At a recent webinar, we tackled how to ask the right questions for your data strategy, and the audience brought a rich set of questions for us to answer. With a varying level of experience in using data and building data systems, the questions reflect the wide range of approaches and needs professional accountants have in making the best use of the data available to them. Franki Hackett, Chair of the ICAEW Data Analytics Community Group and Associate Director, at Grant Thornton, answers a selection here.
Storing data
How to keep and manage data was the source of many questions, especially from those who are already doing some data analysis but want to make this more consistent and ensure their clients’ confidentiality.
What is the view on one database vs multiple?
In a perfect world, perhaps a single database would be the answer to this question, but that would require the size and resources to make it worthwhile building something secure which was capable of storing all of the data, and making it available in all the ways it’s needed. This isn’t a trivial matter, when you think about all of the different data points we need to store and use as accountants, from a client’s annual revenue figure to an employee’s NI number. It’s very unlikely that many of us are going to be seeking ‘one database to rule them all’ in this context, and given the market for data storage solutions is mature and offers a lot of high-quality options, for most of us it is about finding the solution which best works for our different use cases and then making sure our different databases talk to each other.
Having said that, when you have the same data or the same kind of data in multiple places, this can obviously be a recipe for confusion and poor use of data. In these cases, reviewing your data estate (which might feel like a grand term if what you have is spreadsheets, but still!) and identifying the most rational approach to keeping consistent data together in one coherent place is a sensible exercise. The key thing is to know what data you have, where you have it, how you use it, and how it’s kept safe and up-to-date.
Software choices
Some questions asked specifically about which software we would recommend, or how to choose and use software. While ICAEW can’t recommend specific providers, there are some principles worth noting. The recently published cloud software guidance includes a lot of advice on this.
How do I get data into PowerBI that I currently keep in, say, multiple spreadsheets? Is there a particular software type?
PowerBI already has this functionality; you can link PowerBI to all of those underlying spreadsheets within one dashboard and then analyse them together. There are plenty of tutorials on this online. Then you just keep your spreadsheets up-to-date and your PowerBI will always be able to give you the latest analysis.
This is a key point about PowerBI though: it is a data analysis tool, not a data storage tool. If you’re hoping to get rid of your spreadsheets by ‘keeping’ your data in PowerBI instead, you can’t do this. Instead you’ll want to look into alternative data storage solutions. PowerBI can access data from many if not most forms of database, but choosing what kind of database you want is a separate question of how you use PowerBI.
You mention process flow and optimisation, is there a good tool that helps visualise as I always find trying to do that after a discussion devolves to hacking around in PowerPoint or Excel with arrows and text boxes?
There are two different levels of answer to this question, and they start from asking how you are getting to information about your process. The question assumes you’re understanding your processes through discussion: understanding from people how they perform the process and then drawing it. If you take this approach then software like Microsoft Visio, Miro, Figma, and similar design and drawing tools are probably your best bet. They tend to have a reasonably gentle learning curve and let you draw nice-looking process diagrams quite quickly.
This approach gives you a good understanding of what people think is happening within the process, but it’s often quite a poor guide to what is actually happening. For that, you need to take a different approach and look at what the data tells you about how processes really function. There are specific process mining tools on the market: the leaders are probably Celonis, Disco, and Blue Prism and many others are available. These tools do require the right kind of data in the right kind of format, so can require a reasonable amount of data cleaning up-front; but once you achieve that stage, you’ll get an instantly richer picture of the on-the-ground reality of what’s happening in the process. And they come with visualisations built-in.
Is there a good AI model that is well trained on public sector regulations etc. I know somebody at a not-for-profit who has a DFE query and they want to understand it better.
Choosing models is always a tricky one to advise on because the use cases can be so specific, as they are here. The key thing for this question is that there is not yet a Generative AI model on the market which is specifically trained for public sector regulations (or any other) use case. These models are trained widely on data which almost certainly includes the public sector regulations (ask one of the models a question about the factual content of the standards – they tend to give a pretty good answer), but their answers are almost the raw output of that training. None of the quality assurance things that we might wrap around a Large Language Model so we can use its output directly (like for example a grounding database which restricts the model to only using reliable vetted sources) are in place.
So your colleague should feel free to ask any (or many) model(s) the question they want to understand, giving plenty of context but without disclosing private information. To help get good quality answers they should try using some prompt engineering techniques like GCSE (Goal, Context, Source, Expectations) to help the model give them a quality answer. And they should use the response as a starting point – follow up the references, ask the model where they should get more information, and use their own judgement and experience to get the full answer to their question.
Skills and approach
Some questions went beyond the technology and towards how we can, as accountants, approach data strategy.
One might know what they want to achieve but not have the technology background to know how best to achieve it (what products, piecemeal, end-to-end solution etc.). How can an accountant bridge this gap?
There are two answers to this question: CPD and experts. From July 1, all of us as ICAEW members will have to be ‘technologically competent’ in order to meet the requirements of the Ethical Code’s updated definition of professional competence. The way we do this, as recommended by this standard and required by the ICAEW, is CPD. If you know what you want to do, you’re ahead of the pack.
Seeking out training (including considering the training the ICAEW offers!), reading broadly from specialist information providers like Gartner or the trade press, and speaking to fellow ICAEW members and peers in other firms and businesses can be enormously helpful. When assessing vendors, it’s very helpful to have an up-front checklist of what your requirements are that you can measure against to avoid falling for something good-looking but not really suitable for your needs.
Of course, at some point your needs are likely to go beyond what you as an accountant might be expected to understand and be able to do yourself. The standard doesn’t require that we all become database engineers and data scientists! At this point, it’s time to bring in an expert. This might be an external consultancy, a particular vendor partner who you trust and who meets your needs, or hiring someone who does have the skills you need.
At this point, the critical skill for the accountant is knowing how to hold the expert accountable, and how to ask the right questions, which is fundamentally a question of the kind of governance we’re all familiar with. Some businesses and firms might balk a little at paying for this expertise if they’ve managed so far without it, but it’s best to be realistic in this space. If you needed legal advice, you’d pay for a lawyer. The risk with data and IT is no lower than it is with legal compliance, so not paying for the help you need is a false economy.
Any views on use of "Digital worker" solutions being offered, and when might be the right time (data clean-up) to engage with providers of this?
As with the above answers, it’s about knowing your own requirements and then matching these against what the solutions are offering. What is causing you the most pain? Where do you lack the skills or resources? What’s the benefit of bringing in a third-party versus building your own capacity? In this space it’s worth going out and buying things because you need them, rather than shopping because things are for sale.
Where do your thoughts lie re customer led or data team led models? I'm of the mind it should be customer led but getting a lot of push back from the data teams.
Your business or firm exists to serve your customers or clients; if you don’t meet their needs, you’ll find yourself without revenue – not a good place to be! Having said that, your customers almost certainly don’t understand the way your business functions or the data you have and use in the way your data teams do. No one is seriously suggesting you ask your customers to suggest the database architecture you should be using.
The best-practice approach to this is to get an understanding of the problems you need to solve from the customer’s perspective, and then take your expert data teams’ recommendations as to how you should solve those problems. If you let the data teams set the entire strategy you’ll build something enormously convenient for your own teams which makes serving customers very difficult indeed (and encourages people to go outside the processes). But if you ignore your data teams’ advice on how to solve problems you’ll alienate some of your experts and end up with solutions which are impractical in your environment (encouraging people to go outside the processes). Make the best of both sources of information.
