Daragh O Brien on bad data, big data and the future of data

Garreth Hanley:
This is INTHEBLACK, a leadership strategy and business podcast, brought to you by CPA Australia. Welcome to INTHEBLACK. I'm , Garreth Hanley podcast producer here at CPA Australia. Today we're talking about all things data. Data is an integral part of business. It has been described as the new oil and our most valuable asset, but how do we manage our data so it doesn't become a costly liability? Today we're talking with Daragh O Brien, Daragh demystifies data for a living. He's the founder and managing director at Castlebridge and a university lecturer in Ireland. For a full list of Daragh's credentials, don't forget to check the show notes for this episode. In today's show, he explains the costs and benefits of data, where data risks can come from and how to get your data right. I started by asking Daragh about his unusual path into data management.

Daragh O Brien:
The first aspect of it is my name, the name Daragh, has many different spellings and is gender-neutral in the nine languages that there is a phonetic version of the name Daragh in. Ever since I was a small kid, I spelled my name with one R and a GH at the end. And ever since I was a small kid, I was very much in favour of people spelling my name right and it used to annoy me when people spelled my name wrong. And then I discovered all the other spellings as I got older. It made be aware of how data, data, whatever way we want to pronounce it, how important that can be to people and then trundling along through life, I did a degree in business in law in University College Dublin in the nineties. Before that my mother was running an adult education centre and she was getting software to try out and being a slightly nerdy teenager, taught myself the programme when I was eight, I got to play with all those toys. I started understanding there were software to do things with data.

So the degree in business law, then I went into working in a phone company and working for a friends of mine who did computer science degrees, two part-time jobs while I was in university, the phone company, I was doing data projects. I was helping build little things to help run call centres or more efficiently in the call centre working on the business side, I was never in IT and my friend's company, I was helping computer science graduates understand how businesses worked because I was coming from a business background. And eventually by 2009 I was the deputy head of regulatory in the phone company, having been the head of single view of customer for a few years prior to that.

How did I become a data expert? Trial and error hard work and luck, and I've also had the incredible good fortune to have some fantastic mentors both professionally and in the data world, many of whom are quite good friends now, long-standing friends who encouraged me to explore not just the technology side of data, but the human side, the business process side and the human factors side of data as an asset in organisations. That's the various strands of the story coming together. And I wound up now with Castlebridge, my team, we help organisations understand data as a business asset. We help them understand from the people side of the business how to use the technology more effectively so that they can derive the maximum value from data.

Garreth Hanley:
Now you mentioned that your name being spelled incorrectly is what started all this. Is there anywhere else where you've encountered bad data along the way?

Daragh O Brien:
There's a couple of stories here. I'll start with a personal story. A mobile phone operator here in Ireland who had my business mobile account with decided to write to me as Miss. Daragh O Brien, their trusted, their valued business customer, not once butch multiple times. My name is gender-neutral. And as part of cleaning their data, as part of enhancing data, they had decided to assign a gender to me. So long before we had any discussion about gender assignment in the media, I was given a gender assignment by my phone company and I objected to this and a number of iterations back and forth to them.

And what was happening was every time I got them to correct the data in the contact centre, they'd fixed the data on my record. And then a few weeks later, it'll be broken again, it'll be reset back to Miss. Daragh O Brien. Now the root cause of that was that the phone company had bought a particular data quality tool that had in its reference data set a rule that said Daragh was a female name. I understood the reasoning behind that, but it was very frustrating. When I was in the phone company, one of my key roles in the regulatory function was to make sure that the phone bills were accurate. What we had to do in that process was to stop incorrect bills getting out. The billing cycle was four weeks. We had two weeks to find the problem, a week to fix it, test it, and deploy the fix into production before the bills issued. The key element there is, there are small things, everyone talks about big data, but often it's the small data is the problem.

One of the funniest ones I saw over the years, if I buy a toaster on a Monday, Amazon will then send me marketing emails and prompts on the Wednesday and Friday asking me, saying that, "You've recently bought a toaster, would you like to buy another toaster?" Now, if I've bought a toaster, I don't want a toaster, I want some bread. And that's an example of using the data in the wrong way to generate the wrong signal and take the wrong action from the data. Like predictive analytics is fantastic if you're actually predicting things, but if I've bought a toaster recently, it's highly unlikely I'm going to buy another toaster immediately unless I collect toasters.

