Day 2–Unlocking the power of advanced quantitative methods
— Hi there. I’m here to talk about the following quantitative methods and how they can enhance your UX practice. They are:
- Cluster analyses to replace personas
- Perception maps to show qualitative relationships betwen concepts visually
- Binary logistic regression to predict human behavior
- Conjoint analysis to uncover human preferences without asking them
— So let’s start talking about it
— For method one, whether dealing with users or customers, a cluster analysis can beused to classify different profiels of undergrad students tor drive marketing messages
- All customers are not equal and found out having at least three groups
— Helped marketing team figure out claims to understand marketing material and doing right group of users
— Combine with qual research to come up with 11 statements and ask for each person how much they agree with each statement. Grouping users and mathematics to group them together
— Seeking similar answers to different questions and grouping users to figure out best way to segment features, products and so one
- Which questions can we answer?
- How can we segment customers based on consumer preferences?
- How do different patients cluster on disease symptoms and medical history?
— Next, what are perception maps?
— They are strange, but easy to understand
— Before people decide to use it, let’s talk about example of co-branding effort where we would see what would happen if we mix different brands together
— See brand associations above and compare whether good for certain features like photos at night and different aspects are correlated with each other
- One aspect can be reliability or good for work
- If points close to each other, relationship based on space revealed
— Attributes can cluster with each other along a particular brand like Apple
- Each brand fits into different space, and attributes clustered with each one
— Example of search for shopping malls and seeing how mall was losing customers and connection between outdated mall facilities and links to competitors
— Another example of music
- Client playing EDM music in restaurants, but what would happen if they played another genre, what should they play
- Blue represented restaurants
- Red represented customers
— Next, for binary logistic regressions, we predict the probability of a binary outcome (e.g. success vs. failure, purchase vs. no purchase) based on predictor variables
— And example of this: How many factors do people consider when choosing a college 1? 5? 10?
- Answer was 61 factors and taking into consideration each one
— So qual could show 61 factors, but of course qual research can’t cover this
— Showing all 61 reasons to stakeholders, and having them pick would be crazy
— Different aspect of choosing college and based on diagram distance in students and college show 11% of decision
— But using weights it was 70% of decision, and comparing two groups at local college, and competitor college
- Found statistical results that were significant
— Not shown via survey and typical
— Other questions can include:
- What characteristics predict a fraudulent online transaction?
- What variables increase the likelihood of a consumer purchasing a product?
— Helping figure out value of features, and how conjoint analysis works to measure
— Focus is on trade-offs to make
- Choosing anything is hard, as trade-offs require compensation of X
— Customers tell researcher what they want to hear, and surveys measure attitude not behavior, and attitude is just 28% of real behavior
- Conjoint analysis shows choices
— Taking basket of goods for how attributes have different weights based on different clusters and here we have
— Different attributes within each one to have levels, and those that you can understand which ones have highest preference. If you combine levers you will have perfect combination
— Another example of ice cream for flavor, price, and size
—To summarize, here is what these quantitative tools can do
- People’s behavior based on common characteristics, instead of using fictional speculative personas
- Understand relationships among qualitative concepts in a visual manner
- Understanding what differentiates behavior among 2 groups in a binary situation (cancel/not cancel, keep/return items, choose brand A/B)
- Understanding what drives preferences for products and services
— QuestionPro is recommended tool linear regression
— AYTM for perception maps
— DisplayR as another perception map tool
— Conjointly as tool to show collective data and simulate scenarios to show preference share
— Python for all kinds of analysis but demands study from your part
- To learn more follow-me on LinkedIn or subscribe
Q&A
- Time used in designing a service and situation and method applied?
- Favorite question and using all techniques and cluster analysis userful to analysis if creating online school, different segment of customers and cluster analysis can create different prices,
- Regression to grasp why customer losing customers and compare two offers from branches of restaurant and see how behaviors differ
- Conjoint analysis key tool for start-ups and perfect price
- Perception maps to understand user needs and order in qual data and list is infinite