Data mining for startups
Adding another on the business process to your startup’s already list of things to do will seem like a burden at least initially. Typically processes are added to the list of things to do on an agile basis, it goes something like this:
- Something has happened
- Why did this happen?
- Let’s fix this
- Put a process around it so we can be alerted early or prevent it from happening altogether
With each process added there is an overhead so we want to minimise this and maximise the benefits of the process.
Why add data mining so early?
For a typical (tech) startup the process of bringing a product or service market involves developing, marketing and selling amongst other things. At some point along the journey the startup will look at ways to improve, optimise, reduce. My thoughts are that startups should think more about how data is collected and analysed with a specific on improving key metrics that underpins the business.
Reactivity or Proactivity?
Bake data-driven decision making into your startup as early as possible
If this is baked in as early as possible then you become a data-driven company from the get-go. Having an easily accessible 360 view from the start will give you an incredible head start over your reactive competition. If you only look deep down into your data when a problem arises then you’re missing out. The key here where possible we want to as proactive as possible, if our data is telling us that trends are moving in a certain direction then we can be better prepared. In essence, we eliminate guessing as early as possible and back things up with data and analysis.
Data mining = the questions you ask
As a starting point you should ask questions of data and use that as a guide.
- How do we improve X?
- Is this important?
As an exmaple If we look at web visitors or app users, our goal is generally to get as many users as possible on the site then perform some action whethter it’s to read, watch or perform some other action like buy a product. Our starting point here should be to know what the bottom line is. Number of sign ups? Number of products bought? We then should have a method to gather this data and look at the factors that influence the numbers. We then need to ask questions regarding how these factors influence the numbers and look at methods to nudge these factors in the correct direction.
Mining for gems in the dirt:
Startups may not have time & resources to mine everything deeply but startups should be mining those data points that affect the bottom line of the company whether that’s a number of user, churn, or product price points. Every company should on proactive data mining as it increases transparency and identifies problems early on. You don’t need to have an exhaustive list of data to mine but as a starting point you should at least mine those data that directly your bottom line.
Your bottom line. your customers. Your next decision, shooting in the dark or an informed one?
Data mining is critical as it gives us a way to understand what’s going on under the hood, in addition and as a result of mining data:
* We get better & more measurable decisions being made
* We are more likely to address key problems as we can clearly see important correlations and causations And finally
* We have traceability in the decisions made
The importance of data mining for startups Simply put, data mining is important for startups as it gives startups contextualised information from which better decisions can be made. While it’s important for all businesses it is even more so for startups as resources are more often than not at a premium. Having better information at hand will result in better decisions being made with less effort. Automation that helps in gathering and analysing data is a big help in this. It could be assumed that startups, being smaller businesses, have smaller amounts of data than larger businesses and therefore an ad hoc approach to mining will do, while this is true it also offers a false sense of security as even in a one person business patterns in data can easily be overlooked and misinterpreted when dealt with manually. Having some kind of automated data mining processes is therefore necessary.