...

November 6, 2024

Simulated conversations vs fully synthetic data

Share this blog:

Here's how our STRAT7 segment chatbot uses real data as best practice to leverage AI tools.

Welcome to our third blog on STRAT7’s new AI product – our segment chatbot.

At our recent STRAT7 client event in London, the push and pull of synthetic data was a common thread in the presentations.   

The allure of getting access to accurate respondent level data in minutes is a strong pull. Particularly when budgets are squeezed and the technology now exists to generate respondent level data at almost zero marginal cost.  

But at STRAT7 we’ve developed a segment chatbot that simulates conversations, rather than relying on synthetic (artificially created) data. 

Here we outline what is built into the tool to make that happen, and why it’s so important.

1. Real data + synthetic conversations

Our segment chatbot tool builds on actual human respondent level data and the valuable interpretation of the data layered on top (typically in the form of written PPT reports).  

The tool allows anyone to chat to their segments or audience groups, as if they were in the room with them. It simulates or ‘synthetically creates’ a two-way conversation between internal stakeholders and an AI representation of your customer groups. And opens up a wide variety of use cases: 

  • Improve concept testing that goes into primary research.  
  • Test product positioning.
  • Optimise marketing comms.
  • Quickly get back factual information from data tables.

 

Creating a tool with ‘mostly real’ data (see point 2) is also likely to work. But creating a tool on top of purely synthetic data is not something that will lead to anything other than superficial, cursory insights.  

2. Real data + synthetic data boosts

One sweet spot for synthetic data is boosting hard to reach and niche audiences. 

If we consider for example B2B, healthcare and minority audiences, we know that they are expensive, tough to access and often underrepresented on panels. And it has been established that a large amount of responses within these data sets can often be fraudulent. 

STRAT7 Audiences worked with their synthetic partner to find a solution. The starting point was to provide a clean primary data set, verified through our integrity framework, and then use the synthetic modelling tools to ‘boost’ the appropriate audiences. 

Following the successful testing of several data sets, STRAT7 Audiences+  has recently been launched. It’s a solution designed to complement our primary data collection offering, rather than take away from it.  What you get is high-quality boosted synthetically enhanced data, produced on data which is verified at a respondent level.  

This is an innovation which we can continually develop. Leading still with a human first approach – building on verified human data and only being leveraged where it’s not possible to reach robust numbers of those target audiences in the first place.   

BF_Blog_Segment Chatbot_42

3. Real data – why it (always) matters

While there is a fast-growing industry of providers offering synthetically generated data, our current view is that fully synthetic data needs to be carefully and selectively used as part of research: 

  • Fully synthetic data is likely to have less variation than an actual human sample, flattening the nuances and outliers normally seen in quantitative research. This is something that gets worse when analysing data by sub-group. The data may over-generalise and potentially stereotype certain groups, leading to insights that are relatively surface level.  
  • When it comes to connecting different data sets – for example customer survey responses and behavioural data in hybrid segmentations – this will not be possible with synthetic data. 

Our view

Synthetic data is likely to suit those looking for quick, surface level insights, rather than transformative research that hits multiple touchpoints. It will suit clients who want a quick snapshot of the market, rather than creating deep, nuanced understanding of customers and prospects.

Find out more

To learn more about the STRAT7 segment chatbot please contact us, or get in touch with me directly:

Hasdeep Sethi
Data Science Director & STRAT7 AI Lead
STRAT7 Bonamy Finch
hasdeep.sethi@bonamyfinch.com

Infographic

Segment chatbot

Want to talk to customers, as if they were in the room with you?

View our segment chatbot infographic to find out more, including what it is and what it can be used for.

S7_Segment Chatbot_infographic_thumb

Featured content

A bit more about us

Blog

A well trained assistant

Blog

Putting our segment chatbot to the test

We help you understand

Shape your future – where is your future headed and what influences that direction 

Define your difference – what helps you win?

Identify your audience – who are they and what do they expect? 

Galvanise the business – how do you inspire action?

Discover fresh ideas and perspectives

Blog

Lessons in strategy from board games

Case Study

Air Wick

Demystifying consumer motivations and building a cohesive product portfolio