P&G Alexa Chatbot


Conceptualising, building, and testing new research methodologies using Smart Home Devices.


The Work

Given the difficulties faced by researchers in getting meaningful, contextual research data from product testers in a timely, efficient and scalable manner, we worked closely with P&G to conceptualise, build and test a new research methodology using Alexa devices.

The Challenge

Existing research methods come with their own sets of benefits and drawbacks; Quantitative methods are scalable, but suffer from biases in recall accuracy, as well as a limitations of outcomes dictated by available answers. Qualitative methods, such as in-person lab testing, are hard to scale, and may be subject to participant bias due to the lab environment. Overall, the biggest challenge came from logging accurate behavioural data from product testers, over a period of time.

We wanted to test out how a voice based, smart device would fare in giving the researchers meaningful, contextual data, given that these devices were portable, were minimally intrusive, and supported hands-free operation.

Key Insights

VOICE BASED CHATBOT DESIGN

One of the biggest takeaways of designing a voice based chatbot, or any other chatbot for that matter, was that content optimisation is a key factor with regards to creating a great user experience. We found that keeping the messages short and to the point worked better in directing the user, and that keeping answering patterns (such as ranked opinion polls) consistent improved recall. Also, adding “personality” comments at key points in the chat flow eases the monotony of engaging in a voice based questionnaire.

CONCEPTUALISING A NEW RESEARCH TOOL

Utilising Alexa devices as a research tool was a challenge in itself, there being nothing else on the market like it. To achieve this, we deconstructed the current available research methodologies, analysed their objectives, and re-conceptualised it into a form suitable for the Alexa device.

What we ended up with was a research package, with Alexa at its core; instead of filling out surveys at the end of an identified activity, product testers now answer questions given out by the Alexa Skill during the activity itself, improving contextual and recall accuracy. Instead of having a long survey, product testers now engage in multiple quick, short research activities over a period of time. Instead of trying to recall their behaviours during an activity after the fact, product testers are now able to log their behaviours during the activity itself, using Alexa commands.

The Work

Given the difficulties faced by researchers in getting meaningful, contextual research data from product testers in a timely, efficient and scalable manner, we worked closely with P&G to conceptualise, build and test a new research methodology using Alexa devices.

The Challenge

Existing research methods come with their own sets of benefits and drawbacks; Quantitative methods are scalable, but suffer from biases in recall accuracy, as well as a limitations of outcomes dictated by available answers. Qualitative methods, such as in-person lab testing, are hard to scale, and may be subject to participant bias due to the lab environment. Overall, the biggest challenge came from logging accurate behavioural data from product testers, over a period of time.

We wanted to test out how a voice based, smart device would fare in giving the researchers meaningful, contextual data, given that these devices were portable, were minimally intrusive, and supported hands-free operation.

Key Insights

VOICE BASED CHATBOT DESIGN

One of the biggest takeaways of designing a voice based chatbot, or any other chatbot for that matter, was that content optimisation is a key factor with regards to creating a great user experience. We found that keeping the messages short and to the point worked better in directing the user, and that keeping answering patterns (such as ranked opinion polls) consistent improved recall. Also, adding “personality” comments at key points in the chat flow eases the monotony of engaging in a voice based questionnaire.

CONCEPTUALISING A NEW RESEARCH TOOL

Utilising Alexa devices as a research tool was a challenge in itself, there being nothing else on the market like it. To achieve this, we deconstructed the current available research methodologies, analysed their objectives, and re-conceptualised it into a form suitable for the Alexa device.

What we ended up with was a research package, with Alexa at its core; instead of filling out surveys at the end of an identified activity, product testers now answer questions given out by the Alexa Skill during the activity itself, improving contextual and recall accuracy. Instead of having a long survey, product testers now engage in multiple quick, short research activities over a period of time. Instead of trying to recall their behaviours during an activity after the fact, product testers are now able to log their behaviours during the activity itself, using Alexa commands.

 
hi_wht