reinventing the wheelOnce your analysts have a clear business question to answer, do they start new analysis each time, potentially ‘reinventing the wheel‘?

After creating or leading data & analytics teams for many years, I began to notice this pattern of behaviour. What we seemed to lack was a consistent knowledge management solution or ‘corporate memory‘ that could easily spot what should be remembered.

Funnily enough, as became convinced of the need for ‘holistic customer insight‘, I found a partial answer amongst researchers.

Research teams are somewhat better at this, as it has become a more standard part of their methods.

I’ve yet to find an ideal solution, but I think it’s such an important issue for analytics & insight teams, that I’ll use this post to share my own experience.

The lack of secondary research approach for analytics

My earlier mention of somewhat better practice amongst researchers, alludes to their understanding of the need for ‘secondary research‘. Experienced research analysts/managers will be familiar with considering the potential for desk research, or searching through past research, to answer the question posed. Perhaps because of the more obvious cost of commissioning new primary research (often via paying an agency), researchers make more effort to first consider if they already have access to information to answer this new question.

But, even here, there does not appear to be any ideal or market leading knowledge management solution. Most of the teams I have worked with use an ‘in-house‘ development in Excel, interactive PowerPoint slides with hyperlinks to file structures, or intranet based ‘research libraries’. Whichever ‘end user computing’ or ‘groupware’ solution is used, it more or less equates to an easier to navigate/search library of all past research. Normally a user can search by keywords or tags, as well as through a prescribed structure of research for specific products/channels/segments etc.

Some research teams use this very effectively & also recall those visualisations/graphics/VoxPops that worked well at conveying key insights about customers. It is worth investing in this area as it can save a significant amount of research budget to remember & reuse what has been ‘learnt’ already.

However, whilst also leading data or analytics teams (increasingly within one insight department), it became obvious that such an approach did not exist for analytics. At best analysts used code libraries or templates to make coding quicker/standardised and to present results with a consistent professional look. Methodologies certainly existed for analysis at a high-level or for specific technical tasks like building predictive models, but there was no consistent approach to recording what had been learnt from past analysis.

I’ve seen similar problems at a number of my clients. Why is this? Perhaps a combination of less visible additional costs (as analysts are employed already) and the tendency of many analysts to prefer to ‘crack on‘ with the technical work, together conspire to undermine any practice of ‘secondary analytics‘.

The many potential benefits of Customer Insight Knowledge Management

Once you focus on this problem, it becomes obvious that there are many potential benefits to improving your practice in this area.

Many analytics or BI leaders will be able to tell you their own horror stories of trying to implement ‘self-serve analytics‘. These war stories are normally a combination of the classic problems/delays with data & IT projects, plus an unwillingness from business stakeholders to actually interrogate the new system themselves. All too often, after the initial enthusiasm for shiny new technology, business leaders prefer to ask an analyst than produce the report they need themselves.

So, one potential advantage of a well-managed & easily navigable ‘secondary analytics‘ store, is a chance for business users to easily find past answers to the same question or better understand the context.

But the items stored in such an ideal knowledge management solution can be wider than just final outputs (often in the form of PowerPoint presentations or single dashboards).

I have seen teams benefit from developing solutions to store & share across the team:

  • Stakeholder maps & contact details
  • Project histories & documentation
  • Past code (from SQL scripts to R/Python packages or code snippets)
  • Metadata (we’ve shared more about the importance of that previously, here I mean what’s been learnt about data items during an analysis)
  • Past data visualisations or graphics that have proved effective (sometimes converted into templates
  • Past results & recommendations for additional analysis or future tracking
  • Interim data, to be used to revisit or test hypotheses (suitably anonymised)
  • Output presentations (both short, exec style & long full documentation versions)
  • Recommendations for future action (to track acting on insights, as recommended previously)
  • Key insights, summarising up into a few short sentences, to accumulate key insights for a specific segment, channel or product

Given this diversity and the range of different workflows of methodologies used by analysts, it is perhaps not surprising that the technical solutions tried vary as well.

Where is the technology analytics teams need for this remembering?

As well as being surprised that analytics teams lack the culture of ‘secondary analytics‘, compared to the established practice of ‘secondary research‘, I’m also surprised by a technology gap. What I mean is the lack of any one ideal, killer app, type technology solution to this need from insight teams.

Although I have led & guided teams in implementing different workarounds, I’ve yet to see a complete solution that meets all requirements.

An insight, data or analytics leader looking to focus on this improvement should consider a few requirements. First of, the solution needs to cater with storing information in a wide variety of formats (from programming code to PowerPoint decks, customer videos to structured data sets, as well as the need to recognise project or ‘job bag’ structures). Next, it has to be quick & easy to store these kinds of outputs in a way that can later be retrieved. Any solution that requires detailed indexing, accurate filing in the right sub-folder, or extensive tagging, just won’t get used in practice (at least not maintained). Finally, it also has to be quick & easy to access everything relevant from only partial information/memories.

Imperfect solutions that I have seen perform some parts of this well are:

  • Bespoke Excel or PowerPoint front-ends with hyperlinks to simple folder structures
  • Evernote app, with use of tags & notebooks
  • SharePoint/OneNote & other Intranet-based solutions for saving Office documents
  • Databases/Data Lakes capable of storing unstructured or structured data in range of file formats
  • Google search algorithms used to perform ‘natural language‘ searches on databases or folders

These can all fulfil part of the potential, but the ideal should surely be a simple as asking Alexa or Siri & having all completed work automatically tagged & stored appropriately. I’m sure it’s not behind the capabilities of some of the data & machine learning technologies available today to deliver such a solution. I encourage analytics vendors to focus more on this knowledge management space & less on just new coding & visualisations.

Do you see this need? How do you avoid reinventing the wheel?

I hope this petition has resonated with you. Do you see this need in your team?

Please let us know if you’ve come across an ideal solution. Even if it is far from perfect, it would be great to know what you are using.

Share your experience in comments boxes below & I may design a short survey to find out how widely different approaches are used.

Until then, all the best with your insight work & remembering what you know already.