CX STRATEGY / @FJONG / SERVICE OPTIMIZATION / OPEN WEBSITE
FJONG Open Website
UNDERSTANDING CONTEXT
Based on analyzing the FJONG conversion funnel, it was clear that requiring there were two major choke points. The first was turning site visitors into registered users (micro-conversions), and the second was turning registered users into paying customers (macro-conversions). I hypothesized that the site’s log-in wall played a key factor in both of these choke points and set out to investigate further.
FJONG conversion funnel with attention on choke points (actual conversion percentages not displayed at FJONG’s request).
DEFINING THE PROBLEM
From the customer perspective:
By collecting first and second-hand customer feedback (interviewing customers and customer service/sales staff), I observed a common frustration upon first arriving on the FJONG landing page—there was so little information about what FJONG was or what service it offered. FJONG was doing no paid advertising at this time, so most visitors had arrived to the site either through a word of mouth referral or press mention, thus they has some knowledge of the company, but were curious to learn more, and the log-in wall prevented them from easily doing so. Many were frustrated because they knew FJONG rented clothing and their preferred way of evaluating a clothing shop/service was to browse its inventory, which they were unable to do.
The FJONG.com landing page, which provided few clues about the service FJONG offered.
Many visitors simply left the site after encountered this frustration. Others were willing to register simply to get more information to evaluate whether the service was for them or not. But after registering many left quickly again after determining they did not find FJONG’s service offering appealing.
From the business perspective:
The log-in wall was creating commercial problems both by discouraging micro-conversions and creating an unrealistic user base filled with people who were unlikely to ever turn into customers. As an early-stage company, this was problematic for FJONG because it was difficult to identify which users were likely leads that should be nurtured.
FJONG’s decision to keep its website behind a login wall was based on its photography. In order to appeal to customers, FJONG had to quickly accumulate a significant amount of inventory. While purchasing these items and registering them on the site was relatively quick, taking unique photographs of each item was time consuming. As a temporary solution, brand photos would often be used to represent the outfit until FJONG’s own photos could be taken, and some of these photos could not be publicly displayed on the internet. As FJONG was struggling to catch up with photographing all the outfits that needed unique images while also actively acquiring new inventory, this problem situation would not naturally be remedied with time.
IDEA GENERATION & EVALUATION
Site visitors needed to be able to evaluate whether FJONG was a service that appealed to them or not before taking action (registering). FJONG needed to accurately gauge what type of users their service genuinely attracted in order to nurture users into customers. Both of these needs pointed toward the necessity of removing the log-in wall from the FJONG website. By doing so, visitors would be able to evaluate FJONG’s service offering by reading more about the service and browsing their inventory, and FJONG would be able to see what types of visitors and visitor behaviors lead to conversions. However, the photography restrictions made the approach to open up the website more complex. With these facts in mind, three main ideas were generated:
Create a content-rich landing page: Creating a landing page explaining FJONG’s concept in depth with examples of inventory would be relatively easy to build and would give visitors more information on FJONG’s service before registering. But it would not allow them to browse inventory, which users indicated was their preferred way to evaluating a clothing shop/service.
Build two versions of the website: Another ideas was to allow site visitors to browse some inventory before logging in but only displaying outfits with proper photos. Upon log-in, additional outfits would become visible. This option would allow visitors to evaluate FJONG by browsing some of FJONG’s inventory before registering, but not all. The success of this approach would very much depend on how attractive the public outfits were and whether visitors deemed this limited selection sufficient to evaluate whether registering to see more was worthwhile. This option would require some significant UI changes as two versions of the site would have to be navigable and regularly updated..
Open the website: The most comprehensive solution would be to open up the entire website, but to not make outfits without proper imagery available for rent. Like the option above, this success of this option would depend very much on how many outfits were visible on the site (and thus on establishing an efficient photography process). However, visitors would be able to browse all available inventory, which would best address their need to evaluate the service. This option would also require significant UI changes, though unlike the option above, registration would not alter most screen’s UI.
Each approach had its pros and cons, both in terms of how well it solved the problems and in terms of complexity and required resources:
Based on FJONG’s small team and limited resources, the following approach was decided upon: first, FJONG would first implement a content-rich landing page, then focus on creating a more efficient photography process, and finally open the website fully, avoiding the resource-intensive two-view option.
STEP 1:
CONTENT-RICH LANDING PAGE
To define the content to be included on the landing page, I returned to the previously gathered user insights to understand what questions first-time visitors had. It was obvious that visitors wanted to know what inventory FJONG had and they wanted to know more about what FJONG was was and how it worked.
In analyzing these insights with the design team, we also identified that there were two types of visitors, those primarily interested in renting clothes from FJONG as a way to expand their wardrobe, and those primarily interested in lending clothes out through FJONG as a way to clean out their closet and make some money.
