Feedback and Evaluation

I consider the challenge of Databoard.me to be two-fold: half about design and half about development, and I place equal weight on both sides. From the design perspective, there are obviously still a lot of refinements that could take place. In a product review session with the founder of mobile app Piictu, he expressed very positive impression on the overall design and flow, but pointed out that the project is not focused enough. “It is dangerous trying to solve every problem,” he said, “you should figure out that out of the many possible user cases, which one is the most likely to gain traction.” Of course he is coming from a purely startup point of view, but it is that I have been concerned ever since the project started. I intentionally wanted it to be open ended and wanted the users to explore and experiment. But the situation could very possibly be like this: if I never get enough traction, then I will not even be able to reach the niche group that I am targeting for. Perhaps it is indeed necessary to start out with a clearer focus and give users explicit directions on what they should do.

On another occasion, after a presentation, a critic with venture capital background raised the issue that the project lacked a marketing mechanism that can build up the traction in the early adopters. He suggested that I could integrate social sharing features more deeply into the system - since users have already authorized those applications, it would be a waste not to do so. And when they share individual visualizations, there should be a button below says “Get yours at Databoard.me” to drive traffic to the site. This is a fantastic suggestion and I have added it to the next steps list on Trello.

Addition to these feedbacks from critics and testers, I have planned another few features for the near future: 1. More graph modules and more customizable options for each module; 2. The ability to embed Databoard.me visualizations on external websites; 3. An API to allow other developers to write their own modules for Databoard.me.

From the technical perspective, I feel that I have been successful in implementing the front-end user experience as designed. The only drawback being that my back-end skills is not as strong as front-end, so the database code is not as efficient as it could be. The next step would be to implement a crawling system that efficiently sync user data without causing noticeable delay.

View high resolution

Data Visualization Interface

View high resolution

User Experience Flow

The Datasexuals

In a blog post titled “Meet the Urban Datasexual”, author Dominic Basulto identifies a new class of urban demography which he labels the “datasexuals”:

The datasexual looks a lot like you and me, but what’s different is their preoccupation with personal data. They are relentlessly digital, they obsessively record everything about their personal lives, and they think that data is sexy. In fact, the bigger the data, the sexier it becomes. Their lives - from a data perspective, at least - are perfectly groomed.

This seems to be exactly the target audience I am aiming for, and judging from the response to this article on Google+, I’m glad that such a user group does exist.

Design Research: Data Visualization Methodology

Humanized Data

For the first prototype I made, I had a few mockup visualizations that contained data like number of times mentioned on Twitter, number of different activities on Facebook and number of checkins at different places on Foursquare. The feedback I got when I presented it to the thesis studio was that, obviously, the examples were not close enough to life to show anything meaningful. A typical reactions was: now I know how many times I got mentioned on Twitter, so what? I realized that one’s impact/reach on social networks were only of interest to those who use social media as a self-marketing tool, which are in fact a relatively small proportion of my target audience. Databoard.me is trying to help people understand themselves better, so the data has to be more relevant to their life.

What is humanized data? Felton’s annual report can provide some good examples. There are some typical personal data that are tracked in his reports: work/off days, photos taken, places traveled, food eaten, drinks drunk, books read, music listened to, and physical exercise such as miles run. From these data, he then discovered some interesting facts such as the chance of having dinner at home on an average day, or the percentage of days he’s on a trip.

Of course, a lot of such data can only be obtained through almost compulsive data recording habits. It sure can be made easier if we can use existing apps. For example, for a heavy foursquare user, we can almost instantly generate some data reports on his eating/shopping habits. But generalized data just won’t cut it for the level of details we are aiming for. So in the end I decided I have to give weight to custom data recording.

Tell a Story

If we take Felton as an example again, he experimented with giving his annual reports a theme in the later issues. In his 2010/2011 Biennial Report, he quoted Philip K. Dick: “a person’s authentic nature is a series of shifting, variegated planes that establish themselves as he relates to different people; it is created by and appears within the framework of his interpersonal relationships.” And he claimed that the 2010/2011 report ”explores this notion by overlapping facets of Nicholas’ behavior to visualize how his personality varies based on location and company.”

