure steps are always aligned with the manual.
  • Videos, 3D models, or real products are not existing in the IKEA assembly guide system, but in the future, if IKEA provides these features for furniture, this chatbot will also add that.



    Click here to view the new prototype

    Flow Diagram

    The original flow was to let users type which step they are doing, however, typing words is difficult when their hands are occupied with furniture elements. So I decided to present all steps individually, and if users have no problem, they can simply tap the button to skip this step.

    For the checking intent, I planned to add a feature that users can take a photo of their current situation and upload to the chatbot, and then the back-end recognition system can automatically analyze whether it is right or wrong.

    View the flow diagram again


    From my perspective, designing more customized paths for chatbot will make the interaction model more complicated with more varieties. Though it will be hard to write utterances for all intents, what conversational developers can do is to apply machine learning or other algorithms to train the chatbox to recognize keywords more accurately and thus learn what people say.