Wednesday, 22 April 2015

The Harper Lee Experiment

Harper Lee, author of 'To Kill a Mockingbird' wrote the book after a friend gifted her 1 years salary to write whatever she wanted. No-one would argue that that time or money was misspent. It made me think however, when I randomly remembered it earlier in the day about how with some changed variables it could apply to a developer. Then it became a bit of a thought experiment.

Let's say you're random developer with some experience. Not necessarily lots, but enough to be comfortable. On new years eve, a friend (whose reasoning skills you'd likely and understandably question at this point) offers you 1 years salary to create whatever you like under the understanding that after December 31st that this creation would be your livelihood. Through some sort of magic your day job is a thing of the past and you are free to do as you please.

We'll also say as a matter of co-incidence you happen to have been sitting on an idea for some software that a blue box inhabiting man has informed you is a guaranteed hit. You have your idea and 365 days, where do you start?

Getting Started
Well, how much time do you have? That's a good start.

52 weeks - 4 weeks because your friend insisted you take a break.
48 weeks of 5 day weeks, to generalise working for a comfortable 7.5 hours a day.

That gives us a round 1,800 hours of time. It's feeling a bit tight already isn't it? So we already have an idea about the scope of what it might be best to undertake. For example, an application for uploading pictures with captions for your friends to see, that's monetised through some sort of advertising (The requirements stated we have to live off it, remember?).

From there, we can extrapolate a rough feature set, there's going to be some authentication, storage, relationships between user accounts and a phone based UI for your platform of choice. Your at this point unbelievable friend also happens to be a marketing and UI expert and will complete these elements of your work when appropriate to the best of their ability and free of charge, so you can just focus on the software. What's next?

Considerations
How are you going to manage your workload, Kanban perhaps?
Will you start with the back-end first, the front-end, or develop them in a scenario based format in parallel?
Is December 31st the release day or is that the day you expect to be running smoothly?
Is there beta testing? How long for?
How will you handle security?
What kind of testing do you intend to undertake with your already stretched resources?
How will you know that the service is running smoothly after launch?

You might have a rough plan for what is to be done now to successfully unveil your masterpiece to the world. How does it stand against a few test cases?

Test Cases
1) 3 days after launch your primary chosen cloud platform drops off because of a time-related bug and doesn't return for a day, as a result of this are you missing out on ad revenue or is there a contingency?
2) You decide to take a break for a week, as the sole developer you ask a friend to take the keys to your newly up and running service and keep an eye on it...
    a) Something vague goes wrong and users are sometimes failing to upload photos, can your friend solve it easily and how long did it take them to find out?
    b) That massive new feature you checked in just before you left, it turns out, doesn't work, did it make it through to production and can your friend fix it, and do they even know there's a problem?
3) Generic tech review site X found your app and wrote a front page article about how world changing it is and your user-base explodes like you didn't know was possible, will it scale out succesfully?
4) Someone got into your user database, how bad is the damage? Are they looking at password 'Hunter2' or did you go to some lengths to secure it.
5) It turns out that after your app has been running for 3 minutes it runs out of memory, did you catch it during testing and if not, how long will it take to get the bug-fix to the user?

Outcome
It's quite easy to underestimate how much time/effort it takes to get something relatively simple to be reliable and the costs involved in doing so. A year seems like a long time until you realise you probably have less than 6 months of product development time when you factor in testing, security, stability, etc. You could just not test it I suppose, and hope that if it falls down your friend isn't too upset about their failed investment...