Picking a Chart

Since the users are totally untrained, there must be an intuitive guide process that takes them through picking the right chart the represents the meaning in their data. I found this very useful chart on Extreme Presentation:

View high resolution

Screenshot of Databoard.me prototype

Tech Research: Stack Technology Choices

As the only one behind Databoard.me, I also have to make technical decision when hacking out a prototype for the product. Obviously, everything I use is open source. For hosting, I have a $15/month VPS (Virtual Private Server) which basically allows me to deploy any stack. Although it’s a very small machine, it is enough for prototype purposes. Later on I will probably switch to a cloud service like Amazon Web Services or Heroku.

Server Side ———————-

Language: javascript (Node.js)

I love javascript, what else can I say? I want to write my whole app with it. That’s why I chose Node.js as my backend tech choice. Node.js also has advantages in real-time applications but so far I don’t see the need of incorporating such features in Databoard.me yet.

Framework: Express.js

Express.js is inspired by Sinatra, the minimal style Ruby framework. And that’s what I like about it, minimal, clean, easy to work with and maintain. And since Databoard.me is going to be a client-heavy app, the server side is essentially just a data API, so a lightweight framework serves really well.

Database: MongoDB

This is probably the only place where I spent a lot of time deciding. Mongo seems to be natural fit because of its close relationship to javascript and JSON, but it lacks a decent administration interface. MySQL is really easy to work with especially with phpmyadmin, but the mapping between tables and JSON can be somewhat tricky to handle. Eventually I decided to go with Mongo because all the data I will be working with on the client side is heavy JSON, the advantage of being able to store schemeless JSON/BSON directly will be huge.

Client Side ———————-

Code organization: Require.js

Require.js is great for organizing large projects. It allows developers to write javascript code as modules with explicit dependencies on each other, and handles the loading and dependency management automatically. The best part is it comes with an optimization tool that concatenates and minifies all the code into one single file.

MVC structure: Backbone.js

Backbone.js is the go-to solution for client side MVC. It provides a lightweight yet solid infrastructure to build javascript applications on. Since I want Databoard.me to feel more like an app with smooth interactions rather than a website of pages, Backbone.js plays a critical role in the entire project.

DOM manipulation: jQuery

jQuery is almost the standard for dom manipulation libraries - do I need to explain more?

Data Visualization: D3.js

D3.js is another critical component of the client side: it is a powerful library that generates beautiful data-driven SVG graphics. Because it is SVG, it is very compatible across browsers and animates very smoothly. It even works in IE. The author behind D3 works for Square, and they use the library in their transaction analytics services.

Fluid Layout: Isotope

Isotope is an amazing jQuery plugin that provides a very powerful layout system for boxes of content. Databoard.me is essentially a board of cards with varying sizes, and Isotope solves the layout problem.

Precedents and Competitors

Data visualization has long been used in research to gain better insight on large data sets. It is therefore, not surprisingly, an essential part of any analytics-centric business intelligence company. Aside from this type of usage, in recent years there have been a growing number of startups built around data visualizations for the mass public. Since Databoard.me is one of them, in this post I will go over some of the other data visualization startups and discuss if they will potentially compete with Databoard.me.

Daytum

Daytum is probably the most similar to Databoard.me among all the existing data visualization startups. It allows users to record arbitrary data, then turn them into charts contained in ‘displays’, which can then be placed on users’ pages. It is conceived by Nicholas Felton, the designer behind those famous personal annual reports. He is also the designer behind Facebook’s timeline feature. I must admit that Databoard.me share an almost identical view on personal data with Daytum: that it can speak a lot about your identity and help you understand yourself better. The approach is also very similar, but on the execution level, there are a lot of differences to be noted:

1. Daytum’s visualization feature is very limited. It is using the Google charts API, thus they don’t have full control over the visualizations. They don’t even have multiple colors in visualizations, only different shades of the same color. This makes the visualizations much less powerful and expressive than they could be. On contrary, Databoard.me’s visualizations are powered by the D3 library, which can produce amazing, animated and interactive visualizations. The templates are also modularized so that I can easily roll out new templates.

2. Personally I think Daytum’s data-entering interface is still not simple enough, creating a motivation barrier for users to enter more data. I believe it can be easier - fewer clicks and less time spent figuring out how to use the system can result in more data recorded.

3. Daytum does not have social features, which is essential to Databoard.me. In this sense, Databoard.me is combining part of the Visual.ly features with Daytum.

