BI AI ROI in 2018

6 Hidden Secrets to Improve BI & AI ROI in 2018

As another year approaches the end, it’s a time to reflect on where we are in developing our cognitive capabilities. Its been a year where media has reflected a continuing plateau in the return on investment [ROI] with business intelligence [BI] or artificial intelligence [AI]. There are many reasons for this – some well recognised, such as data quality, and others frustratingly silent in their menace. So, I thought I would share 5 things that have the potential to improving BI ROI.


1. Set More Realistic Expectations

We all know that ‘analytics’ is the darling of the day, but unless you get the fundamentals right with BI, there are going to be some major hidden time-bombs coming your way. Sound overly dramatic? Perhaps; but it is very real.

Businesses often expect that BI and analytics are like a magic pill, one that is going to reveal unique information to foil competitors and escalate market share. Yes, that does sound a bit naïve, and I don’t mean any disrespect, but it is essentially what many projects market to the business in their business cases.

BI tells us a lot about what has, and is, happening in our business. Adding in predictive and prescriptive analytics, and we start reaching into the zone of potential for our future value. However, analytics is not fool proof. Nor is it totally, independently logical.

There are three aspects of BI and Analytics ROI that I urge you to keep top of mind.

  • Firstly, analytics is a very powerful capability, but analytical predictions are only as reliable as the data upon which it is based.
  • Secondly, analytical models are not independently logical. There is inherent bias in the raw data; some of which can be filtered out during the preparation process, but there is still inherent bias that is retained through collection methodology. There is also the bias of the data scientist that develops or configures the model.
  • Thirdly, insight revealed from the data must be integrated into a robust, logical process. That process may be mechanical, triggering business-rules governed actions, or it may be personal. Either way, we need to have a simple, structured framework to integrate data-driven insights into our businesses – and, with the risk of a overused cliché – in the right way, at the right place, at the right time.

Attempting to implement BI tools without an holistic approach to how the insight is best utilised is not only a waste of money, it screws up the value perception of BI and analytics projects in the future.


2. Don’t Get Romanced by Big Data

Big data is reaching the point of media saturation – it is at ‘over-hype’ stage. We now know a lot more about big data, where it comes from, how it is collected, and what we can expect from it. Yet, it still retains its prominence in executive discussions. The bulk of big data is just irrelevant noise. Until we have the tools and the skills to filter out that noise, including bias, it is not something that will radically transform a business. The data is just too dirty.

A 2012 survey by the Economist Intelligence Unit, reported that in processes “where Big Data analytics has been applied, on average, they have seen a 26% improvement in performance over the past three years, and they expect it will improve by 41% over the next three”. Data from those three years has not really backed up that claim. And, it seems that success is very industry specific – with those operating in manufacturing or high-volume transactions, where data production is very high, the numbers continue to improve. These are the industries where the dream of a ‘data-driven’ business’ rings true.

Used for more tactical purposes, big data is proving its worth. For example, to prevent brand damage – social media data can acts as an early warning system to alert businesses when customers are turning against them, or their products.

So, there is definitely real value in combining big data and analytics, but there are also challenges and limits that few talk about. Challenges relate specifically to data quality, governance, consistency, and the big one – organisational silos preventing better use of Big Data. Even with the emergence of “trusted data aggregators” to break down silos, the quality of big data is still highly unreliable on several levels.

On a more strategic level – it is not an easy journey applying advanced predictive analytic techniques in real time, especially where unstructured data is in the mix. It is in the layers making up unstructured data that the real competitive edge is hidden. Whilst visualisation tools such as Tableau may be able to ‘represent’ and categorise unstructured data, there are yet to be any really effective tools [other than image/facial recognition] to extract really valuable analytical insight from it.

There was a time, not that long ago, where we would have 12-18 months to gather the insights needed to make strategic decisions, but now that cycle has reduced to often no more than 3 months, making early access to data a critical part of the decision process. And we are not yet able to deliver in time, enough of the time.

