Website review: ASOS

My love affair with ASOS began several years ago when I was at university. I was procrastinating in the computer lab when I noticed a girl in the next row over doing the same thing. However, instead of staring at a 3D model and willing it to complete itself, she was scrolling through a seemingly endless digital smorgasbord of pretty.

Unbearable curiosity took over and I Googled that four letter acronym-turned-word and the rest is history. When it comes to fashion, ASOS has just about everything I could ever need and is the subject of this week’s website review. Here’s what happened when I put ASOS’ US website to the test using the Optimal Workshop suite of tools.

Getting started

Website: ASOS US website

Research scope: Tab labeled ‘Women’ and tab labelled “Men’ ASOS

Not in scope: Marketplace, remaining labels on the ASOS global navigation bar and footer content

Tools: Treejack and OptimalSort

For this study, I decided to use both Treejack and OptimalSort. Treejack is our online tree testing tool (also known as reverse card sorting) that assesses the findability of content. OptimalSort is our online card sorting tool that shows how your users expect your content to be grouped (taxonomy). By using these two tools together, you can assess and understand your information architecture (IA) from multiple angles and pinpoint exactly where improvements could be made.

I also chose to only test the two global navigation tabs labelled ‘Women’ and ‘Men’.  That is where the bulk of the content lives and I was curious to see how it was performing and where the opportunities for improvement were.

Setting up the study: Treejack

Task questions

I decided to include 8 task based questions and divide them equally between the two tabs that I was testing. For tasks 7 and 8, the same content lives under both tabs.

Task questions I used in this Treejack study:

1. Your friend Rose has been invited to a formal event and you’re helping her find a fancy frock for the occasion. Where would you go to help Rose?
2. Your sister’s birthday is coming up and you think a shiny new case for her phone would make the perfect gift! Where would you go to shop for this on ASOS?
3. You’d like to buy a pajama set as a gift for your mother. Where would you expect to find this on ASOS?
4. Father’s Day is rolling around again and you’d like to get Dad some new pants to wear on his daily run. Where would you go to shop for this on ASOS?
5. All of your boyfriend’s t-shirts are looking rather tatty and you think it’s time he got some new ones! The distressed look is so 2016. You’re hoping they might come in a packet of 3 or more- where would you go to find them?
6. You’re on the hunt for a pair of swim shorts for your uncle and you know he’s a size 5XL. Where would you go to find these?
7. You’re starting college a few months and would like to get a Hershel backpack for your day to day needs. Where would you expect to find these?
8. A friend mentioned that ASOS sell advent calendars containing beauty products for the holidays and you’re curious. Where would you go to find them?

When developing these questions, it was important that I kept them gender neutral while still testing two tabs with binary gender labels. To ensure I was being inclusive and also not potentially influencing my participants, I phrased my questions around helping someone else find a specific item or shopping for a gift. Learn more about choosing task questions for Treejack.

Building the IA

Getting ready for testing is really easy! All you have to do is pop your IA into a spreadsheet and then it’s just a quick cut and paste into Treejack. If you’re like me in this case, and need to pull it from a live website, build your tree by starting on the home page and then drill your way down through all the links, copying and pasting them into a blank spreadsheet as you go. You can also build the tree directly into Treejack.

For this study, looking at the screen grab image below, the beginning of my tree looked like this:

Home

Women

SALE NOW ON

Sale: Dresses

Pulling it all together in Treejack

Once I had my tree and my task questions, it was time to create the Treejack study. Treejack guides you through the study creation process and also comes with instructions for your participants that have been developed through a design process and in my experience, work really well.

Next, I bulk imported my tree from my spreadsheet into Treejack, copied my task questions in and set the correct answers. The tool uses these correct answers to add structure to the results.

I set the participant identifier to anonymous because I don’t need to know the identity of my participants for this research study. I also chose to randomize the task order and I allowed participants to skip questions. If they get stuck, it’s better for them to be able to move on to another question otherwise they might just shut the whole thing down! Treejack tells me when and where a participant skips a question so I can see what happened.

