This was a task for the Hobby Squad/Team within the company.
Project Manager, Data Scientist,
6 people Engineering team, UX Designer.
Product Designer responsible for end to end process. Ideation sessions, research, design development, UI, Journey mapping and delivery.
Figma - Wireframing and prototyping
Miro - Ideation workshops
Fullstory - Research and heatmap analytics
Looker - Dashboard for metrics
Experiment 1
Show a birth record match in onboarding
Experiment 1
Show a birth record match in onboarding
Showing the user their civil birth record introduces them to hints/records, setting them up for success with the core tree building tools.
Newly registered users that sign up via the "Family tree" landing page and have a record match.
Users who generated a record match were presented with their birth record.
Out of the users who received a match 72% accepted the record match.
Users who entered the experiment interacted with hints 18% more
The experiment is introduced to users as part of the wider registration to onboarding journey
The experiment is introduced to users as part of the wider registration to onboarding journey
The experiment is introduced to users as part of the wider registration to onboarding journey
This experiment was based around ideas from "The Hooked Model" which theorises that in order to create a habit loop of re-engagement, we must first create a trigger point, where the user is likely to engage, ask the user to perform an action and as a result of that action give them a variable reward or confirmation of success. After they experience this, they are more likely to make an investment within the product, this in this case, it is continuing to fill out the onboarding forms.
What do we know about our users?
Most of our users drop off after onboarding and continue to drop along the in-boarding phase. We want to increase the number of users completing onboarding and in-boarding because we know that they are more engaged, experience value and therefore are more likely to convert to paying customers.
Customers who have British or Irish ancestry and are born between 1940-2000 are the most likely customers to engage with the product as they benefit most from our record and newspaper offering.
Experiment 1 results:
13% increase in users adding more "nodes" (People) to their family tree in their first session
32% increase in members conducting a search
11% increase in "hints" being managed and interacted with.
We can conclude that by introducing users to the concepts of hints and records in onboarding, they are more likely to engage with them in the main product interface as well as search for these records on their own. This proves our hypothesis that by explaining to users how to use the product and helping them experience value upfront they are more likely to engage with the product and conduct meaningful research.
Experiment 2
Quick add for suggested relatives
Experiment 2
Quick add for suggested relatives
A suggested relative is a valuable point of value exchange. When a user encounters a suggested node with detailed information, they are more likely to make a decision and expand their tree.
Allowing users to swiftly accept a suggested relative on the canvas enables them to grow their tree and gain value more rapidly, it also iterates to the user that they don't need to be an expert or knowledgeable to do the hobby.
All users with suggested mother or spouse hints.
Currently live - awaiting statistical significance.
Old experience (Control):
We show suggested mother and spouse hints on the tree canvas but they are generic tiles which take users to a drawer and then finally to a complicated table where they review the information to see if it is correct. This is an industry standard practice, however is not new user friendly and research tells us that new users find the record-merge process confusing and difficult to understand.
We show suggested mother and spouse hints on the tree canvas but they are generic tiles which take users to a drawer and then finally to a complicated table where they review the information to see if it is correct. This is an industry standard practice, however is not new user friendly and research tells us that new users find the record-merge process confusing and difficult to understand.
New experience (Experiment):
If there is only one hint (Smart record match) we allow the user to accept the suggestion on the canvas with one click. Accepting hints allows the system to create more matches and provide the user with more tools to continue their research.
This experiment was based around ideas from "The Hooked Model" which theorises that in order to create a habit loop of re-engagement, we must first create a trigger point, where the user is likely to engage, ask the user to perform an action and as a result of that action give them a variable reward or confirmation of success. After they experience this, they are more likely to make an investment within the product, this in this case, it is continuing to fill out the onboarding forms.
If there are multiple hints for a relative (multiple potential record matches) We show the match with a higher accuracy rating but inform the user there are alternative matches, they must then review all of the matches and accept the one they think is right.
What do we know about our users?
Most of our users drop off after onboarding and continue to drop along the in-boarding phase. We want to increase the number of users completing onboarding and in-boarding because we know that they are more engaged, experience value and therefore are more likely to convert to paying customers.
Customers who have British or Irish ancestry and are born between 1940-2000 are the most likely customers to engage with the product as they benefit most from our record and newspaper offering.
We show suggested mother and spouse hints on the tree canvas but they are generic tiles which take users to a drawer and then finally to a complicated table where they review the information to see if it is correct. This is an industry standard practice, however is not new user friendly and research tells us that new users find the record-merge process confusing and difficult to understand.
Old experience (Control):
New experience (Experiment):
If there is only one hint (Smart record match) we allow the user to accept the suggestion on the canvas with one click. Accepting hints allows the system to create more matches and provide the user with more tools to continue their research.
Old experience (Control):
We show suggested mother and spouse hints on the tree canvas but they are generic tiles which take users to a drawer and then finally to a complicated table where they review the information to see if it is correct. This is an industry standard practice, however is not new user friendly and research tells us that new users find the record-merge process confusing and difficult to understand.
If there are multiple hints for a relative (multiple potential record matches) We show the match with a higher accuracy rating but inform the user there are alternative matches, they must then review all of the matches and accept the one they think is right.
Experiment 3
Celebrating your first Census hint!
Experiment 3
Celebrating your first Census hint!
Explaining the value of unique hint types will help the user evaluate the record and understand the difference between hint types.
Users who generate a census hint and have not managed one before.
In development - aiming to go-live in early November
Hint prioritisation:
We will always prioritize hints from 1921, as they offer a unique value proposition compared to competitors.
Age-Based Variation: For users with multiple 1921 hints, we will randomly show either:
A hint for a person around 35 years old (deeper data on employment and marital status - enables tree depth)
A hint for a person close to 5 years old (potential for upward tree growth)
Goal: This split will help us understand whether new users benefit more from tree growth or tree depth.
Hint prioritisation:
We will always prioritize hints from 1921, as they offer a unique value proposition compared to competitors.
Age-Based Variation: For users with multiple 1921 hints, we will randomly show either:
A hint for a person around 35 years old (deeper data on employment and marital status - enables tree depth)
A hint for a person close to 5 years old (potential for upward tree growth)
Goal: This split will help us understand whether new users benefit more from tree growth or tree depth.
Hint prioritizer - we are aiming to build a data product that maps out the next best action for users and we will use learinigns from this experiment to feed into this as well as help us display these hint types to users in hopes for better engagement and adoption of the product.
Hint prioritizer - we are aiming to build a data product that maps out the next best action for users and we will use learinigns from this experiment to feed into this as well as help us display these hint types to users in hopes for better engagement and adoption of the product.
Tech limitations
Due to complexities around loading the family tree in the background we compromised from a design perspective by rendering a “demo” tree that was not interactive as per the initial vision.
Project velocity
To prove out certain hypothesis and decisions made with low tech effort we ran lots of small iterative experiments on the onboarding to figure out the optimal flow instead of a whole re-design.
This was a challenge because the design had many aspects de-scoped in first iterations which were gradually built upon.
Experimentation with the location of paywalls helped us to learn that showing a user a paywall twice led to better conversion.
We also saw more users complete more input fields by 60%
Users who completed the onboarding were also more likely to engage with the tree canvas better after exiting onboarding.
Trying to introduce the concept of hints within onboarding.
Provide personalised experiences for various cohorts based on their record matches.
Include some intent capture in onboarding to better tailor the product experience for users.
