Recognition Engines Don’t Die… They Just Fade Away
As I was planning for our 2013 OrboGraph Client Conference, I was reviewing materials from a past years conference, and have to admit, was a little disturbed on a few memories! I remember that conference like it was yesterday. Beyond the excitement of a conference in Salem, MA around Halloween, and a really fun theme called “A Supernatural Opportunity from OrboGraph”, three very distinct topics gave me flashbacks:
- Many attendees thought that check recognition performance was “good enough” for distributed capture applications like teller and branch capture. Remember, in 2002, we had just introduced OrboCAR Accura and Super Trinity had been around for a few years already.
- Teller capture would make branch and centralized capture obsolete within a few years (which we now know is more like a 15 year window).
- OrboGraph was proclaiming, as others had been doing in the industry, that a pure algorithmic approach for check recognition had reached the “practical” asymptote. As you can guess, this was Avikam Baltsan’s presentation, where in analytic geometry, an asymptote of a curve is a line such that the distance between the curve and the line approaches zero as they tend to infinity. (Check out this Wikipedia link if you want to relive your high school geometry days: http://en.wikipedia.org/wiki/Asymptote) What does this mean in layman’s terms? That the investment to achieve incremental recognition performance could not be attained without a major investment, making it unpractical from a business perspective.
In December 2012, OrboGraph will release Accura XV Version 1.2, a scheduled upgrade to Version 1.1. Accura XV has been a huge success, as we’ve upgraded hundreds of customers. Accura XV also powers all of the new Automation Services flavors. Version 1.2 includes a number of handy improvements, which make the system run faster, with less footprint, for a wider range of industries and with more features. And, as we plan for 2013, there will be an Accura XV 2.0 with better recognition rates, increased accuracy and…
So what’s going on here? Why would business professionals 10 years ago claim that recognition was pretty much maxed out, and expectations were set for read rates of 70-80%? We had clearly reached a plateau.
Customers Drive Demand, Which Drives Innovation
Although check recognition had temporarily reached a plateau, innovations in handwriting and OCR continued in other areas. For example, the first tablets hit the market with only modest OCR success, but companies continued to drive improvements because customers demanded more, and were willing to pay for it. Today, the advancements in not only tablets, but OCR on mobile phones, illustrate the major steps recognition processing has taken in the past ten years. During this time, certain products essentially became outdated, and either lost market share or were discontinued from technical support (They faded away).
In 2012, we discovered that customers were willing to pay for improvements and value when measurable and significant. So although we did plateau for a few years in performance, essentially the market needed to adopt the technology; which practically spanned for seven to nine years.
Today, OrboGraph continues to “find a way” to inject technologies and innovation into its core product line; check recognition using CAR/LAR/OCR/ICR and many other technologies. We leverage this technology to guarantee 90% overall read rate performance with check read rates in the mid 90’s. Customers seem to want more… so that’s what they will get in 2013!