Setting up the study: OptimalSort

Open, closed or hybrid?

The binary gender split on ASOS presented a challenge when determining which type of card sort to use. Many of the labels under the two tabs are the same but contain different content. An open card sort would have resulted in an enormous mess of duplicate cards, while a closed or hybrid card sort could potentially have only told me which cards participants identify as being male or female. Given that I wanted to know how users expect content to be grouped, this was not helpful. To make things even more interesting, after removing all the duplicate cards, I was still left with an eye watering 200+ cards per tab. Nobody wants to sort that many cards!

I decided the best course of action in this case would be to run two smaller open card sorts: one for each tab and only with the cards that related to non clothing items, for example accessories, gifts, shoes and more. This brought the study back to 60 cards for the ‘Women’ tab and 58 cards for the ‘Men’ tab.

Choosing the cards

Figuring out which cards to include in each of the two card sorts was fairly straightforward because I already had them in the last column of my IA spreadsheet tree.

Once I picked out all the clothing related items and divided the labels into their respective ‘Women’ and ‘Men’ tab piles, it was just another two quick copy and pasting efforts into OptimalSort and I was ready for launch.

Recruitment

For all tests in this study, I used the Optimal Workshop recruitment service. All I had to do was fill out the recruitment order form under the ‘Recruit’ tab.

The recruitment brief for ASOS was:

Treejack: 30 participants aged 18-35 all residing in the United States

Optimal Sort ‘Women’ tab: 30 participants who identify as female and are aged 18-35 all residing in the United States

Optimal Sort ‘Men’ tab: 30 participants who identify as male and are aged 18-35 all residing in the United States

Interpreting the results

Treejack

My ASOS Treejack tree test study had a total of 35 participants and of that number, 34 completed the study and one abandoned it. The abandonment rate tells me how many people closed the tab or window without hitting the complete button. Before we jump into the results, let’s take a quick look at the Overview tab in Treejack.

This tab provides useful high level information on the completed Treejack study and also tells me where my participants were located. This time around they were all in the US which is exactly what I wanted.

Further down the same page, we have a really cool diagram that shows overall task success and failure at a glance.

This one shows that the results are quite balanced. Tasks 5 and 8 appear to have been a bit tricky for participants in this study, so let’s take a look at what happened there.

Task 5 analysis

I included Task 5 because I was curious to see how the label ‘Packs Save’ would test when participants were looking for a multipack of t-shirts. Only 1 person in this study was able to find the correct location. By looking at the interactive Action Taken pietree that shows an overview of all pathways taken for Task 5, I noticed that 18 participants in this study went into one of the three ‘T-shirts and Tanks’ areas (16 full price and 2 to the sale bin) and didn’t turn back.

That’s more than half of the participants in this study and technically they weren’t wrong. The multipack items do exist on those pages, trouble is they’re mixed in with every other type of t-shirt and tank, and at this point in time there are no filtering options for multipacks on that page. This means that in the real world, more than half of the participants in this study would be required to pick through what currently stands at 5,183 styles of shirt to find what they’re looking for. That’s quite the haystack. I also find it interesting that when I visit the ‘Packs Save’ page, I get the below page titled ‘Multipacks’.

I wonder what might happen if ASOS were to change the ‘Packs Save’ label to ‘Multipacks’. There are many, many wardrobe staples under here and the boyfriend with the tatty t-shirts could also do with some new jeans, socks and much, much more. The potential resulting from one label change, could be quite significant. Including a filtering option on the pages that these items are mixed up in could also provide a lot of value for such a small tweak.

Task 8 analysis

I included this task because I was delighted to find that ASOS not only has advent calendars for the holidays that are filled with beauty products, but they surface that content under both the ‘Women’ and ‘Men’ tabs! I found 5 locations within the IA where this content lives, but in this study 88% of participants were unable to find it. The Action Taken pietree for this task is quite large and dispersed. When I switch the view to the Time Taken pietree (below) I can see that participants spent the bulk of their time scouring the content under the ‘Women’ label.