Garreth Hanley:
Daragh, I've heard a lot about data being called the new oil and Amazon is one of those companies that's leveraged data for their business operations. Recently we've heard about lots of data breaches both globally and also here in Australia, and there's been a lot of talk about how we can protect our own data. Now Daragh, you are the expert here. How would you describe what data is? Why is it being called the new oil and what are the benefits for businesses who are collecting data?

Daragh O Brien:
Okay, what's data? And it's simplest, it's a record of a fact, about a thing, an event or a person that when used with an understanding of its context can tell us something about that thing, that event or that person that we can then use to make decisions or take actions that will add value. If you look at the fundamental accounting definition of an asset, something which on its own or when combined with another asset can be used to generate net cash inflows or deliver services. That's data. It's describing the fact of the thing, the event or the person. It becomes information when you have context around it, that enables you to understand it, that enables you to apply it and then it becomes actionable knowledge once you can put that into a process or a procedure for taking action. I'll give an example. 007, what is that? We don't know what that is without some context. If I say, "007."

Garreth Hanley:
James Bond springs to mind.

Daragh O Brien:
It's also the international access code for Russia.

Garreth Hanley:
There you go.

Daragh O Brien:
Who says data nerds in the telecom sector don't have a sense of humour? I'm trying to find out the chicken and egg here, which came first, 007, as the character or 007, as the international access code? Because either Ian Fleming had a sense of humour or telecoms and nerds had a sense of humour. I'm not sure which. That's what data is. Data is the new oil. Well, the origin of that phrase is a woman called Michelle Finneran Dennedy, who is a very good friend of mine. She was the first chief privacy officer in Silicon Valley in the nineties. She was privacy officer for Sun Microsystems. She now runs a privacy unit, enhancing technology startup called PrivacyCode in California. But she was actually misquoted. She said, "Data is the new oil, except that it's not."

And everyone missed the second part of that sentence because data's the new oil to an extent in that it's a valuable asset, it's a consumable asset. It's something that can be used to fuel and do other things. It's an easy, lazy metaphor. But data's different to oil because it's a fungible resource in that you can use it many places at many times it's non-exhaustive that the information you have about invoicing in a business can be used to inform marketing planning, can be used to inform credit decisions as well as managing the bank balance and the bottom line of the business. In that context, data is incredibly useful for fueling organisations and fueling businesses.

In terms of data is the new oil, the only way I would say data is new oil applies is when you talk about those things of data breaches and data leaks, data is the new oil and every day we see the equivalent of a digital Exxon Valdez where the oil is escaping and bad things are happening, as a result. I often say that data is the new plutonium because it's really powerful, really useful, but you wouldn't want to have too much of it on you. And one of my students in university coined the phrase data is the new nuclear fusion because once you get it right, it becomes a very powerful and self-sustaining source of energy for your organisation.

Garreth Hanley:
If data is the new nuclear fusion, let's talk about the risks. What are the risks for businesses who are involved in data collection and data storage?

Daragh O Brien:
Okay, first of all, every business is involved in data collection or data storage. Everyone focuses on the big tech, yay, they have lots of data, but if you're running a bakery and customers and you have a loyalty card scheme or you know about your regular customers or about your suppliers or you're generating invoices or you're having invoices coming into you, that's all data. In terms of the risks, you've quality risks. Is the data accurate? Is it up to date? Is it complete? I've worked with one business in Ireland here where one of the key challenges they had they encountered just coming into the pandemic was that the accounting firm they were working with wasn't updating their internal management accounts on a frequent enough basis. They were doing the financial accounting side of things fantastically well or reasonably well, but the company actually didn't have a real bottom line accurate understanding of where they were month on month or quarter on quarter.

They went into the pandemic with a realisation that there was a hole in their bucket and they needed to fix that hole while dealing with lockdowns. And that was a challenge for them. And that's data, that's data completeness, consistency, timeliness, all these measurable characteristics of quality. The risk is if you are mishandling that data in any way, if you are having proper controls over that, that can impact the trust that customers or staff or suppliers have in your business. And that's not just security, it's things like spelling customers names right. It's things like ensuring that you apply marketing suppressions correctly on data, so that you don't send spam emails to customers who don't want them. And finally, there's the obvious one, security. It's a valuable asset.

