Next Generation
Check Recognition Technologies
For Teller Image Capture &
Remote Deposit Capture
Written By: Joe J. Gregory,
Vice President Marketing
March 2013
Introduction
New recognition technologies continue to be ‘invented’ for the
financial industry, particularly for teller image capture (TIC) as well
as remote deposit capture (RDC). “Next Generation Recognition
Technologies” will review the key technologies now available
which enable image workflows to move from a state of average
performance to exceptional.
Early Generations vs. Next Gen
Generally speaking, when first generation products are introduced
to a new market, the primary objectives include:
First generation
products are focused
on confirming the
business case,
achieving stability
and penetrating the
market.
•
•
•
•
Prove out the concept and confirm a value proposition
Achieve product stability
Penetrate the market with awareness and use cases
Provide attractive packaging during the product introduction
stage to achieve overall market growth
TIC and RDC both followed this path during the mid 2000’s,
bringing innovation to distributed capture. The focus during this
period was on the overall deliverable, rather than targeting
implementation of best of breed components.
Related to the check recognition component, early adopters were
less concerned about read rates and misread rates, and most
focused on proving out the workflow. The fact that many
implementations only achieved read rates in the 60’s or 70’s was
overlooked for the bigger picture.
Now that platform stability and added value functionality for TIC
and RDC has been achieved, it is time for new TIC and RDC
implementations to take advantage of “next generation” check
recognition technologies that solve various business problems
associated with these workflows.
Solvable Business Problems
#1 – The promise of “straight-through-processing” in a check
processing workflow is very difficult to achieve due to capacity
limitations in toolkit vendor’s CAR/LAR (courtesy amount
recognition/legal amount recognition) engines. Although
exceptions, including rejects and unreadable checks are now
passed to tellers and merchants, the desired process is to
eliminate human intervention. With technologies that deliver
performance levels in the 60’s, 70’s and even 80’s, most
implementations are still not even close to this goal.
Once the
infrastructure is in
place for TIC and
RDC, more advanced
concepts can be
pursued to address
several remaining
business problems.
#2 – Balancing continues to be a problem for medium sized and
large deposits. When items are misread or miskeyed, an intensive
balancing process can be required, starting off with a manual
review of most items. To put it into perspective, many recognition
systems deliver 2-3 misreads per 100 items. It is problematic
enough to locate these 2-3 items, but when a batch of 1,000 items
is deposited, that’s 20-30 that must be corrected!
#3 – Fraud and operational risk are major considerations when
looking at TIC and RDC. Today, the ability to run real image fraud
detection for on-us and transit checks including automated
signature verification, check stock validation and alteration
detection provides protection against counterfeits, forgeries and
alterations.
#4 – Eliminating over-the-counter paper completely is not yet
realistic, but there are ways to dramatically consolidate and
minimize the number of documents with new recognition
capabilities.
Impact Points
Let’s assume for the moment that new generation recognition
technologies solve the aforementioned business problems. The
benefits are simply impactful; both tangibly and intangibly. When
looking through the eyes of a cost analysis, savings can be over
$1M per year for every 1M items processed.
There are many
impact points which
are improved with
Next Generation
Recognition
1. Smoother customer experience
a. Faster customer transaction time
b. Additional interaction & cross sale
c. Greater customer satisfaction
2. Reduction in teller effort
a. Reduce teller keystrokes and maximize efficiency
b. Detect missing endorsements & signatures
c. Easier training for part timers
d. Less IQUA false positives
e. Minimize rejects and exceptions enterprise-wide
f. Improve the employee experience
3. Large and medium sized deposits
a. Reduce “provisional” transactions and adjustments
b. Reduce balancing frustrations
4. Operations
a. Superior IQUA = Less NCI
b. Research and adjustments savings
5. Management
a. Staffing and high volume branch optimization
b. Manage & monitor the performance systematically with
minimal overhead
c. Optimize staffing models
6. Support/IT
a. Less issues with teller computers for recognition rates
b. Easier to control and verify
c. Push out changes seamlessly
7. Fraud
a. Reduce on-us check fraud in cash checks and split
deposits
b. Reduce deposit fraud on transit checks
Technology Behind Next Generation Performance
The recognition vendors in the marketplace continue to provide
incremental functionality and improvements. However, the true
jump to Next Generation performance is predicated on these
capabilities:
1. Multi-engine correlation with new voting and merge schemes:
Each recognition toolkit utilizes somewhere between 3-10
recognition technologies depending on how they utilize
CAR/LAR/ICR/OCR/ALR, etc. These unique technologies or
algorithms need to be aggregated and profiled based on the
strengths and weaknesses of each set of algorithms. Next
Generation recognition utilizes a sophisticated data mining
approach to profile each engine and merge the results
together to a single, usable score which minimizes misreads
and maximizes read rates.
The result: nearly 100% recognition performance on personal
and business checks for smaller transactions.
2. Dynamic thresholding: This approach utilizes new engine logic
to on-the-fly modify thresholds based on transaction size.
A unique approach
is needed by
implementing new
technologies to
reach Next
Generation status.
The result: reducing misreads to 2 per 1,000 items, which
dramatically reduces the number of errors in mid-sized and
large deposits, thus improving balancing processes for TIC
and RDC.
3. Item verification and “rebalancing”: This technology
improvement leverages detailed multi-engine results to identify
and flag potential items which are misread or miskeyed.
The result: a reordering process can be implemented to
streamline balancing even further.
4. Check box detection: Internal bank tickets are challenging to
manage and can be very expensive. Although teller image
capture does reduce some percentage of internal documents,
the ideal scenario would be to consolidate to a single
document for branch or teller image capture. The Universal
Teller Document concept can achieve this by assigning
specific tran codes to “check boxes” on the ticket.
Conclusion
Next Generation recognition solutions provide a wide range of
benefits including cost reduction (drive down unit costs), efficiency
improvements, improved employee experience, increased staff
utilization, fraud reduction and reduced paper.
However, the most important benefit is an improved customer
experience for both Teller Image Capture and Remote Deposit
Capture.
Feel free to contact Orbograph at info@orbograph.com or
joe.gregory@orbograph.com for support on any of these
concepts. Orbograph has extensive experience and provide cost
analysis models to help quantify these potential benefits.
www.orbograph.com
Phone: 800-995-2502, Extension 5042