Algorithms for generating new ideas for startups

I used to rely on spontaneous inspiration to generate most of my ideas, normally while doing some activity where my mind was relatively free to wander like showering or going for a long drive. Recently though I’ve started to get frustrated at the rather unpredictable rate at which I amass ideas and have started developing more systematic approaches for generating new business ideas.

In many numerical disciplines, there are often functions that we want to sample from but where doing so is difficult. So in general, we cheat. We sample from a simpler approximation and then apply some form of correction afterwards. Consider these different approaches as approximations. For each technique i’ll explain how it works and then go through a few examples of applying it. Feel free to add more examples in the comments and i’ll update the article with them where possible.

Servicing trends

A fairly simple technique to kick us off.

  1.  Identify a list of technologies, platforms and services on an upward trend
  2. For each item in the list:
    1. Find an existing concept that has similar characteristics
    2. Identify services and complementary products
    3. Consider the application of these to the trending item incorporating any new aspects it brings

Let’s look at a few existing servies as examples first.


If we look at the trends for around 2008, it’s fairly easy to see Git trending.

Looking at similar, incumbent version control systems there were numerous companies offering hosting for open source projects using Subversion and CVS (Sourceforge and Google Code being two notable ones) and commercial repository hosting for private projects. Github successfully offered both.


Another good example is Heroku, which originally catered for Ruby developers. While other languages like perl and PHP were often supported on shared hosting environments, Ruby (and specifically Rack) projects required specialist or dedicated environments.

Potentially new ideas

It’s fairly easy to rationalise any successful idea retroactively so let’s try applying this method to generate ideas.

Let’s follow the algorithm we defined earlier. First we need a list of things that are trending. Google Trends as we used to look at the VCS trends earlier is one option. A few other useful sources are Crunchbase and Google Insights.

Step one, we need a list of trending items. We’ll start with: Hadoop, MongoDB, Stanford and MIT’s online university courses.

So let’s start with Hadoop and its related projects. Google Trends shows usage is trending at a fairly rapid rate.

Now for step two. We need to identify an existing concept that has similar characteristics. For Hadoop, probably the closest would be databases.

Step three. We now need to find services and complementary products that already exist for traditional databases. There are two good tricks here. The first is relatively simple, use a search engine to search for related items and looking for advertisements or heavily search engine optimised product pages. For example, searching Google and Bing for ‘mysql’, ‘postgres’ and ‘oracle’ coupled with ‘tools’, ‘services’. The second involves finding a few advertising-heavy sites dedicated to the particular thing you’re looking for and studying the contextually-relevant advertising. Places like eHow and wikiHow are great for this.

I spent twenty minutes doing both of the above and here’s a few things that jumped out:

- Consultancy (Performance tuning, version migration and platform migration seem to be popular)
- Automated performance monitoring/analysis
- Auto-scaling cloud hosting
- Cloud backup
- Query editor and database explorer desktop GUIs

Step four. Consider each of these services or products as a potential starting point for a business involving Hadoop while, if advantageous, incorporating any unique characteristics that Hadoop brings.

Many of these could equally be applied to MongoDB, another trending technology.

How about the recent trend of large online university classes kickstarted by Stanford and now continued by Udacity, Coursera and MITx? We’re probably very early on in the trend but that’s normally a good place to be.

How can we service these trends? Cranking the handle with the techniques from earlier and a little brainstorming:

- One-on-one tutoring for popular courses (or a marketplace for them)
- eBooks of questions and worked solutions{{ footnote(“you could even provide the questions for free but charge for the worked solutions or upsell having your solutions marked by a human with feedback.”) }}
- Specialised recruitment events/platform (‘eMilkRound‘)
- Blog/social media widgets showing your completed courses and current studying

I’ve not tried to filter any of these ideas, there may already been well entrenched competitors and they probably all need a little further development to find a good fit in the market (a little hill climbing).

Well, that’s all for this technique. If you’ve got good examples, please post a comment!

Leave a Reply