As a content design agency, it has felt like all aspects of our work are evolving at light speed these recent years. And this is especially true of our work in conversation design. While many organisations are only just beginning to explore the role of simple chatbot conversation design, others are embracing the advance of SaaS capabilities, shifting how we manage data to create truly incredible digital experiences.
For conversation designers, it’s important to understand what’s happening in other areas of tech that might impact the future of our work. AI is the big one, but so is how we use, collect, store, and manage data.
New ways of interacting with what we call ‘structured’ and ‘unstructured’ data are creating fresh opportunities in conversation design. So let’s explore what this means for organisations and content specialists alike.
What’s the difference between structured and unstructured data?
Structured data
Structured data is very organised, and usually arranged predictably in tables with rows and columns. It commonly refers to information managed within tables or spreadsheets, such as product information in e-commerce.
For example, this could be a product and its many attributes (e.g., ‘lego’, ‘toy’, ‘medieval’, ‘price-high’, ‘discount’). Because all information is categorised and neatly compartmentalised, it can be easily managed in the backend (e.g. editing or analysing data) as well as the front-end (through a user’s search experience).
Unstructured data
Unstructured data typically refers to longer forms of content and information, such as blogs, newsletters, social media posts, emails, etc.
Unstructured data can be more free-form in nature, which can make it harder to process or analyse with automated systems. It can go beyond text-based data to include images, audio and video, which have been traditionally hard to manage and analyse in the same way as structured data.
A great example of this for conversation designers is call centre recordings. It can be hard to automate the analysis of these recordings to inform chatbots. But this is changing.
Today, technologies like data ‘lakes’ and ‘warehouses’ help us easily store and manage both data types. At the same time, new tech capabilities allow us to 'tokenise' unstructured data so it can be processed in the same way as structured data. Advances in AI mean this tokenisation can be increasingly automated so these forms of data can increasingly be used to build digital experiences.
This gradual blurring of the line between the two data types means organisations can pull information for users more easily and provide more personalised conversations.
Why does data structure matter for conversation designers?
Conversation designers need to understand the type of data and information their end users want to engage with. Not doing this is like giving directions to someone on the street without listening to where they want to go. You might be able to have a lovely chat with that person but your instructions will be useless without understanding their end goal.
Understanding whether the data users will interact with is structured, unstructured, or both, can help a conversation designer in several ways.
1. Plan conversations with more ease
If data is structured, like in a database, a conversation designer can plan for more predictable and straightforward user interactions. For instance, if someone wants to make a payment through a chatbot, responses to questions and fields are going to be predictable. But if data is unstructured, the designer must account for variability and ambiguity in user prompts and responses. This might mean a conversation designer needs to do more research into different use cases and the language users typically apply to unstructured data contexts. For instance, if someone wants to make an insurance claim and needs to describe an incident, there is going to be more variability in their language.
2. Develop clear natural language processing (NLP) techniques
Unstructured data often involves natural language, or more casual and less predictable language. It's important chatbots and virtual assistants understand user intent from a range of differently phrased inputs. For conversation designers, it's important to understand the structure of their data and the nature of that data so they can work with others to create effective conversation experiences.
Imagine a brand wants to build an online shopping assistant. A conversation designer needs to know what sources of unstructured data could be valuable in building that experience. They might take unstructured data from product descriptions, customer support conversations, social media comments and more and then process these (e.g., 'clean' the data and 'tokenise' it so it can be easily analysed and managed by a chatbot). Analysing this unstructured data can help the conversation designer anticipate problematic user queries as well as different ways a user might communicate a query and then design fallback responses that ensure a smooth experience.
3. Create more personalised experiences
A conversation designer can use structured and unstructured data to offer personalised recommendations, reminders, or relevant content. Understanding the nature of data helps create contextually relevant conversations.
For instance, the conversation designer building a virtual shopping assistant might use unstructured data (say, from social media posts) to anticipate both positive and negative sentiment from users and design conversations that are personalised and reflect this analysis. They might even identify characteristics of a social media comment (e.g., demographic data such as someone's age, sex, location), which is a form of unstructured data, and increasingly be able to anticipate the sentiment and needs of similar users to provide personalised conversations.
4. Avoid biases and risk
Ethics in conversation design is becoming increasingly important for brands and organisations. It's important conversation designers understand the sources and nature of their data, especially unstructured data, so they can identify and manage unhelpful or even harmful biases or risks in that data. For instance, this might be most prevalent in social media data, requiring conversation designers to source data from a diverse range of social media platforms and even use AI tools to analyse data for bias.
How to use unstructured and structured data together
Using unstructured and structured data together can help organisations understand their end users and have more effective ‘conversations’ with them. With tools like Digital Experience Platforms, organisations can consolidate the management and use of data to build more advanced and personalised digital experiences.
Easily pull data from multiple sources
DXPs help organisations centralise the management of their data from different places, such as from forms, social media, app analytics etc. This means conversation designers can access data from various sources, such as databases, APIs, or CRM systems like Salesforce and build a chatbot/voice assistant experience that can easily retrieve this data.
Use AI to understand complex inputs
DXPs can often integrate with AI, helping conversation designers surface unstructured data more easily by accounting for all the variations in natural language.
Build insights to iterate
The centralised analytics DXPs can provide mean conversation designers can build a clear strategy for measuring user interactions, identify patterns, iterate on how their conversation design is structured.
Cross-channel conversations
Users can increasingly enjoy cross-channel conversations with brands. A user might start a conversation with a chatbot on a website and continue it later in an app or be served personalised experiences through email, advertising or other forms of content. This presents incredible opportunities for using conversation design to build brand relationships and support users in the long-term.
Create granular journey mapping
Finally, DXPs help conversation designers understand user behaviours across a brand’s different touchpoints. They can then create more precise journey maps. This can lead to more useful interactions. Instead of broad questions, a chatbot might apply user data to offer specific answers that are more useful for a user.
We've got our own specialist conversation designers on our team, and are always keen to hear what others are doing and how they're meeting the challenges and opportunities that new technologies present. Want to share your thoughts? Let us know in the comments below!
Image by pch.vector on Freepik