Back to case studies Panda

Multi-user indoor positioning mobile app for social discovery.

http://joinpanda.com/
App Store Google Play
Panda

Nobody is a stranger

Remember yourself sitting in a crowded restaurant, waiting for your order. It takes ages. Trying to tame your boredom, you start stealing glances at people from neighbouring tables. That gentleman with glasses is typing something really quickly on his phone. And that girl drinking milkshake in the corner has just glimpsed at you and immediately looked down with a slight smile. Who are those people? What are they thinking about at this very moment? What brought them here?

Imagine you have a map of the premises that shows the location of every other person – you can see their names, interests, statuses and also instantly begin a conversation. This is exactly the vision of mobile app that Hares Youssef, the founder of Panda, had in his mind when he came to YouTeam.

Panda

Technology constraints

Indoor positioning mobile app is the next Holy Grail of technological business. When Google has launched its indoor maps in 2011, many tech pundits and sages proclaimed the dawn of the next information reality – full of microlocation-savvy services and IoT devices.

Although outsourcing mobile app development for four years, the indoor navigation services have not yet become omnipresent. The reason is that drawing the floor map is the easy part. Far more challenging is to determine your location on this floor. Various technologies have been utilised in an attempt to clear this height: GPS, GSM data, WiFi, Bluetooth, ultrasonic sound, motion tracking cameras – you name it. However, when we started analysing each of those solutions for Panda, it turned out that they are either prohibitively expensive or produce a gargantuan amount of error. At best, the positioning precision was up to two metres, which is OK if you want to navigate your way to a check-in desk in an airport terminal, but far too vague if your task is to distinguish people at different tables in a crowded restaurant. LTE Direct has the potential to be a game-changer here, assuming it will ever be released (which is so far announced for 2016).

ibeacon

After screening all available solutions, we have chosen the relatively cheap and accurate Bluetooth Low Energy. The key perk here is the existence of a purposed device – the iBeacon. Since Apple has started licensing this technology, a number of devices from different producers have appeared on the market. After having experimented with several models, we finally stopped at Estimote – a Krakow-based Y Combinator start-up that advertises their indoor navigation SDK as a separate solution. The thing is that iBeacon is de facto not designed to track your location in premises. In a standard scenario, it simply detects that you have entered the shop where the beacon is installed and sends you a push notification with the latest offers and discounts. Our case is far more complicated.

We started with a number of tests using a tri-lateral and tri-angular data from several beacons, measuring their relative signal strength. This video, featuring Yours Truly, shows one of the early micro-location tracking prototypes with six beacons installed in the room:

As you can see, everything works pretty well – with a single user in an empty room ☺. The problem is that with Panda this is never the case. Panda’s environment is by definition full of other people – with their bodies full of blood – which in turn is full of iron-containing haemoglobin. This iron blocks or diffuses the weak Bluetooth radio-wave signal – which turns the resulting coordinates data into a total mess. Now imagine that you have to track not just one person, but 20 or 30 – all in real-time!

The case for a Minimum Viable Product

Quickly, it became clear that the current state of technology makes it impossible to build a full-fledged indoor-positioning platform. Therefore, we had to figure out what level of technological solution would be sufficient for the particular Panda application. After several business model workshops with the client, we have finished with three promising applications:

Panda-Applications

The founder has decided to start with Panda Club, since the market here is the closest to his initial vision. The social capital also has played a role – as Hares happened to know the owners of a couple of club and restaurant chains. Then, the segment has been narrowed down to nightclubs only, since their business aims at solving the same problem as our application: the social discovery.

I have paid several visits to different nightclubs to study the behaviour of the audience. I must say this has been quite a fresh experience for me – as I haven’t been in a nightclub for a good decade. It turned out that the club goers’ behaviour follows a simple and permanent pattern:

  • Male visitors spend ¾ of the time sitting at the tables talking, drinking and watching the dancing crowd. Mostly, they come in same-sex groups. Obviously, they are the most active smartphone users.
  • Female visitors also prefer to come in same-sex groups, but they spend only up to 1/3 of their time sitting by the bar or at the table. Instead, they go dancing at the first opportunity. As you can imagine, the smartphones are rather rarely used on the dance floor – and mostly for selfies.
  • People do not move around the premises too much. They usually adhere to 2-3 locations: their own table or bar, the dance-floor, or the table of their newfound friends.

Those observations suggest several important aspects of the product’s functionality. This allows us to arrive to the MVP scope within the present technological constraints:

  1. Our platform does not need to track the real-time movements of each user. It is enough to show the current location in general. This means we do not need the trilateration, but we can instead just split the premises into functional zones (bar, dance floor, VIP-zone, etc.) with one or several beacons in each. It will show the users in their corresponding zone.
  2. Table tops are perfect for additional beacons if we want high precision, i.e. to set every table as a separate zone. Smartphones are mostly used when people are at their tables. Placing beacons there minimizes the chance of a shielding obstacle between the beacon and the user’s device. It also decreases the distance between them which improves the accuracy. As a result, we’ve decided to design special menu holders with beacons “secretly” installed into their bases.
  3. We need a simple tool that will allow guys to demonstrate their interest in a girl and make her return to the conversation-starting app. For this we’ve designed a special “Wink” sticker as a messege, an analogue to the Facebook “Poke”. The app will then use push notifications and those “Winks” will drive retention.
Presentation-Panda

The Prototype

There are many approaches to the MVP design. Some industry experts advocate the so-called “Lo-Fi MVP” approach, which means that you don’t even need a product to get feedback from you target audience: a simple video or look-and-feel mock-up can be enough.

