Let me get the following statement out of the way before I begin my rant: most people have very little understanding of what IDP is let alone how to go about achieving great performance – and this includes many of the vendors you’re considering. Just sayin’.
So now to that supposedly simple form you’re trying to automate and having a hell of a time getting the level of performance you expected.
Simple form data capture should be…simple, right? It can be, but there are a lot of gotchas involved when you apply real-world situations to your planning.
For instance, what do you think about this claim form? Nice right?
What about this one? Hmmm.
The first example was scanned using grayscale which can capture more of the details on the page. For instance, we can easily see the difference between the pre-printed text on the form and the actual data we want. The second one is more typical of a scanned image – using a bitonal setting otherwise known as black-and-white. When those settings are used, it is a bit harder to discern between the form and the actual data.
Now take a look at this one:
That’s the same as the first example, right? Nope, this one has a slight tilt to it, known as “skew”. Skew is very common with scanned images and, if not corrected for it, can introduce problems with locating and pulling the data you need.
So far I’ve only shown you the easy stuff. The reality with any scenario where you are dealing with scanned documents is that the amount of variation across a lot of factors makes even a simple structured form a bigger challenge than you think it should be. For example, check out these real-world claims and the amount of variation:
All of these types of issues conspire to create a situation where reliable automation of even 10% of these “simple” forms is incredibly difficult. Admittedly a CMS-1500 claim form is a hard case as there are a lot of fields and they are packed densely together. But the same problems arise again and again regardless of the form. And if you are considering as part of your scope forms that are captured via a smartphone, well… I think you get it.
So you have a number of challenges even for the seemingly simplest project: dense form structures, the amount of information (bitonal, grayscale, or color) you have to work with, and then all of the image quality issues that get introduced by the various ways the forms are scanned.
What to do? Most organizations just deal with it by manually entering the data that cannot be accommodated via their existing automation. The organizations that do attempt to automate as much as possible of these real-world scenarios have to spend a lot of time creating various rules and templates to accommodate the differences they observe. Note the italics because with a traditional approach, you can only account for what you notice; few organizations have the time or interest to analyze even a small percentage of their data.
So back in that real world, a lot of organizations are only achieving minimal levels of reliable automation on even the simplest (there’s that word again) forms.
Something has got to change. And it has. Check this out:
Contact us to find out more…