What is the most important feature of an Intelligent Document Processing (IDP) product? Hint: You already know but you don’t spend enough time on it. Before I get to the point, let’s discuss how IDP software is different from most other enterprise software.
Features & Functionality – The Hopeless Scorecard of Enterprise Software
We’ve all been there, either from the buyer or seller side. A problem is identified within an enterprise, a team is put together to scope the problem and identify key requirements of any resolution. Then a shopping list is constructed designed to vet one vendor’s offering from another. Commonly referred to as an RFI, this uber checklist often goes to great lengths to ascertain the level of capability of a given solution. For document automation needs, we often see questions on architecture/scalability, user experience, administration capabilities, reporting, and even how we use machine learning within our product. More detailed RFIs even inquire about the types of APIs provided, document classification algorithms employed, and whether or not support for tag-value pairs is provided.
Most of the questions are aligned with the ability to quickly verify the presence or absence of a given solution attribute. Certainly, all of these questions are geared towards real needs, and some are critical. For instance, if an enterprise needs a SaaS offering, an on-premise solution is out of the question.
But while many enterprise solutions can be easily compared with one another based upon verifiable attributes such as those within RFIs, when it comes to IDP software, one attribute stands head and shoulders above the rest: the ability to achieve high levels of reliable automation. That’s it. Anything else is secondary as no single organization would spend any amount of money on an IDP solution that could only produce single-digit or even low double-digit automation. Organizations want 100% automation at 100% accuracy, but that is highly unlikely in any scenario. This is where the “accuracy question” comes into play but oftentimes enterprises treat the attribute of accuracy as just another easily answerable line on a spreadsheet. The reality is that it is far from easy.
Why? Unlike organizations that need sales force automation, or a CRM system, or even a procurement system for which any number of solutions can readily solve problems out of the box, no two organizations are alike when it comes to the range of document-based information they manage or the types of tasks they wish to automate. While there can be similarities from one organization to another, when it comes to the discussion of accuracy, a lot more detail is involved.
Did You Know You Can Get 99% Accuracy and Achieve Almost 0% Automation?
You heard me correctly. Claims of 99% accuracy are all over the place yet they provide very little information regarding the automation potential. For example, if you need medical records identified and sorted by service date, it is entirely possible to configure a system to achieve 99% accuracy. Yet this level of accuracy might only be applied to 5% or less of your overall number of medical records. The upshot is that the true value of IDP software relies upon TWO factors:
- the number of tasks that can be automated (think document splitting, document identification, or document indexing) and
- the percentage of those tasks that have a high level of accuracy. Multiply those two numbers together and you get a baseline for complete automation. Without one of them, you only have a very small part of the entire picture.
So why do most organizations only ask about accuracy and why do most vendors gladly only talk about accuracy? There are many answers to those questions. Regardless, organizations that do not prioritize overall automation potential by focusing on verifying both task automation rate along with task accuracy rate are doomed for mediocre project results. A SaaS solution heavily vetted to ensure the user experience is optimal and that scalability is always managed offers nothing if it cannot deliver on its true value.
Want a playbook for how to measure automation performance in IDP and conduct a proof of concept? Take a look at this one.