Join us for a live demo of Avora on Thu 04/03 at 3PM GMT. Register now
Root Cause Analysis: Going Beyond Anomaly DetectionFibre Marketing2021-02-22T09:56:26+00:00
What is Root Cause Analysis?
So, what is root cause analysis anyway?
It’s Thursday. You’re in a performance meeting with the executive board, and the CMO asks: “So what happened with sales?”
You say, “Well – my anomaly detection tool showed that Sales were unusually low yesterday in comparison to last Thursday.”
You don’t know why, so you start making educated guesses. “It might just be that Monday’s a bank holiday, so all our prospects have taken today off for a long weekend.”
But the CMO wants more.
“What about that campaign we ran? I think we need to figure out exactly what happened here.”
You need to be able to say “Well, Sales were £1M lower yesterday, vs. same day last week, driven by poorer conversion rates across the Paid Per Click and Performance Social channels, for 3 specific campaigns targeting our Pay Monthly and Mobile Broadband products.” This is root cause analysis, in a nutshell.
But it may take your analysts days, maybe even weeks, to get to this conclusion. And by that time, the actionable insights they find will be out of date.
Decision making is getting more difficult for humans – and relying on gut instinct ALONE can be dangerous
We live in a complex world where we, as humans, are required to make quick decisions, based on incomplete information. And this can be a real cause of stress – on individuals and on businesses. Too often important business decisions are reduced to a combination of gut instinct, analyses based on incomplete or poor quality data or assumptions that contain unconscious bias.
There is nothing wrong with gut instinct – if the decision you’re making isn’t business critical.
If we want our businesses to be successful, we need to make the best possible decisions as space for errors decreases. We need to get to the root of the problem – fast.
How do I use root cause analysis to find out why performance changed?
At Avora, we use root cause analysis to answer this question. In commercial settings, most datasets measure performance of some sort. These datasets could be large – with billions of rows, and hundreds of columns wide! We’ve made it straightforward to ask this question, and bring back the most relevant reasons behind that change in performance, in seconds.
Can’t my analysts do this?
Your analysts should, in theory,be able to write complicated formulae and figure out the root cause of a problem. But this takes time.
They have to collate the data. Bring it together from all the different data sources you have. Joining it together takes hours, best case scenario, but quite often it’s more like weeks.
Then they have to analyse it. This involves slicing and dicing the data by different segments, recording the results for each segment. And rinse and repeat.
Next up is insight generation and presentation. Communicating these results concisely, that shows the information in context. This is a separate skillset. Many analysts will need to involve another colleague to help with this part.
It’s only after all of these steps have been completed that the crucial debate and decision making on what actions to take can begin.
If you take until Friday to conduct an analysis to answer a question that needed to be answered at 10am on Monday morning – you have to start asking yourself “what’s the point?” You’ve solved the problem – but how much vital time have you lost? And even more importantly – how much money have you left on the table?
So the real question is – should your analysts be doing this?
From a people perspective – nothing is more demoralising than pouring your heart into a piece of work that then isn’t needed anymore. That’s a surefire way to create a staff retention nightmare.
I’m better than a computer at finding out why!
Of course you are better than a machine at deciding what to do. At Avora, we strongly believe in the power of humans to make the most important decisions – and human decision making power is at its strongest when enough information is available. And gut instinct thinking of the type Daniel Kahneman focuses on in Thinking, Fast and Slow, is at its best when you are engaging in an action you have taken thousands of times so your brain has been able to develop patterns and intuition based on years of practice.
The reality is that every business decision is unique and comes with its own set of drivers impacting the final decision. Kahneman makes the distinction of the stock market in comparison to a game of chess. The former is unpredictable and can be chaotic; the latter is based on a set of patterns and rules (albeit complex) that, over time, a human can learn to navigate.
So the role of the machine in this chaos is to get you 95% of the way there. You allow the machine to filter out the least relevant possible reasons for the “why” you are trying to get to. You are then left with the most likely reasons – and those may well fit into a pattern that you recognise and can use to make a confident decision.
You get to play to your strengths, and the machine plays to its strengths – to your benefit.
And that’s what Avora is all about. Giving you the tools to test out the hypotheses only you, as the expert in your business, could come up with – but in seconds, not hours, days or weeks. You can test your ideas as you see fit, adding and removing reasons you think might be contributing to the anomaly, to give you the “smoking gun” analysis.
And that combination of human plus machine – not human vs machine – is a powerful weapon.
What are the real world examples of root cause analysis?
Sharp, forward-thinking businesses who use root cause analysis have gained:
A significant drop – 10-20% reduction in Cost Per Acquisition – in how much it costs to acquire customers. (Based on data from Avora’s Telco, Performance Digital Marketing customers.)
A 6% increase in Conversion Rate (Retail)
A 30% drop in data warehousing costs overnight
Time saving – analysts actually get to do what they’re good at – solving business problems that drive incremental revenue, or reduce costs (see Alvin’s Data Warehouse cost blog for an example of what can be achieved when we give ourselves the gift of time and capability).
Give yourself the freedom to make decisions based on actionable insights based on the best work of humans and machines.