It doesn't appear on your balance sheet as an asset, but if you have a ransomware attack and you suddenly can't do business, you become very painfully aware of how important an asset data is to your organisation and the risks there are malicious actors outside your organisation trying to leverage your asset, trying to steal your asset in the same way as malicious actors outside your organisation might try and get in and steal physical assets, but also there's the accidental or malicious actions that staff might take in your organisation. A well natured, well-meaning member of your staff does something accidentally and sends a customer list to one of your competitors by mistake. Shock horror, bad thing has happened to you. A valuable asset in the hands of a competitor or a disgruntled soon-to-be ex staff member decides to copy a customer list to bring to one of your competitors because they're changing job.

These are all the types of risks that occur and 69% of security risks are happening inside the organisation. People accidentally doing things or not thinking about what they're doing with data. And ultimately when we look at the statistics on data quality, data quality costs between 10 and 35% of turnover in the average organisation because every time you have to correct incorrect data, you're paying for that data twice or three times. There's the cost of collection, the cost of correction of the data, maybe the cost of correction of the relationship with the customer, you have to give a refund or a voucher or something, those are the sorts of risks and issues that can arise.

Garreth Hanley:
You're talking about a combination of collection, usage and regulatory?

Daragh O Brien:
Collection, usage, regulatory, because you have to think about it holistically.

Garreth Hanley:
Okay Daragh, so for our listeners who are running or managing a business and collecting data, if they're aligned with their local regulatory requirements and have assessed their collection and usage, but a concern they might be getting bad data, what do they need to look out for? And if they do find bad data, can they turn it into good data?

Daragh O Brien:
One of our clients here in Dublin is a dog re-homing charity and they have a mantra that there's no such thing as a bad dog. There is a dog that has been mistreated, mistrained or isn't the right dog for the environment that the dog has been put into. And that's where you get behavioural issues with dogs in their experience. And those are things bad dog. There's also thing as bad data because it's all about the context. What's good data? In an organisation context, it's about looking at the process that you're using the data in and understanding what is critical to the quality execution of that process, the delivery of the outcome you want. If it's invoicing data, what are the business rules around what you need to have in an invoice for it to be processed accurately?

What are the things that cause delays or impacts on your invoice to pay process? What's the lead time there? If a client decides that there's a purchase order not on the invoice, the absence of that purchase order number is a data quality problem and we need to understand where that came from. We've now made sure our processes are aligned that if we always ask is the client's payment process and invoices requiring a purchase order? And if it does require a purchase order, boots do not go on the ground until we have a purchase order number. And that's a control we have in place so that when we send in our invoice, we will get paid on time. And that's the process we've mapped to make sure we know the data and what's critical and that's how we get good data. Can you turn bad data into good data? There's lots of tools out there for cleaning data and correcting errors in data. Going back to my first example about my name, you have to be very careful and clear about the rules you are applying to the data in terms of what good data looks like. Defining what good is the first step in moving from less than good data or bad data to good data.

But it's much more important, Garreth, to actually look at the process that's creating the data and fix the process. The best way to fix data problems is to fix the process that is causing the data problem because if you focus on, there's lots of businesses, lots of companies out there selling tools and technologies or services to clean data, to clean marketing lists, things like that. But again, that goes back, you're paying for that data twice. Whereas if you fix the process, if you plan for your data as an asset and understand your process and identify where the problems are arising, that's how you start to improve the quality and usability of your data. Something as simple as a business over the years, we've changed accountants a few times and it's things like the mapping of the chart of accounts as you're moving from one accounting package to another, migrating data from one accounting package to another, it's one thing to take all the balances and start a clean slate moving into your new accounting package.

But from a management accounting perspective, an internal reporting perspective, I need to be able to compare quarter on quarter and period on period at a more granular level. We need to be able to map and migrate that data. We've actually worked with our accountants here in Wexford in Ireland to map, we did a proper data migration mapping, which they'd actually not done to the level we would do it before because we wanted to make sure we had consistency of the data because we wanted to make sure our reporting would run smoothly as we moved from one accounting package into another because we didn't want to have the overhead and the hassle of running three or four Excel spreadsheets on the side to keep a track of our internal management account.

