Business Goals

• Reimagine a popular deal-search app as a universal binary for iOS and Android phones and tablets.
• Create an experience that supports multiple online shopping habit styles
• Create a concise and modern visual language to communicate simple concept
• Drive up Monthly Active Users metric and establish user base value
• Cultivate and grow company marketing

 
 
 
 

The story of this app really begins around iOS5. Its design has stayed the same for the first few years, until the release of iOS7 which had such a radically different visual style that basically everyone who had an app had to either redesign theirs or risk looking dated (and all the negative connotations that may carry).  Fortunately the new visual language had rules straightforward enough to pick up.

At the start, the app was designed for the largest screen - the ideal way for users to come into contact with large images of lucrative products. The app started out as a universal binary.

 

At first we were out to support our most loyal user base, the dealmakers.

Later:
The goal was to grow our Monthly Active Users steadily and develop features that would provide the ‘hockey stick’ rapid acquisition of users to establish product value.

 

 
Progression from first release of the redesigned experience and rigid 'dashboard' structure of content to a less user-centric experience with variable content. Between the first version (left) to the last (right) user base has grown 60x in loyal user base (returning consistent users).

Progression from first release of the redesigned experience and rigid 'dashboard' structure of content to a less user-centric experience with variable content. Between the first version (left) to the last (right) user base has grown 60x in loyal user base (returning consistent users).

 
 

Challenges

The biggest challenge of the development was having to manage a universal binary for iOS supporting both phones and tablets (and supporting 4s devices onward, requiring legacy implementation) and an Android binary that would work on all Android phones and tablets. It was imperative to keep the experiences consistent even though some of the core navigation functionality would have to differ due to the disparities between the iOS and Android User Interface Guidelines.

 
 

Audience

 

The company's web-based search engine had around 10 million searches a month, so the target audiences were established. However extension to popular hand-held devices widened the band for the Dealfinders (people hunting for deals/lowest price online) and Local Shoppers (Who find local deals online and go shopping).

 

Disposable Income
15-25

 

Dealfinders
25-45+

 

Local Shopper
30-45

 
 
 
 

Roughing it Out

 

Expectations were firmly in place, not just from the product team but from the users of the previous incarnation of the app.

The wireframes were built in-house, then a boutique visual shop was contracted to prototype it out as a native semi-functional skeleton app which the in-house team would fill with features. Then the visual style was reskinned.

The original app had a core 'trendsetter' audience who later signed up on to the beta before we even had a stable build, allowing us to get some some insights.

 
 
 

Method to the Information

Late snapshot of the unified iconography map.

Late snapshot of the unified iconography map.

If you've ever bought anything online, chances are you've experienced that your average buying session is a highly orchestrated madness. You might have 20 tabs open in your browser and you rely on remembering which tab matched up with which product.

 

The legacy of the app empowered us do what we've done but better-  provide the users with a content-rich homescreen that would be the many-points-of-entry springboard into the rest of the experience. It was perfectly logical - you want to expose users to as much content as possible. 

 

Exposure to each piece of content = chance to engage.

 

Certain data was catered to users based on their onboarding, which would also pre-populate their Brands and Stores and encourage them to add others to their favorites. Making the users actually choose the stores they'd like to subscribe to during on-boarding was tested with minimal impact on the web's homepage. It did not go well; resulted in 46% drop-off during onboarding. Thankfully this was tested on the homepage which rarely received direct traffic.

 

All these features were secondary to product search, the primary purpose for most of our users, which the metrics of the previous generation app suggested.

 

In the end, the most stable flow for the core functionality was a simple and intuitive hierarchy:

Home < Search Results < Product

 
 
 

The Ever-Shifting Marketplace

 

Soon after launch we discovered two important things:

• Phone installs accounted for over 2/3 of all installs.

• Tablets accounted for about 28% adoption initially but saw a steady incline onward

 

Because of their common use case, tablets were then considered a leisurely device. Since tablet installs were growing in number and slowly catching up to the phones, we took an educated guess that by exposing them to more content and faster, we could increase session time exponentially. Mission was accomplished when we implemented a 'feed' system with content from the user's saved stores and brands (and their respective facebook feeds). Traffic doubled and then tripled within two and three months.

