Pitfalls to Avoid When Starting out with Analytics

You start with good intentions.  But the track record is that many analytics initiatives are often unsatisfying and sometimes abandoned entirely.  All while you keep hearing about the “successes” that others are having, and how they’re going to steal away every profitable customer you have.

The following six pitfalls can kill or seriously impair a project:

  1. You don’t have a clear set of goals (or even early goals) to achieve from Analytics.  “Analytics” is very big, and as a term permits the speaker to mean many different things. Be clear on what your goal is, and what you want to do first. Is it sales? Is it cost control? Is it risk reduction? Marketing? Must it start with “big data?”  There is a lot to do regardless of where you start, so having a clear goal is a really good starting point.  “Analytics,” as we use the term, means predictive analytics – digging deep into the data, learning the patterns that typically indicate something you care about, and then figuring out how to drive an outcome you care about. (You'll hear this called Data Science, Data Engineering, Big Data, Business Intelligence, and many other things - for simplicity, we call it analytics.)
  2. Analytics is not reporting.  If you want “good reporting” – don’t start with “analytics” – it will slow down a reporting project. And this is more important that it may seem - shockingly many companies have very unsatisfying reporting.
  3. Your data is incomplete, sometimes wrong, and certainly messy. (Trust us on this - everyone's is.) While that issue can be problematic, a bigger one is often just understanding the data. This knowledge turns out to be sparse in many companies, and sometimes totally unavailable. The understanding you can build, though it may take time. This will lead you to needing serious data manipulation skills, and some data architecture. Some of these skills are technology and database related, some are statistical, some are mathematical, some are logical, some are programming related - this is a complex and varied skill category. But without this skill package, you really won’t be able to get much from your data. If you don’t have it, you’ll have to hire it, build it, or get it some other way. 
  4. Typically, your data is not enough. You’ll quickly need to combine your data with third party data sources. There is a research component to this – you’ll need someone who can understand external data.  Some sources are obvious (census data, often).  Others are more obscure.  Industry data, for example.  And some data that you’ll want may not even exist.
  5. Picking through the hardware, software and cloud options is often also a bit of a landmine.  Each approach has strengths and weaknesses.  Sometimes you need “big iron” to help out.  Sometimes you have a challenge that can be addressed quite elegantly with limited software investment.  Putting data in the “cloud” solves some issues, but brings a host of different challenges.  Picking badly can leave you with large bills, and a very difficult business case to make “analytics” or “insight-driven” investments pay for themselves in the first few years. 
  6. Finally, what you’re planning to do with the output of the analytics is a larger consideration, but can often be left for a little later.  For firms that are starting out, learning to understand the data, dig deep and really analyze it can lead to some simple, easily adopted improvements.  But in time, you’ll want the products of your analytics research to be integrated with your key systems to drive continued performance improvement. Taking on this step too soon is a final pitfall. 

This list is not comprehensive, but it is a good summary of some of the ways that internally launched analytics projects go awry. Another way to get started is to ask us about our Analytics in Days offer.