Hidden merchandising opportunities. UX approach baked by data.

8 min readNov 9, 2023

How normal people are looking at product listing page

How I see product listing page due to professional deformation

What’s going on when people trying to improve PLP’s performance

One might think.
Hey, let’s fill this listing page (aka PLP or product category) with top selling products and that’s it.

Others, more advanced, try ML that arranges products based on prediction for each individual user.

Or, probably, use hybrid approach.
Well, these might work for specific cases and probably works or don’t work as expected.

Some think of filters which is a fair point.

Quick background
For 14 years I help businesses enhance UX along with CRO, keeping in mind lifting AOV and Revenue. For a long term.

UX Problem
Find optimal balance between users’ needs and expectations from PLP and elevating business outcomes.

I can confirm several solutions below proved to be working.
Each case is different. Try and test.

Let’s start with fundamentals
Terms we’ll use
PLP — Product listing page (aka category)
PDP — Product detail page
ATC — Add to cart
AOV — Average order value

Goal of the PLP
Increase # sales and AOV over time.
At least I set for myself for the purpose of this story.

Let’s break it down:

PLP -> PDP CTR
(PDP views / PLP views) * 100
PLP -> ATC rate
(PLP views / Products from this PLP added to cart) * 100
PLP -> Initiated checkout rate
(PLP views / Products from this PLP initiated checkout but didn’t convert rate) * 100
PDP sales
# of transactions of products from this PLP
PLP Conversion rate
(PLP views / # of transactions of products from this PLP) * 100
Additionally, predictors of interest (low to high intent)
% Returning users
% Added to favourites, comparison etc
% Filter applied
External factors
Ads campaign engagement
— Campaign CTR
— Direct traffic performance

Almost any PLP above the fold is very valuable real estate where each element should work for the general goal (remind it here: Increase # sales and AOV over time).

Several data-driven approaches I apply to reach this goal.

See what’s going on first.
Start with EDA (exploratory data analysis) over time.
Aggregated numbers are rearly actionable even if you compare periods.

Let’s blend Google Analytics (GA) data + CRM

Disclaimer: Due to NDA restrictions in examples below I’ll be using dummy data but close to reality.

The main thing here is to grasp the concept first.

Pull data from GA

Pull data from CRM

Now we have to join the date and category tables to get this table

We are getting close to EDA except few things to be considered.

Let’s calcualte AOV and conversion rate.
In this case we assume that user who visited category page made a conversion.

And filter just to category “Lips”

Important notes to make sure you come up with the right analysis:
For the simplicity:

1. I’d suggest to remove outliers that won’t impact your analysis.

2. Make sure this dataset a complete record of all transactions for the given time period?

3. I take summer period that might not affect purchasing behaviour.

4. In this category there were excluded gift with purchase items not to skew the analysis.

5. We are assuming that key traffic sources are relatively stable for the given time period.

6. No changes to the webpage layout, navigation, or checkout process

Not a rocket science so far, right?

Let’s explore data.
Question 1
In what AOV ranges there is a relatively high conversion rate over time?

Key things to know about this question:
— Relatively high conversion rate to double down on it
— Over time, meaning the conversion rate is relatively steady
— AOV ranges, meaning what products in that range should we promote to lift revenue, AOV, and probably conversion. Some really high performing products could have a lower visibility (>20 position) on PLP.

These numbers can be used as a baseline before you start experiments

Explanation
Conversion rate over time for 5 different AOV quintiles labeled Q1 … Q5. The quintiles divide the AOV data into 5 equal parts, ensuring a balanced representation of AOV ranges.

Q1: The first quintile with AOV ranges from $5 to $16, representing the lowest AOV range, has the highest average conversion rate of 0.76

Q2: The second quintile with AOV ranges from $17 to $25, has a lower average conversion rate of 0.3

…

Q5: The fifth quintile with AOV ranges from $44 to $58, representing the highest AOV range, has the lowest average conversion rate of 0.15

Logically, less expensive items often tend to convert better and vice versa.

Next questions
1. Given that Q1 + Q2 give us relatively stable higher conversion rate , what would be the lift if we bundle products?

2. What would be the best bundling combination Q1 … Q5?

3. What combinations should I consider to elevate AOV, Revenue & probably conversion rate?

Start with a baseline first
Unlike the previous table, It helps further to compare your efforts before and after in general.

Hope it’s straightforward so far.

Let’s answer above questions 1 by 1
Given that Q1 + Q2 give us relatively stable higher conversion rate, what would be the [estimated] lift if we bundle products?

Hard to imagine, right?
Let’s visualize it.

Revenue

AOV

Conversion Rate

The estimated lift for new revenue is shown separately to clearly indicate the potential increase or decrease after bundling, whereas AOV, and conversion rate won’t be harmed.

Background for calculation:

Best-case Estimated New Revenue (Bundle Q1 & Q2)
Formula:
Best-case Estimated New Revenue = Revenue % of total (Q1 + Q2) × (1 + Best-case Lift)
This assumes a 10% increase in the combined revenue percentages of Q1 and Q2

In numbers
Combined Revenue % of total (Q1+Q2): 39.87%
Best-case Lift: 10%
Best-case Estimated New Revenue: 39.87% × 1.10 = 43.86%

Best-case Estimated New AOV (Bundle Q1 & Q2)
For the best-case scenario, we assume the increase the average AOV by 10%.
Best-case Estimated New AOV=Average AOV×(1+Best-case Lift)

Firstly, calculate Average AOV (Bundle Q1 & Q2)
The Average AOV is calculated by taking the mean of the AOVs for Q1 and Q2.

Average AOV= AOV (Q1)+AOV (Q2) / 2
Average AOV=(29.48+21.91)/2=15.695

Best-case Estimated New AOV = 15.695 × 1.10=17.2645

Best-case Estimated New Conversion Rate (Bundle Q1 & Q2)
For the best-case scenario, we assume increasing the average Conversion Rate by 10%.

Firstly, calculate the Average Conversion Rate by taking the mean of the Conversion Rates for Q1 and Q2.
Average Conversion Rate= (0.76+0.30)/2 =0.53

Best-case Estimated New Conversion Rate = Average Conversion Rate × (1+Best-case Lift)
Best-case Estimated New Conversion Rate=0.53×1.10=0.583

Quick summary
For this example with bundling Q1 & Q2 one can expect definitely increasing revenue having relatively stable averaged conversion rate and AOV.

More optimization opportunities

In this case let’s iterate over top 5 bundling combinations that have the opportunity to maximize the lift in AOV, Revenue, and Conversion Rate

2 highlighted segments worth trying to test first for this specific PLP.

Compare these estimations with previous baselines

and general baseline

Worst-case Estimated New Lift Revenue:
Worst-case Estimated New Lift Revenue = Combined Revenue % × (1−Lift Percentage)

Worst-case Estimated New Lift AOV:
Worst-case Estimated New Lift AOV=Combined AOV × (1−Lift Percentage)

Worst-case Estimated New Lift Conversion Rate
Worst-case Estimated New Lift Conversion Rate=Combined Conversion Rate×(1−Lift Percentage)

The least you can do

Once you identified these segments make sure the products within those segments are visible in the first 20 positions.
Prioritize unobvious performers
— Identify for a specific PLP top 20 most converting products (preferably on SKU level). Make sure these products also contribute to revenue.
— Check their positions in PLP. If it’s below 20+ move them to the top 20
— In case you see product views for high converting products less than avg, make sure to add to product feed to promote on PPC.