In observing FJONG’s social media activity, we also noticed that just after the service was launched, there were many concerns shared about the hygenic factors of renting clothing, but that these had largely disappeared as more FJONG renters began to share their experiences and enthusiasm for the service. Consequently, we concluded that this sort of social proof would be very helpful to display on the landing page.
These insights helped us to map out the informational hierarchy of information to include on the landing page. Presenting information about what FJONG does and a preview of inventory would be helpful to all visitors (potential renters would understand what type of garments are available to rent and potential lenders would understand what type of items FJONG would accept from them). Then visitors would be directed into either a renter or lender track, where more specific “how it works” information and testimonials would be visible to them, in order to reduce the confusion of displaying the details of both sides of the complex service to everyone.
The FJONG design and marketing teams collaborated on the specific content and imagery to display on the homepage.
My role was in leading the team, ensuring that we all remained focused on the purpose of the redesign, and drafting the written content.
In order to keep this new landing page from feeling too static, we collaborated with marketing (and the dev team), to define a process for updating the hero image with each new marketing campaign, as well as the garment category images, customer testimonials and photos on a regular basis.
Within a few weeks of launching this new landing page, we observed a significantly decreased bounce rate and modest increase in registrations, suggesting that the change was having the intended effect of educating visitors about FJONG and helping them to make an informed choice whether or not to sign up.
STEP 2:
MORE EFFICIENT PHOTOGRAPHY PROCESS
Stage two of our agreed upon solution consisted of finding a more efficient way to photograph FJONG’s inventory, so that all existing inventory could be publicly visible and newly acquired inventory in the future could be visible and available for rent as quickly as possible. For the stage in the project, I collaborated closely with FJONG’s lead photographer, Benedikte Lie, to ensure we were both aligned in understanding the current process, need for improvement and available options.
In terms of context, we first reviewed background research on the importance of photography to e-commerce platforms. In traditional shopping settings, customers can physically inspect clothes, feeling the fabrics, seeing the color in different lights and looking at the fine details of the garment. In order for our digital product photos to convey enough information for make customers comfortable enough to rent online, it was important that the photos be adequate in number, detail and realism.
With this in mind, radically reducing the number of photographs per outfit in order to save time would not make sense. Nor would it make sense to shoot the clothes on mannequins or laid flat, as it would be more difficult to imagine how the garment would fit a real person (not to mention be inconsistent with the FJONG brand that celebrates women). We did consider taking fewer 3D images or adding a zoom feature to negate the need for detail shoot, but technical limitations made these solution unsuitable at this time. It was clear that, for now, we would need to focus on improving efficiency within the process itself.
Mapping the Current Process
There are four main steps in the FJONG photography process:
Booking models for a photoshoot and assemble garments to be shot (as well as shoes, bags, etc. for styling)
Dressing the models and taking photographs (5-7 photos per outfit has proven optimal for FJONG)
Post-processing the images from shoot (select which to use, color correction, cropping, edit out tattoos, etc., export)
Uploading final images to website and making outfit publicly visible
In terms of time (per garment), post-processing takes the longest, followed by actually shooting the photos. Uploading them takes minimal time, as does preparing for the photoshoot. Accordingly, our efficiency efforts would be focused on steps 2 and 3.
Relative time to complete each step of the FJONG photo process (specific timings withheld at FJONG’s request).
More Efficient Photoshoots: Ideation & Testing
By observing a photoshoot and discussing it with Benedikte, it was clear that actually taking photos took up the minority of the time—more time was spent on the model getting dressed/undressed and selecting accessories to pair with each outfit. Benedikte herself spent a lot of time moving around the room, helping with these various activities. To figure out a better way to do things, we conducted experiments around several different scenarios:
Scenario 1:
We first experimented with including a separate stylist, who would be incharge of helping the model get dressed/undressed and selecting accessories, so Benedikte could focus on the photography. This was somewhat more efficient, as Benedikte did not have to change roles continuously, but now she spent time just waiting for the model to be ready, which was clearly not efficient. This sparked the idea that we should use multiple models, so that one could be getting ready while the other is being photographed.
Scenario 2:
We scheduled a photoshoot with two models, one stylist and one photographer, on the assumption that it would take roughly the same amount of time for a model to take off one outfit, put on another and add accessories, as for the photographer to shoot one outfit. This assumption proved correct and we were able to photograph almost three times the number of outfits as Benedikte and a model working on her own in the same amount of time.
Scenario 3 (untested):
Based on this success, we considered whether we could reduce the actual photography time even more with practice and use three models and two stylists at a time, so that as one model is getting undressed, another is getting dressed and another is being photographed. However, we were never able to test this scenario as aligning the schedules for three of our intern-models proved near impossible. As paying to hire professional models was out of budget for FJONG at this time we landed on scenario two as a viable solution for the timing being.