4. Daytum provides little to none guidance in how to use it to its full potential. This is something Databoard.me is prepared to address by providing a visualization creation guide which is designed to be as intuitive as possible, plus examples and social exploring features.

Visual.ly

In the big picture, Visual.ly aims to build a “platform (that) democratizes the creation and sharing of visualizations, making it possible for everyone to participate.” It started out with only the upload, share and explore functionalities, but that already caught quite some attention and has secured considerable traction. It just recently rolled out the “Create” functionality but so far it’s quite limited - all a user can do is link up a Facebook account or enter Twitter handles, and the visualizations are all pre-designed templates with little customization.

Although Databoard.me is also trying to become a platform for data visualization, but there are fundamental differences in between. First, Visual.ly is so far still mostly a place for users to upload data visualizations they found elsewhere. It’s not their data, not their story, not personal at all. Second, judging from Visual.ly’s list of big-name partners (The Atlantic, BuzzFeed, AskMen, CNNMoney, GOOD, The Economist, NASDAQ, Skype, National Geographic, Wall Street Journal), their business model seems to point to business-oriented direction (which is still kind of blurry to me). On the other hand, Databoard.me is a straightforward product built to provide value to revenue generating individual users.

Tictrac

TicTrac is also a personal data hub which integrates a great number of third party data APIs. It seems to provide very well designed personal data aggregation and presentation, and will definitely be (at least partially) competing with Databoard.me. However since it’s still in closed beta, I can’t get a hand on it to see if there are more overlapping areas.

Vizualize.me

Vizualize.me has a very simple concept: create better resumes by automatically visualizing your LinkedIn data. Visualized resume is a great concept, but personally I think there are two problems with this company: 1. They are too dependent on and limited by a single data source. I’m not sure if they have any plans on dealing with that; 2. The quality of the produced visualizations are somewhat mediocre and repetitive. I hope they are working on improving and providing more templates, otherwise I don’t see it going too far.

Geckoboard

Geckoboard focuses on providing a real time analytics board for companies or websites. What resembles Databoard.me here is its presentation and aesthetics - a board that shows your visualized data. Obviously they target completely different markets and serve different purposes, but one important thing I learnt from them is looking at the components of the board as widgets - they should be modularized so that they can be added, removed and developed with ease.

ManyEyes

This is not a startup, but is a very important precedence so I want to mention it here too. It is one of the first website that provides the users the tools to create visualizations from data through relatively simple interactions. It guides the user through a step by step process to point them to the right graph to use. It is dated and of course has its drawbacks, which is the reason why I am building Databoard.me, but its methodology in creating visualizations are extremely helpful.

How Databorad.me Delivers Its Core Value

In the business model I have outlined the core value proposition of Databoard.me: helping the user get better self understanding, and express their unique identity. Now the question is, how do we achieve that through an application? This can be further broken down into smaller questions, as discussed below.

1. What kind of visualizations can help people understand themselves better?

(The answer to this question will be further elaborated in another post on the research of data visualization methodologies)

• Humanized data.
  
  Without a connection to humanity, the data is just boring numbers. We want to look at things that are closely associated to our lives: food, drinks, books, movies, music, spending, travel, work hours, physical exercise… the system must be able to facilitate the recording of such data, or manage to mine it from existing APIs.
   
• Reveal something by putting the data in context.
  
  We can often discover interesting things when we juxtaposition two data sources together. For example, if we look at miles run per day together with daily weather, we will discover some kind of correlation in there. A more effective example would be looking at mood in context of weather, work hours or food eaten. This kind of visualizations will be the key that makes the whole thing meaningful.
   

2. How can users learn to do this?

• The interface must be easy to use and guide the users intuitively.
  
  This makes the design and implementation of the interface a key activity for the success of the project. Tracking data is a repetitive and sometimes cumbersome task, so the friction of performing an action must be lowered as much as possible to make users more likely to do it. Currently, the plan is to build a step by step guide with very simple, easy to grasp questions and instant preview.
   
• There must be examples.
  
  No matter how easy the interface is to use, it can still be hard to come up with good strategy to use the data. That’s why they need examples like the ones mentioned in (1) to inspire them, to show them what can be done with the system.
   
• Users must be able to share with and learn from each other.
  