Bad data becomes good data by improving the environment, improving the process, and thinking about what's important. And that allows you to maintain the level of quality of data because you've planned for it as an asset in the same way as if you're hiring staff, you've planned for the staff you need, you have controls in place through interviewing, et cetera to try and get the right staff into the role. And you have measures you take in terms of performance management to improve staff, to make sure they're doing the job they should be doing. Or if you're buying something as humble as a calculator or a computer, you've specification for what you want it to do and the functionality needs to have. Having that specified from a data perspective is the first key step to making sure that you don't have bad data, that you are capturing good data as much as possible and that you know how to turn your bad data into good data.

Jackie Blondell:
If you’re enjoying this episode of INTHEBLACK, you might like our Excel Tips podcast. Each week our resident Excel expert, Neal Blackwood CPA brings you tips and tricks for Microsoft Excel. Search for Excel Tips in your favourite podcast app or check the show notes in this episode to subscribe.

Garreth Hanley:
It sounds like what you're saying there is that having a lot of data if it's not collected properly, can be problematic. Before businesses collect data, they should firstly identify their needs, then develop a process for collecting the right data and then think about starting to generate and collect the data. Is that right?

Daragh O Brien:
That's the ideal world, Garreth, yeah. The life cycle of data is exactly the same as the life cycle of any other asset. And the problem we have in organisations most often is that people jump into the obtain stage. The generic life cycle of data plan, obtain, store and share, maintain, apply and dispose. They're the key steps we go through in handling any asset from shoes to laptops to data. But we jump in at the obtain stage quite often with data where people go, "We can get all its data, so let's have it and then we'll figure out what we're going to do with it later." Anytime you've filled out a loyalty card application form in your supermarket or a gym membership form, all those questions they ask you, sometimes they don't need that data. Sometimes that's them gathering information and then they'll figure out what they want to use it for later. The best way for organisations to manage data risk is to be clear what is important? What's the critical data you need? What are the critical things you need to do with data and have a strategy for data in your organisation that's supporting your business strategy?

And that's for organisations of all sizes. And we work with organisations from government departments and multinationals down to small, not-for-profit organisations and for all of them, while the mechanics and the methods might differ slightly in how we go about doing it, the fundamental is what's the business objective you have? What data do you need to deliver on that business objective? And then what do you need to do to either get the data or manage the data, store the data, apply it in the organisation, and also think about what level of quality you need, what level of formatting of data do you need and to make sure it can be used consistently so you're not having to incur additional hidden costs. I said earlier on 10 to 35% of turnover, that's a number that hasn't changed in 25 years. It's the little things of people having to re-key phone numbers or correct customer addresses, things like that. Part of your plan for data then needs to be where you're storing us, how you're managing it so that you can apply appropriate security controls, but also so you can make sure you don't have multiple copies of the same data.

From an accounting perspective, if I told you I had four different versions of the books for my business because the marketing department didn't like the accounting system, so they wanted to do things their own way, that would probably cause auditors to shutter significantly. It's hard enough getting your core data set correct and managed and maintained without having all these satellite copies popping up around the organisation. And that's what the plan comes in. Sometimes redundancy is important and useful. You might, in an accounting context, take data from your accounting system and use it for and take copies of that for analysis in a management accounting or management reporting perspective, same data, different bucket, but that's something you're planning for and you've processes and controls around. And that's the key thing for organisations of any size. What's your business objective? What data do you need to achieve that objective? What's your plan for that data? And then implementing that plan in the same way as when you're setting up an accounting system, what's your chart of accounts, how do you structure it? And then what are your processes for invoices in, bills out and tracking the cash? That's all data.

Garreth Hanley:
Should businesses be focusing on big data at all, or are they better off to focus on small or bespoke data?

Daragh O Brien:
Big data is just small data that has lots of friends. I come from the telecom sector and long before there was this hype around big data we had call data records, every individual telephone call that you make generates a significant amount of data, even a landline call. And on an individual level that's used to generate your bill, on an analytics level, we can start to look at that at scale and go, "Hang on a second. This is the pattern of calls this person makes and we can use that data for other purposes." Don't get me started on what we can find out about you from your mobile, from your cell phone, where you were when you made the call, how fast your car was moving when you made the call, based on how quickly it took you to get between two cell towers, things like that. I always hated the term big data because it's a marketing word, it's just data. I actually coin the phrase morbidly obese data because that's big data that's sitting on the couch doing nothing. And for most organisations, that's what they have. They think they've got big data, but they don't. Because most of what they have, they're not using.

