top of page

Bad Analytics

In all my research to date one thing is clear. No one is safe from bad analytics! You have them, you know you have them and you have learned to live with it. It does not have to be this way. I am going to use examples and illustrate how you can break this trend and leverage the knowledge provided to recover. There may still be an intervention needed but I will lay out the basics and you can decide.

The first set of analytics I noticed was my Utility bill. Now it does not seem to be the place you should look but it is the best example of the absolute worst analytics that will never help you lower your bill. It is almost like they do not want to really help you. You use a product (energy) and at the end of the month you get a bill. You know what that reminds me of? The revenue forecast. Did you hit that number? When the books close for the month and you see the forecast was off then it is just too late. That is ok because you are going to do the same the next month, in fact you are going to do it through the whole year. If you take one piece of advice from this blog it is this: If your growth remains flat or has gone down for the year without much external influence (Covid etc) then ditch your analytics they are doing you no good.

The whole purpose of analytics is to give information that can help you decide what to do. In a timely manner is the other part of the equation.

So the first law of analytics is: Analytics must be effective.

While this sounds obvious I can assure you it is quite complicated to have effective analytics. It combines information about what is happening with your sound business understanding about what to do about it. Lets look at the energy example. What if you knew a week into the month what your usage looks like and an estimate of your bill. The odds are better you might pay attention to the usage and manage it better but its really still not a lot of information. Basically you are just being told that you will pay more at the end of the month just budget for it. What you really need to ask is what exactly should I do? What lever at my disposal can I pull so that this will be managed? Does the analysis work for you then? It is not enough information, also it is not enough Actionable information. Now what if you found out that your Pool pump is using 40% more power than it was last month at this time. Suddenly you have a lever you can pull on. Checking the system is working ok, the filter is not full these all become areas to suddenly explore and quickly. The information is actionable and provides you (the one with the knowledge of your home business) the levers you need to sort the problem. In business analytics do not work this way because often they are bestowed upon a business unit from accounting without a clear understanding of the drivers of the business. The business unit often builds its own analytics and still there is little resemblance to a really good set of analytics because the business unit is good at its business not the business of creating analytics that are useful. The creation of good analytics combines the business unit working in tandem with some top notch analytics experts. A good analytics consultant should be able to benchmark you against the competition in various areas to make sure you are competitive.

To summarize, the fact that your bill is going to be high is an Indicator. On its own it is not a lot to go on. You could also call it a performance metric. The next level down could be House, garage, Pool. Another level down below Pool could be pump, lights, skimmer etc. To have actionable analytics you need to have an indicator and then be able to go down two levels to find the lever on which to pull. The Utility company can give you an indicator but it does not really do much more than that. Analytics need to be much more sophisticated and useful to a business but in reality just like the Utility company you do not have the next level of data because many companies have data silos. Insight across silos does not exist. It is almost like saying Production systems do not talk to IT systems so data about how much IT they are consuming by product is simply not available. If you want to at some point really work on customer 360 or initiatives to push sales then I say first fix your silo problem internally so that you can manage externally.

Thus we announce the second law of analytics: You must be able to ask questions and get answers.

This harks back to the point of being able to get to the next level to find the issue and pull that lever. If you have a backlog of data requests in IT you have a problem if you are going to start asking questions. Analytics needs to become its own thing, if you tie it to IT operations you are going to have tremendous backlogs. If you separate your analytics you risk data silos, redundant data sets and duplication of effort. Sounds bad on both sides. A better strategy is to look at data as an asset and focus on managing it as such. Data becomes valuable when it does something valuable for the company. Note if you look at the top ten companies in the world they are valued on their data not their infrastructure. Data efforts pay off when you use analytics effectively. A best practice approach is to build an architecture that is agile and can ingest information quickly in all forms and get it to actionable insight quickly. When you apply scale, hardware and regulations to that construct everything slows down. A good way to manage this is to build on Cloud and remove some of the complexity.

The third law of analytics is: Once you have answers you should be able to get recommendations on what to do.

Prescriptive analytics have really come into play helping companies to answer more complex questions. If you are having an event and start signing up customers online, when do you drop the price to attract more people? Or indeed what if you lower the price will that increase attendance to a level where it makes up in total revenue what you lose in margin? AI and ML models are now being able to answer these questions, albeit every time you ask a new question you really need to create a new model quickly but as long as the data is there you should be able to manage it. Most companies we have talked to do not have this level of capability across the enterprise and vendors are now beginning to start building that capability. Research has shown us that the less complex systems you have in place the easier it is to start these initiatives or indeed wrap them into the DNA of your implementations. If you have legacy old school AS400s and lots of Access databases around the office it is not easy to bring things together.

In future articles I will get into how to really build good analytics. The right metrics, the right indicators and KPIs as well as what is a forward looking KPI and what is a lagging KPI. You have seen a lagging KPI in this article and you know the result. You need to move your analytics to a forward looking capability and build a quality analytics system that allows your business to ask questions and get answers.


About the Author

Asim Razvi is the head of data management and data strategy at Onis Solutions with over 25 years experience in delivering world class solutions in data to clients. He has architected some of the largest hybrid data management solutions for the Fortune 100 and also worked closely to deliver Business Intelligence strategy assessments to them as well. He works and collaborates closely with a number of CDOs and maintains a busy schedule of events and speaking engagements. Outside of work he trains outdoors to maintain a healthy lifestyle and spends time with his family in the wilds of the California mountains.

Asim Razvi

Vice President Lead, Data Management

165 views0 comments

Recent Posts

See All


bottom of page