Until we can overcome these challenges, and the quality and management of big data is viewed strategically at senior levels, we will not unlock its full value. Making better use of business activity data (e.g., sales, purchases, costs) still adds the greatest value to most organisations.


3. Delay Rollout Until Data Quality Hits Minimum Acceptable Levels

With any BI or analytics initiative, data is around 80% of the project effort. It’s the biggest decision of the day. When faced with dirty data, you are faced with two options:

  1. You get your data cleaned up now and enjoy a much smoother, faster, more successful and less stressful path forward with business intelligence, analytics and all cognitive technologies along the AI roadmap, or
  2. You hold off, and just get the software implemented first so you can at least score a win with that and then work on the data over time.

Let me share with you what I have found consistently over participating in, and studying numerous BI implementations.

Option ‘a’ gets you just what you think – in the short-term the business won’t be that appreciative because they don’t see what is going on under the hood and have no idea how dirty their data is – and the disastrous impact that dirty data can potentially have on their decisions, and productivity. They will feel frustrated that they cannot get you to provide them with the tools they want, now! But, you can use that frustration to get them onboard as data owners and stewards to get the data cleaned up faster. Then, once they engage with the data through the BI and AI tools, they will get the results they want. So you eat crow for the first 6-12 months, and then become the hero for the rest of your tenure.

Option ‘b’ makes you the great person that delivered our tools in record time, but when the results [inaccurate outcomes based on dirty data] don’t meet expectations, and who do they blame for that – no, not the vendors or the tools; they blame the person that is closest, and over whom they can exert the greatest pressure – YOU!!. So, you are a hero for a short term, and someone to distrust for the rest of your tenure.

Now think about this for a moment – how would you rather experience the next five years?


4. Leverage every iteration

Analytics is an enterprise-wide capability, end of story. Whilst getting analytics accepted by the business starts with one passionate head of a business unit with a problem that BI or AI can solve, it is the enterprise scope of analytics that delivers the value. It’s a familiar approach, finding the biggest problem that can deliver the greatest value, but even with this approach, the numbers rarely stack up post implementation.

Most point solutions have not, and will not, meet ROI expectations. It is not until an analytics implementation is shared across multiple business units that the ROI starts to look exciting. This cannot happen with localised point solutions culture. BI and analytics need to be enterprise-wide initiatives that are governed and managed by a central group – outside of IT.


5. Ensure Insights Are Effectively Communicated

Why is it that around 70% of decisions makers claim that they don’t understand the data they are seeing in their dashboards. There are two reasons for this:

  • The dashboards are poorly designed – with too much emphasis on ‘presentation graphics’ and not enough on visual design principles that work with the brain.
  • Users are not educating themselves on how to integrate insights into their thought and decision processes

Whether its BI, analytics of big data, it takes more than just the data and the tools, it takes a cultural transformation that includes accepting insights from data, integrating them into thought processes, operationalising insights into automated decision making, and a whole new language that supports establishing communication between IT and the business.


6. Review Decision Making Processes to Leverage BI

Adding in personal insight is an essential part of any decision – it is how this is done that is most important. We know that emotions impact how we ‘perceive’ things; with unconscious biases leading us to either not seeking out information that we need, discounting that which we do receive. Layer upon that how our unique personal values taint our perception of reality and it doesn’t matter what the real data is saying, your brain is going to interpret it in a very unique way.

In spite of technical limitations being increasingly overcome for operational BI/analytics, the accountability for important business decisions still ultimately rests with a human, not a machine. Unless human intuition is logically factored into data-driven decisions, there is a major weak link in the chain.


Onwards to 2018

There is a common thread through many of these challenges. We need to spend more time weaving human factors into the use of BI and AI. Whether that is in better designed dashboards, a more holistic approach to AI across the organisation or helping decision makers gain more awareness into the role bias is playing in their ‘intuition’, or the negotiation of a common language between the business and IT. It’s #theHumanFactor that is being overlooked. And that’s one problem only humans can resolve.

So my strong recommendation for 2018 is to step back a moment, and start considering where the human element of BI and AI is not being resolved.

Leave a comment

Your email address will not be published. Required fields are marked *