Garreth Hanley:
Is this a case where businesses are looking for the solution before finding the problem?

Daragh O Brien:
In a lot of cases, that's what it is. Organisations have gathered the data. The disciplines and the plan for whether it's big data or your small data, or bespoke data, it's the same thing. It's the same steps you go through. What's your business objective? You mightn't need big data, you mightn't have enough customers to have big data, but it might be really, really essential that something as simple as your invoice to cash process runs smoothly. And that when you're chasing invoices, you have the right contact information for people in your customer organisation so that you can get through the layers quickly to collect. And for most businesses, that's actually more important than a future looking predictive analytics piece. I'm really passionate about this aspect of there's the fundamentals of doing data and the fundamentals of doing business. And then everything else is value add, but it's only value add if you have a business. I've worked with lots of startups doing AI technology. It's one of fantastic, fascinating stuff. But the key question keeps coming back, how is this going to add value to the business? And that's where your business objectives data strategy is big data part of that and what we do. Because ultimately all the data quality problems you have as you aggregate your data up to create big data, it's made up of your small data. And if that data has problems in terms of accuracy, tidiness, consistency or data quality issues, the analysis you do on your big data will be wrong.

Garreth Hanley:
Do you see that a lot where businesses are collecting the wrong type of data or interested in something that's less important than what they should be interested in?

Daragh O Brien:
Yeah, we see that quite a bit. Organisations who are just gathering data because they've been told big data's an important thing or you need to have improve your analytics capability. Fantastic. What data? How? Why? A lot of what organisations really need to do isn't glamorous stuff. Big data is great, has a wonderful marketing department. AI has a wonderful marketing department. Clean data, doesn't really have a good marketing department, but it's actually probably more important in an ongoing operational context in businesses. In that context, we worked with an organisation a couple of years ago that was looking to spend quite a lot of money on a predictive recommendations engine for their product. And the problem they had was they didn't have good enough data about their product to actually inform the predictive analytics engine, the things they needed to use to analyse to make those recommendations to customers.

They just didn't have that data consistently to a high level of quality and completeness in their product catalogue, so that when they pointed this recommendations engine at that data to tell customers, "We see you bought a toaster, would you like some bread?" They didn't have that data to make that next best product recommendation. And in the end, they actually pivoted their data strategy towards focusing back internally on improving product quality first. And that actually turned out to have a bigger impact for the business because it reduced the admin overhead they had fulfilling orders because on the admin side of their business they were having to correct orders, correct shipments, correct invoices because the wrong product codes were being used and things like that within the business because the data they had wasn't as good as they thought it was. But that had been seen as the cost of doing business for a while until we looked at it and said, "Well, it's 11% of your turnover. Are you happy with that? Giving the target of improving your profitability by 5% this year, by fixing this, you exceed your target and if you spend your money on this fancy thing, it'll fail." And that's quite a common thing because people think data is a magical thing and it is, but only if you've cared for it and fed it properly.

Garreth Hanley:
Okay. Looking forward, Daragh, what is the future of data collection and data quality? You've mentioned AI is having a great marketing department and there is a lot of talk around tools like ChatGPT and other AI systems at the moment, do these tools have potential to improve data management? If organisations are looking for these sorts of tools to help, will they help? And what do you think the next 12 months of data collection might look like?

Daragh O Brien:
I'm always wary about making predictions. What I would say is that any tool will help if you are using it in the right way and if you are putting the right resources to use with that tool. Going back the definition of an asset, something that on its own or combined with another asset can deliver a net cash inflow or deliver service. Tools like AI tools, there's a lot of fantastic use cases and examples of AI being used really, really well in a variety of industries to improve processes to speed things up. But they're usually industries that have large amounts of data available. Tools like ChatGPT, I think will help raise people's awareness about the importance of quality in to deliver quality out. Garbage in, garbage out remains a fundamental concept in data and in life. As organisations start to adopt these tools, they're going to start finding that they make stuff up because all they are, I don't know your vintage Garreth, but I'm old enough to remember Microsoft Clippy, the annoying paper clip in Office 97.

