1 day from UX Design routine. How to think before design. 3 furniture cases
Many people ask about the process and why did you design this way and not the other way around.
14 years I help businesses enhance UX along with CRO, keeping in mind lifting AOV and Revenue. For a long term.
Let’s get straight to the point.
I wrote about briefing and outlined a generic process.
Now want to share some approaches to design. What matters and why.
You run a D2C furniture online store and want to optimize website conversion along with usability.
Pull data
Choose Metrics and dimensions to see the impact on Revenue and ARPU (avg revenue per user)
Metrics definitions
Product List CTR
The rate at which users clicked through on the product in a product list
Product Detail Views
Number of times users viewed the product-detail page
Buy-to-Detail Rate
Unique purchases divided by views of product detail pages
Product Adds To Cart
Number of times the product was added to the shopping cart
Quantity Added To Cart
Number of product units added to the shopping cart
Quantity Checked Out
Number of product units included in check out
Product Checkouts
Number of times the product was included in the check-out process
Item Quantity
Total number of items purchased. For example, if users purchase 2 frisbees and 5 tennis balls, this will be 7.
Unique Purchases
The number of product sets purchased. For example, if users purchase 2 frisbees and 5 tennis balls from the site, this will be 2.
Product Revenue
The total revenue from purchased product items
To get a general overview without seasonality factors i usually pull data for the last 12 months
Use basic modelling
For above data points I usually apply XGBoost model.
Feature importance that affects on Product revenue.
The total revenue from purchased product items.
As retailer was international, I extracted top features by country.
Here is an XG Boost prediction for 1 country.
The most important features that affect product revenue and unique purchases are:
— Item q-ty
— Product adds to cart
— Buy to detail rate
In order to achieve highest possible items purchased and add to cart from product page, we should provide more and better recommendations from top pages that have high probability of converting users.
Let’s see probability of conversion rate for the next 30 days.
Desktop is higher
Having this segment we discover top landing pages
For example, we shortlisted
— Home page
— PLP 1
— PLP 2
— PDP 1
— PDP 1
— Another landing page from highly ranked by SEO
— Another landing page from highly performing on CPC
Having distilled these pages, it’s also important to look into clickmaps and heatmaps and see which elements are performing better.
In my case users actively using:
— Cart
— Search
— Menu
Problem with home page
Only 50% scroll to the middle of bestsellers.
Possible situation maybe due to not interested in current offerings.
Decision
Segment products into categories to faster navigation
For PDP
Users actively using:
— Colors
— Add items
— Add to cart
Observed problems above
50% don’t even read description on desktop
Very low engagement on the product page due to lack of proper recommendations.
Solution
Notice instead of description we show how it looks in interior first.
So to set expectations before user will dive in into details.
What about listing pages?
Having data on
— top performing categories in terms of revenue
— items quantity impact unique purchases and revenue
— data which items are frequently sold together using market basket analysis
Next step
Launch Looks so user will preview how it will fit their interior.
Now how page with specific look looks like
We are not going to stop there and introduce contextual blocks that will likely increase probability of navigating to PDP.
Let’s also explore data from another perspective.
The shop is underperforming (online) by selling only low AOV items.
It’s intuitively understandable to mitigate the risks.
More seroious purchase is happening offline.
But below we’ll see some hack to takle AOV.
Next
Let’s spot problems early in the journey.
BTW, see my another detailed and comprehensive research for spotting the root cause of the problem in user journey.
It’s an issue as being top-3 traffic pages having 85% bounce is close to disaster. Will show you my appoach later.
But now, we’ll explore buying habits broken down by categories.
Will reveal preferences by style and product specifications that later we can reporpose for other pages.
Now instead of products we sell category.
In this case this is “Living room furniture” and “Home and styling” with better image than original.
This will 99% motivate user click.
For the category like “Home styling” in the first row we show top 3 categories we want user to consider first. Make them bestsellers.
Oh, important to keep in mind.
Tables with numbers are boring.
Visualizing helps quickly grasp the key.
For example, say what’s happening in 1–2 sentences.
It gives us the 1st signal and question.
Is there a way to double down on this and to have every 4–5th transaction with 2 unique items?
Will see it later.
After market basket analysis, it’s more clear about people’s choices.
Will also use it later in design.
Before that we can apply time series clustering of those who converted for desktop and mobile.
Time series clustering helps to understand patterns of behaviour and think about the way to enhance that experience.
Here is how it looks like in a simple way.
In different clusters we may see the same pages that users visit.
It’s totally true as those who convert most likely will visit cart and checkout.
Time series clustering revealed the following patterns.
— On Mobile, unlike desktop, customers visit only limited number of categories (mostly top 5 distict).
— Cluster 1 with website searches (meaning top search page pageviews) mostly looking for tables and chairs as well as rugs that are in stock.
Most of them are visiting sofas and make searches.
Potentially interesting one
— Cluster 2 of logged in users who lost passwords and visiting address
Similar to this cohort, they often check their order history and looking for chairs in stock or probably checking for delivery updates
— Cluster 3 who start with home are searching for products and visitinig chairs and tables in stock. Less tend to add items to wishlist.
Mid intent. Worth discovering further.
— Another cluster 4 consists of custmers who are looking looking specifically for armchairs and cupboards, they seem to plan visiting an offline store to see live.
Clustering relatively fast way to understand pattern of behavior and focus on that one where people often make purchases like tables and chairs.
Keep in mind to allocate resources in 2 cases
Case 1:
where users convert and start research from here.
Like digging deeper into search terms.
Case 2:
Where users NOT Convert and start research from here.
It’s non obvious for many folks out there.
As we naturally tend to focus on that things that convert better.
Think about unexplored opportunities where users might convert but didn’t. We talk about not window shoppers, but those who indicated low to mid intent.
Beyond the data it’s worth checking out CX and address issues accordingly.
Here is the optimization opportunity for Home page
In design process often involved PM, SEO, SEM, DEV and other stakeholders.
Keep in mind their limitations and expectations.
In this case focus on top categories that user shouldn’t even bother searching and they generate 50% of money
Or navigation with deals
Here is the way to optimize PLP for stimulating the purchases of sets as we know we want to increase the revenue per user in top revenue categories.
I’m working on more focused approaches to PLP merchandising, as there are so many hidden gems worth sharing.