However, the truth is that MVP is nothing but a learning tool that allows you to test the most risky assumptions of your business model. Eventually, it will pave the way to a final product tailored to the true needs of your customers. Provided that all you need is an answer to the simple “go or no go” question, the Lo-fi MVP may be sufficient for you. But in reality, you often need more than that.

At YouTeam, we have developed the following rule of thumb that helps carving a balanced MVP scope:

MVP must have one feature for every KPI of your business model.

By KPI I mean simply the well-known “Pirate Metrics” by Dave McClure: AARRR!
Here is how those were translated into the features of Panda Club App:

  • Acquisition: Single-tap login with Facebook account. User profile data (name, photos, friend list is imported automatically. Also Facebook “common likes” has transformed into “common interests”.
  • Activation: We believe that a user has activated the app after he/she has accessed the map, picked another user and engaged in a chat. For the sake of a truly intuitive UX, I did not want to use any tutorial here. Immediately after the login, the user will receive a welcome message from the club they are in at that moment. After they authorise the app to share their location, they will see a beaming map button on the top and the list of all present users in the club sorted by gender.
  • Retention: Besides the above mentioned push notifications (which is pretty standard for the messenger apps), the user will be notified every time they enter another Panda-featured place.
  • Referral: A “check-in with Facebook” button on the club screen and club map allows the user to check-in using Facebook API. In this way, the information of their current location and the link to Panda website will appear on their Facebook timeline. They can also send their friends personalised invitations to Panda via Facebook messenger.
  • Revenue: We have chosen subscription as a revenue model – with nightclubs being the paying customers. Obviously, an improved customer experience is not enough to make this deal attractive . This app must either drive the additional audience to the subscribed clubs or generate certain upsale. Ideally – both. For the former purpose, we have included a street map with all Panda places geotagged. For the latter, the functionality that enables users to place orders via the app is currently under development. This features the EPOS API integration as well as a special “Drink?” sticker.

The platform of choice was, of course, iPhone, as the software development for iOS takes on average 20% less time than for Android.

Researching the Pandas

If you have ever experienced building an innovative product, you know very well that MVP is just the first part of the story (from my own experience – usually no more than 1/3 of the story). The real fun begins when you start interacting with real customers. At this stage, the question of cost and time optimisation is even more acute.

This is why we recommend launching the MVP on a small model market.
For Panda, we’ve chosen a large two-level club in our development team’s headquarters, Lviv, Ukraine as our experimental lab. A total of 48 beacons had been installed and a team of 10 promoters recruited. Six of those (three boys and three girls) interacted with the audience, literally grabbing people’s smartphones and installing the Panda app, while four others acted as “decoy-ducks” – sending “winks” and texts to all the newly-registered. Google Analytics was used to gather the quantitative data, while UserVoice and direct messaging were used for the qualitative feedback.

The key findings and implications on the functionality are:

  • A large share of the young audience do not use Facebook. Yes, you didn’t mishear that, Mark… We’re adding a sign-up using a phone number.
  • The indoor map paradigm is new for the users – so they often get lost with it. We’re changing user experience to more Tinder-like: starting with large photos of the users of opposite sex and a map button on the top to show their current location.
  • The percentage of Android-powered smartphones turned out to be much larger than we expected. We are developing an Android app – as this is a social platform which values an exponential function of the users’ amount. We cannot afford losing accounts due to the platform’s limitations.
  • The most active segment constitutes of males between 20 and 25 y.o. Women are more difficult to engage – they need additional incentives to stay registered with the app. Gamification with achievement badges can be one solution here – with the opportunity to exchange badges with drinks at the bar.

The open beta tests and corresponding app tweaks will continue until the proper level of KPI is reached – most importantly until the growth becomes organic driven by high referral rates.
Then, Panda Club will be launched at at target market, which is, of course, nowhere else but in London.

panda-splitscreen
Design by
QubStudio
Development by
devabit
Team
  • UX/UI engineer
  • 2 iOs developers
  • Back-end developer
  • Android developer
  • QA
  • Product manager
  • Research manager + 10 promoters
Time:
  • Business model and Product Vision: 3 weeks
  • Micro-location prototype: 2 weeks
  • iOs app and back-end: 12 weeks
  • Android app: 8 weeks
  • Open Beta testing: 8 weeks
Technologies:
  • java
  • ibeacon
  • c#
  • .NET
  • estimote
  • objective-c
adware anti