  Once users start creating their own visualizations, there sure will be a lot of creativity floating around. I think it would be greatly helpful if users can easily discover and share great visualization ideas, making the app a platform where they can learn from and innovate upon each other. I plan to make it really simple to apply others’ visualizations to one’s own board. For example, you find that someone is using Databoard.me to track their weight change from day to day. You think that’s a good idea, and you can hit “I want this on my board too”, an identical card will be created for you, ready to receive data input from you.

3. How can users express their identity through this?

Each user will have a public board where he can place visualizations that he would like to show to other people. These can usually be about some sort of bragging rights, or something like a public status update. It also shows how they think their data represent themselves.

The Business Model

A brief breakdown of the business model for Databoard.me (subject to change):

1. Customer Segments
   
   Databoard.me targets a niche market (as of today) of individual internet users who are interested in understanding and expressing themselves better through the collection and visualization of their personal data. A typical customer would be a curious, organized and tech-savvy person. In the future when the importance of personal data becomes more universally acknowledged, the market will potentially expand to cover a much larger audience.
   
   More detailed segmentation is described in item D and E of section 2 below.
   
   
2. Value Proposition
   
   The reasoning:
   A) People produce a lot of data on existing internet-based services.
   B) These data have meaning that either helps self-understanding, or can express the uniqueness of identity.
   C) A service that makes it easier for them to enter, track, visualize and communicate their personal data will help the customers gain the values proposed in B. It would also help if the service visualizes the data with a pleasing design aesthetics.
   D) There are plenty of people who like better self-understanding or expressing their unique identity. They become likely customers.
   E) Among the customers described in D, some are willing to actively enter additional data in order to get better access to values proposed in D. They become key customers.
    
3. Channels
   
   Awareness and Evaluation phases are (planned to be) covered by a combination of the following: user-to-user reference, tech blog appearance and social sharing reference.
   
   The main value delivery channel is the service itself in the form of a desktop web app plus a mobile web app. It will also cover the purchase phase.
   
   After sales phase will be covered by on-site FAQs and a support forum.
    
4. Customer Relationship
   
   For the most part, Databoard.me will be a self-service system that allows the user to explore and experiment on their own.
   
   One core mechanics of the system of the selection of visualization modules - in addition to the default modules, new modules will be continuously rolled out to address customer retention.
   
   The ability to share a part of their creations also makes the service a platform where users can interact and form a community.
    
5. Revenue Streams
   
   The service will operate on a freemium model: for basic usage, the account is totally free. This is also to lower the barrier for initial user acquisition. However, free accounts only have access to the most basic selection of visualization modules. They can choose to individually purchase modules that they like (one-time payment), or they can choose to upgrade to premium account, which charges a fixed monthly subscription fee. If they decide to upgrade after they have already purchased some premium modules, previous charges can be deducted from subscription bills after the upgrade.
   
   One potential source of revenue is advertisement or sponsored recommendation based on data-mining. This is a debatable option and I am deferring it to a later stage, to be discussed when there are actual revenue and data to look at.
    
6. Key Resources
   
   The key resources will be dominantly people. To be more specific, developers and designers. Hosting infrastructure can also become a key resource, but only when the service has acquired a large enough user base. There are also a number of cloud services that can be leveraged to turn actual hosting infrastructure requirements into the need for several dedicated system engineers.
    
7. Key Activities
   
   The key activities in the initial stage is production of the app. More specifically, we need to design the core mechanics of modularized data visualization and a easy to use data entering interface. Then we need to implement the system with a database backend.
   
   When we enter the user retention stage, we will then focus on designing and deploying new visualization modules, integrating more third-party APIs and data sources, and promoting user interactions on the sharing platform.
    
8. Key Partnerships
   
   The most important partners are existing services that already possess or track user data. Most of such services will likely provide an API to work with, but sometimes we might have special needs that cannot be fulfilled through a regular API, and we will need to work with these service providers to get access to those data.
    
9. Cost Structure
   
   There are two major parts in the cost structure: human resource salaries and hosting costs. For the proposed system so far, the team will most likely remain small and compact in order to move fast. It is only me for now but it probably will not exceed five. Hosting costs, thanks to cloud hosting services, will be trivial in the development and testing stage, and can be scaled proportionally along with user and revenue growth.