CloudTrade Podcast - Episode 2 - Trusted data
Watch the podcast here, if you’d prefer to read about what was discussed, please read on.
In their latest CloudTrade podcast, David Cocks, CEO, and Steve Britton, Director of EMEA Sales, debate the concept of “trusted data” and why this is such an important and vital tool for the C-Suite, particularly the CFO/Head of Finance. They highlight the inevitable difficulties faced by companies using OCR and robotic data capture, and explain how the CloudTrade hybrid, rules-based solution gives 100% accurate results – to truly deliver data you can trust.
Steve: Morning David. Welcome to this first CloudTrade TV podcast, a second one that we have done in the series. It’s been a lovely weekend, I don’t know what the weather was like with you down in your part of the world, but we had beautiful blue skies and sunshine and it was quite warm over the weekend, so hopefully a good start to spring.
David: That’s right Steve, we took my little dog for a walk, he’s only 6 months, a good pandemic pup, and he’s got kennel cough this morning so he’s in isolation.
Steve: Oh no!
David: Either that or COVID-19.
Steve: Hopefully not!
David: A dry cough in his throat. So he’s either got COVID-19 or kennel cough.
Steve: Oh dear, well I wish him well and I know that you and I have both had our first round of vaccines and are slowly but surely building our immunity we hope.
So, this is as always lovely to talk to you. The point of this series is to look at challenges we have come across in the industry with our customers over the many years that we have been working together in the area of digitalisation and business process automation. One of the areas I wanted to focus on today was what I call “trusted data”. We are all aware of big data aren’t we. That has been around for a long time. Businesses today have many challenges around the constant change in IT. Looking at document processing, we focus on invoices and orders as the two areas with which businesses who come to CloudTrade require assistance. These businesses have the challenge of asking themselves, “Do we do it on premise or do we outsource this?”.
The market is constantly changing. But one of the things we recognise all the time is that documents come in different formats, different shapes and sizes, with different content, so there isn’t a constant there. As I mentioned earlier, IT constantly changes and one of the challenges I often hear from executives in business is they just want data they can trust; data that they know when they process it in their business is going to meet their business demands. As I mentioned before CFOs can outsource, but at the end of the day CFOs can’t outsource their responsibilities.
They have to put their returns into the tax authorities, to their shareholders and so on. So, having the ability to receive business documents, human-readable documents, and process them with accuracy is critically important and that’s where I come to my “trusted data” question. So one of the things that I wanted to debate with you this morning is “Why do capture solutions, which have been around a long time, and there are some very good solutions out there, why do you think they find it so hard to capture data in a way a CFO could actually sit back and go: I can trust that data?”
David: Yes, Steve. I think the answer to this is in two parts. I think one of the problems is, and you are right there are a lot of solutions out there, but an awful lot of them are quite legacy. One of the problems is the way they do the capture, which is really right at the root of this. They usually use OCR, most of them still do, where it is unnecessary now in a data-rich document, and that by definition introduces errors. Even the very best OCR systems – and some of the most modern OCRs are really good – used on anything but pristine documents make errors. They generally also use generic searching techniques, and at best this gets 80-90% of the fields that they pick up from invoices correct; so where you have ten fields on the simplest of documents, one of them is going to be wrong. You are almost in the situation of building-in the fact that every document is going to be wrong somewhere.
Steve: That’s interesting David. I suppose the challenge of what does that mean to a business then, if you’ve got inaccurate data? The fact that you can’t trust it, as a CFO in a business, I guess you are going to have to employ people downstream, to check that that data is accurate, because you are not going to publish false information. But just explore, if you would, some of the challenges of having inaccurate data. What does this mean to a business?
David: Well, at best, and this is going to be really at best, something downstream captures the inaccuracy and you have got to have somebody to correct it. That’s the best thing that can happen. The worst is that it corrupts some downstream business process. That could be something as straightforward to understand as you just pay an invalid invoice. Not great. You pay one of your suppliers the wrong amount. But it could be even worse. It could be, I think you mentioned early on Steve with SOX Compliance and regulatory approval it would mean you are reporting invalid data. Your numbers are wrong. Your business numbers are wrong. If you have invalid data, what we call at CloudTrade false positives, it is really, really bad news.
Steve: And I suppose when you think about all the press we have had over the last couple of years around robotic process automation, if you have data that you present to a robot that is fundamentally flawed, I suppose you have a dumb robot in this regard. You have a robot that can’t process information correctly if it is presented with inaccurate data.
David: RPA can always enter incorrect data into the screens. Most robots are just screen scraping and entering stuff into screens, and an invalid amount whether it’s entered by a robot or a person, is still the wrong number.
Steve: Absolutely. Interesting. I suppose the other challenge – we have got the capture of data and the accuracy levels that we have just been debating – but actually, in the real world, do documents coming in always contain the right information that is necessary to process the data downstream?
David: Well, no, they don’t. And I was going to say early on Steve, you’ve moved on a bit, that the capture is the problem of solutions, but also it’s not only about the capture, it’s about making sure the data is right for the end system. No invoice, that I have ever seen, has the receiving party’s vendor ID on it, so by definition if you are going to automatically process invoices you have to add the vendor reference. Now that’s a very simple example but with sales orders, for example, if you have got inbound sales orders, it’s very, very likely that your customer sees or describes what they are buying in a different way to the description you have in your fulfilment system. There has to be a mapping between what your customer thinks he is buying and what you are actually selling. You either do this by hand, or through automation, because your fulfilment system is not going to understand your customer’s description of your goods.
Steve: Yes, interesting. I suppose if you think about it in a human process way, you or I would open an envelope and look at the piece of paper we are reading and say, “Ah this is a customer order”, as an example, to take your analogy, and we know what they generally order from us, and therefore we apply some logic. So, coming back to the question, “How do we get to this ‘trusted data’?”.
With the different shapes and formats in which the data arrives in businesses today, the majority of it being electronic, what is the answer then, taking what people know in the business about, “This is what I should do with this document”, and the way that data is delivered to us today, is there an answer that we can go back to a CFO and say “Actually this will give you trusted data so that you don’t have to touch it, it will deliver to your backend ERP system” – what we call in our world first-time post rates – i.e. You will deliver accurate data. Is there an answer?
David: Yes. And that is what we do at CloudTrade. It’s about using the correct tools for the job. As I always said, to make everyone laugh, I said hammers are great for banging in nails, but they are not any good for screws, so don’t use a hammer to put in screws. RPA is about repeatable simple processes. It’s not about coping with the variability you get on human-readable documents like invoices and purchase orders. You are applying the wrong tool, as great as RPA tools are, you are applying the wrong tool to do the job.
What we build at CloudTrade is a tool that we build on a mix of AI, both natural-language processing and machine-learning neural networks, but also with the ability to program-in bespoke business rules. And it is that combination of what the machine can learn, the simple parts, plus you need human knowledge, captured in the form of business rules, which gives you the solution where you can get that 100% data. That’s what we are about.
Steve. It is, absolutely. So, David that’s really interesting. Using that rules-based approach with a human-assisted front end, as it were, to provide end-to-end automation. That is something that we are very proud of in CloudTrade and which we have been delivering now successfully to our clients for the last 10 years. So, interesting debate. I hope there is some interesting information for our listeners today that will be useful for them in their challenges around addressing their own business processing, but we will be delighted to hear from anyone who has challenges, or indeed requires to have data they can trust in their end-to-end business processes. So, thank you again for your time today, David, and looking forward to the next podcast that we will make next week.
David: Yes. Thanks Steve, and I am off now to see the dog who is in isolation at the moment.