From here, only 3 participants made it to the correct location and only one visited ‘Face + Body’. Beauty product filled advent calendars are a seasonal item but in my years as an ASOS shopper, I’ve seen their range increase every year and they usually sell out faster than I can get them into my saved items list.

Other insights gathered from this Treejack study include:

• >When looking for a dress for a formal occasion, the ‘Wedding Shop’ received only 1 visitor during this study and 18 participants hit up the sale bin for their fancy frock finding>71% of participants in this study were unable to locate a phone case intended for a female recipient

>Only 4 participants selected the label ‘Joggers’ as the place where they expected to find a pair of mens running pants

>Less than a third of participants in this study visited the men’s plus size section when looking for a pair of swim shorts in a size 5XL

>And lastly, the hunt for a unisex branded backpack resulted in an only slightly uneven distribution of visits between the ‘Women’ and ‘Men’ labels (see below)

OptimalSort

For the two OptimalSort tests in this study, the participant stats were:

• ‘Women’ tab: 53 participants in total with 34 completed responses and 19 abandonments
• ‘Men’ tab: 51 participants in total with 32 completed responses and 19 abandonments

‘Women’ tab results

Participants in this test sorted the cards into groups that largely matched reality and overall, it tested quite well. I started my analysis journey in the Participant-centric Analysis (PCA) tab in the results section of OptimalSort. The PCA tab (below) serves as a high level overview by surfacing the top 3 ways the cards were sorted. In this test, I noticed the major difference between the top 3 card sorts and the current website was the number of groups. Participants sorted the cards quite similarly to the current state, but into broader groups with less granularity.

OptimalSort’s Similarity Matrix (below) shows clear detailed primary clusters in the darker shade of blue along the edges for this study but also shows secondary clusters in a slightly lighter shade of blue connecting the primary clusters. These are the participants who were more comfortable with the broader groupings. The numbers show how many times each card pairing occurred.

Before I move onto the results from the ‘Men’ tab, an interesting group name caught my eye. The results showed that 22 participants created a group called ‘Beauty’ and filled that group with all the items that currently live under ‘Face + Body’ on the live ASOS website. This is very interesting because up until recently, ‘Face + Body’ was called ‘Beauty’ on ASOS. That makes me wonder — why the name change?

‘Men’ tab results

Once again, I started by looking at the top 3 IAs surfaced in the PCA tab and it told a very similar story — it tested quite well and closely matched the current website but with slightly larger groups. This time, I’m going to show you what I found when I peeked under the hood of this test and looked at the Dendrograms tab. A dendrogram is an interactive diagram that illustrates data clusters. There are two different types: Actual Agreement Method and the Best Merge Method. Which one you use depends entirely upon how many participants you have. The Actual Agreement method dendrogram works best when you have more than 30 participants but because I only had 32 participants in this test, I decided to look at both.

The Actual Agreement Method dendrogram (below) shows that many of the data clusters for the current website groupings were hovering around the 50% agreement level in this test.

When I cross-referenced this with the Best Merge Method dendrogram (below), I found that 48% of participants felt that just about everything belonged together which is consistent with my other findings.

Much like with the OptimalSort results from the ‘Women’ tab, I don’t see this as a serious showstopper. Participants in these two OptimalSort tests seemed to prefer larger, broader categories over specialized niches. It could potentially be more of a problem if it was the other way around!

Overall, all the three tests in this study performed quite well with the biggest insight value coming from the naming conventions used. A few very small tweaks around labeling could potentially have a significant impact on findability for ASOS’ US website.

Key findings and recommendations for ASOS

• Participants in the OptimalSort study sorted the cards into groups that largely matched the current state but were broader and with less granularity around item types
• Consider changing the ‘Packs Save’ label to ‘Multipacks’ to not only be consistent with the child page but to also improve findability
• Consider revisiting the label name change of ‘Beauty’ to ‘Face + Body’ as both studies showed a lack of resonance with the new label
• Overall both the Treejack and OptimalSort studies tested well