As we repeating the new two-model photoshoots, we were able to learn more to refine the process even more. For example, the faster pace of constantly the new photoshoot format was more strenuous for the models. We found that the standard four-hour shoot needed to be broken up with a break with some food and that it was important that the stylist and photographer encourage the models to take pauses to drink water as needed. We also happily discovered that being a stylist for these shoots was viewed as quite fun by many members of the staff and they were happy to volunteer for these positions, reducing the need for additional paid staff.
More Efficient Post-Processing: Ideation & Testing
I began investigating the post-processing step by observing Benedikte at work and discussing the process with her. Aside from some one-off type edits, such as removing tattoos or other distractions, such as one diabetic model’s glucose monitor, a large amount of Benedikte’s time was spent selecting which photos of each outfit to use. Many images would have to be discarded because the model was making a strange expression. It was clear that FJONG was committed to continuing to use live models, so eliminating the human element was not an option altogether.
In researching the photo approaches of other e-comm sites, we discovered something interesting: some sites cropped their images at the neck so their models’ heads were not shown. This approach would obviously save a great deal of post-processing (and photography) time as fewer images would have to be taken and sorted through in order to select an adequate number of “good” images, as facial expression would be irrelevant. However, FJONG’s team was split in their their reaction to this option. Some viewed such “decapitation” as inhuman and not in alignment with FJONG’s brand values of celebrating women’s individuality. I turned to academic research on the subject to help the team facilitate a more objective discussion on the matter.
“...Cropped [headless] product pictures made it
easier for female consumers to identify with the target group for the products, which meant higher self-referencing – and consequently higher purchase intentions – for the products. ”
Dr. Hanna Berg of the Stockholm School of Economics, wrote her dissertation on just this subject, concluding that there are higher purchase intentions for the clothing displayed in pictures with cropped, “headless” models for two main reasons. First, as female models are rather dissimilar to the average female consumer, removing some of the characteristics of the models made it easier for female consumers to identify with the models in the pictures. This higher self-referencing leads to higher purchase intentions (e.g. it’s not the case for women shopping for men’s clothing). Second, when models have a face, human neural processing dictates that we focus on their face rather than their outfit. Furthermore, women are willing to pay less when an ad features a very attractive model versus moderately attractive model when they are thinking about their gender and body (as when shopping for a dress), and faces are the primary key factor in establishing “attractiveness.”
This research suggested that removing the head’s from our model’s photos would not only save us time, but could also actually increase rentals for FJONG. But, of course, not if we alienated our customers, or our models, with the choice.
We talked personally to all of our models to hear their perspective on the matter. Rather surprisingly (at least for me), all of them expressed relief at the idea that their faces would not be displayed for product. It made the job feel easier and less stressful.
To gauge the option of our customer community, we turned to social media, producing two very similar product-focused instagram stories, one showing models heads and one not. There was very little difference in reaction to these two stories and no negative comments about not showing their heads.
Product photos not showing the model’s head became the default after testing showed positive reactions from our models and customers.
In sum, these various sources of data confirmed that the efficiency decision to publish photos without heads was a good one. And indeed, combined with the above improvements to the photoshoot process (and some quick tech wins to speed up the uploading process), it now takes less than 25% of the original time to complete the photos for one outfit.
Now that we optimized our manual system and have a better understanding of our options and pressure point, we are currently investigating options for additional standardization and automation, such as using a robotic photo environment (such as Style Shoots), as our future budget permits.
STEP 3:
REMOVING THE LOG-IN WALL FROM THE WEBSITE
Once the photo process was optimized and most inventory had publically viewable images, it was time to tackle the technical and UI aspects of removing the log-in wall from the website. For this project, I worked closely with several members of the FJONG development team and our in-house UI designer, Jens Bringsjord.
I began by discussing the goals of the project with the project team, particularly since it had been some time since initial discussions began. And since we would be making changes to our analytics enabled website, we also agreed upon key trackable metrics that would help to evaluate whether or not these goals were being met.
Goals
To make the FJONG website visible to more people (build awareness of FJONG)
To make it easier for new visitors to understand FJONG’s value proposition, increasing expected utility
To make it more attractive for visitors to convert into users (micro-conversion, consideration) and customers (macro-conversion, purchase)—because they have found something(s) they want to rent.
To enable digital marketing initiatives (links to specific pages, retargeting ads, SEO, paid search, etc.)
To achieve these goals with the least amount of time and resources spent = maximize ROI
KPIs
Site visitors (new and total) (total, and specifically from organic search and from digital marketing) - expected to increase
Bounce rate (leaving site after viewing only one page) - expected to decrease
Micro-conversion rate (% visits that lead to a sign-up) - expected to decrease initially (more info available to decide if you want to be a user before signing up)
Macro-conversion rate (% visits that lead to a rental) - expected to increase over time, but influx in site may show decrease or now change initially
Percentage active users (registered users who have placed at least one order in the last six months) - expected to increase over time
Strategically, because several other major changes to the website were underway, the goal was to remove the log-in wall with as few other changes to the site as possible.