The main challenege onward was whether the content was lucrative enough to convert casual tablet-era window-shoppers into buyers; a slight increase may have been a causality of specific types of sought-after products appearing on specific stores' feeds and their deals. The other dependency was on the product suggestion alorithm.

 

During Thanksgiving 2014 the mobile traffic exceeded desktop traffic for the first time. For the first time in history, online sales were the strongest on mobile devices - phones and tablets.

Even major retailers's mobile traffic accounted for at least 2/3 of their turnovers.

 
Source: IBM Experience US Retail Online Holiday Shopping Recap Report 2014

Source: IBM Experience US Retail Online Holiday Shopping Recap Report 2014

 
 
 
 

Universal Binary

 

At the start we used iPads to lay out the experience with the intention of having phones be the shrunken version which may or may not drop a feature or two.The next major iteration - replacement of the dashboard for feeds proved that users with compulsive unending browsing for related products, rather than general searches - as the problem is people don't necessarily know what they want to search for, so the entire experience needs to be able to guide them exactly where they want to go without requiring too much thinking or effort on the users' part.

 
 
 

Android

 

Because there was no groundwork for the Android version of the app, the Android development was 3-4 months behind, piggybacking off the iOS' design, and giving its adversary adequate time to test new features on real users.

The Android team was mostly based out of Russia and Romania.

 
 

Testing

 

Recorded Think-Out-Loud Testing (both screen with touch tracker) and a webcam recording of the user’s emotional status/reactions was done with the first fully-functional build. The users were asked to play out scenarios (eg. you're going camping, find a tent and cooking equipment). This repeatedly showed that users needed many visual cues to pay attention to the searchbar and were mostly looking to tap and browse rather than perform their entire shopping experience in one go.

From that came the idea of implementing store and brand feeds - which would provide users with exactly the type of content they should want to look at, putting the ball back in the store's own court.

 
 

Marketing

 

The company's had a strong presence at several trade shows it frequented, meaning the branding will have to have grown organically and concurrently with the experiences. The concepts are built around the  users could find and buy via the search engine. For targeted marketing the best conversion rates were always those aimed at specific currently-hot items (eg specific Nike trainers, seasonal Tory Burch bags, Xbox One etc.)

 

Although this was a splash screen that animated left to right, a modified version also used for trade-show booth backdrops.

 
 
 
 
 

Publishing

 

The original experience was already fairly popular, but thanks to Apple featuring the app in 17 lists over its lifetime, the product was on the brink of the 'hockey stick'. App held consistently in the top 10-20 in Lifestyle and held in top 10 on Shopping once the section was separated with the arrival of iOS8.

 

App Store updates were pushed out bi-weekly with bugfixes and feature releases after two weeks of QA.

Over two million users installed the app on their devices.

 
 
 

Statistical Analysis

 

With statistics the first things to solve for are the rapid declines.

Through our own detailed in-app analytics we were able to make informed decisions on what blocks the users from proceeding. Although this completely misses the conditional component, we could identify features that needed to be reworked.

For example, initially the marketing team wanted to incentivize the actions users can perform on separate interstitial screens before they actually land in the experience. While this made sense in concept, the falloff rate for users was as much as half; users found this too intimidating even though the concepts were very simple and nothing they wouldn't know how to do from using the system.

A more pragmatic approach was the inclusion of in-experience coachmarks that fire off at appropriate times to guide the user (and don't take the user out of context of the interface they're trying to familiarize themselves with it).

Marketing also wanted a drive traffic with a binary action words- Browse and Search. Statistics quickly showed that users don't understand the concept, particularly because Browse required different kind of interaction than than searching. Once we simplified that down to toggling between different store feed containers, no extreme drop-offs or unforeseen statistics appeared.

 
 
 

Speccing

 

General and feature specific specs made to be as easy to understand as possible, never including already-specced items (and rather referring to the appropriate page) and being as concise as possible.

 

 

 
 

In Conclusion

 

The 2014 IBM benchmark report suggested that 51% of users feel more comfortable shopping on their own devices, which is where the surge of use came from.
In retrospect, to make the product succeed and graduate from MVP to a rapid expansion in audience, the experience could have been more user centric.

 

The company was acquired by Facebook, who retains the brand and all product which may be resurrected at a later time. Until then, the backend algorithms are used for the very thing that drove Shopping's paid conversions - the ad suggestion system.

 

Last but not least, a few well-earned, digestible lessons...