Garreth Hanley:
I thought maybe they'd bring Clippy back with some of this new technology.

Daragh O Brien:
Clippy was great because Clippy could identify you were trying to write a letter and would offer to help, but knew nothing about writing letters. And these AI tools are just an iteration above that where they're collecting data and they're responding to prompts, but they're a predictive model. They basically crunch data to figure out what the thing should be in response to your answer. And they've already been identified as making up things. I asked ChatGPT to write a profile for me as a biography for a book. I said, "What would you write as my bio?" And it was wrong. It credited me writing a book that a competitor wrote. And it also told me that I was on a data governance board run by the Irish Data Protection Commission, a board that doesn't exist, but apparently I chair it.

Maybe in a parallel universe, maybe that's happening, put garbage in, garbage out and understanding what the tools need. I think what we're going to find over the next 12 months is that we might see the beginnings of a demystification of some of these tools, which will be very, very useful, which means businesses won't necessarily dive into spending large amounts of money on them. In the Australian context, I know your attorney general has just announced or performed a proposal to improve Australia's federal data privacy laws to align with European standards, to align with GDPR. That's a big wake-up call for businesses and any of your listeners who are advising their customers, time to get your house in order. It takes three to five years to get this right from an internal operations perspective in a moderate to large sized organisation, the basics are relatively straightforward, but it's a culture change and mindset change, not a technology change in most organisations, and that's something that's coming down the pipeline in Australia. But in terms of new tools, we're going to see lots of hype, lots of hoopla.

We're going to see lots of companies that have spent lots of money trying to develop these technologies, coming under pressure to deliver a return on the investment for their investors. And that's where we're going to start seeing the tide going out. And the question will be, which of these tools is actually wearing their swimming togs when the tide goes out, which of these technologies actually can be demonstrated to deliver consistent value? And that's going to really come back to how are they being used, and how good is the data they were being built on and trained on.

Garreth Hanley:
Is that regulation piece something that's happening worldwide?

Daragh O Brien:
Yeah, it's happening globally. Personally, I've been involved in working with countries and sub-Saharan Africa, my companies currently involved in number of projects in the Middle East where these laws are being enacted, and organisations having to put in place these processes, these controls, and they're having to actually start thinking about data and thinking about it in the context of their processes, not as a byproduct or as a second level order of concern, but as a core part of how they do business. And it's the data about people. That crops up in places you mightn't expect. You might be in a business to business context, but unless you are selling to a robot on the other end of the phone, you are talking to a person, you are sending an email to a person, you are phoning a person and you have data about that person. And even in a business to business context, if you have a mishap with that data, if you have your equivalent of the digital Exxon Valdez and the contact information, or information about your interactions with a contact in another business, are disclosed or leaked or become subject to a ransomware attack or something like that, that's embarrassing to your business and potentially harmful to your business and almost certainly costly to your business. And the regulation will cover that because it's still data about people.

Garreth Hanley:
Are those regulatory changes due to a shift in the way governments are thinking about data or as a response to recent events?

Daragh O Brien:
One of the key things we're beginning to see is organisations and governments are beginning to realise that data is a social thing, not a technology thing. And it affects society and how people interact in society, partly disinformation and misinformation on social media. But even simple things like how governments use data to inform decisions on policy. And that's going to become an increasingly important aspect of social and political life going forward. And that leads to potential ethical issues and ethical concerns as well about how we use data. Even before we have laws and regulation, we have to think about how we're using data to benefit people, benefit our organisations and benefit our societies, rather than necessarily having a very narrow focus on how we are going to maximise the bottom line rigorously, data is a slightly bigger consideration in that regard.

Garreth Hanley:
Daragh, thanks so much. It's been great talking to you. Unfortunately, that's all we have time for today, but thank you for joining us on INTHEBLACK.

Daragh O Brien:
Thanks very much, Garreth. It's great to talk to you.

Garreth Hanley:
If you've enjoyed this episode, help others discover INTHEBLACK by leaving us a review and sharing this episode with colleagues, clients, or anyone else interested in leadership, strategy and business. To find out more about our other podcasts, check out the show notes for this episode. We hope you can join us again next time for another episode of INTHEBLACK.