[I am in Las Vegas for the TAWPI Payments in Transition conference – these are my notes from one of the sessions. An index of all of the sessions and links to the rest of my notes is here. – EMc]
Image Integrity—A Critical Component of Check Electronification
Kerry Atha, Director of Product Development, Viewpointe
Perry Bailey-Kopp, Group Vice President, Check Production Support, SunTrust Banks, Inc.
Financial institutions are continuing their rapid adoption of check electronification, raising new concerns about maintaining image integrity—ensuring that images and their associated metadata files match. Should mismatches occur, financial institutions could face considerable risks with significant financial and customer trust/public relations repercussions. Speakers will present the challenge today’s financial institutions are facing in maximizing image integrity, a critical component of quality assurance; explore the financial and PR/customer satisfaction ‘fallout’ that occurs when mismatches happen; and detail the actions FIs and the industry must take to preserve the integrity of image presentment and storage.
Image integrity = correct match between image itself and its corresponding meta data. (Vs. Image quality which has to do with the image itself and was the subject of another session this morning.)
Exponential growth in image exchange in the last year. Thousands of banks, not just the big ones. Volume data from ECCHO and Federal Reserve (slides 3–4). Anticipate that 2008 will bring tipping point (more electronic checks processed than paper checks).
Kerry Atha, Viewpointe spoke first:
Ability to "share" vs. exchange item when both participants are in the same exchange.
- Image quality – characteristics of the image
- Image usability – legibility and completeness of info
- Image integrity – wrong image associated with transaction data (can be a privacy issue if wrong check image is displayed to customer)
Initial industry focus was on image quality. How to translate subjective qualities to objective measurements that everyone can agree on, and then develop software tools to check quality of images.
Image integrity is the focus of conversations between image exchange partners (particularly at beginning). They made an effort to understand one another’s processing environments, efforts made to ensure quality of images and integrity of data files. But as image exchange grows and more and more end points are added how do you evaluate the quality of the incoming images and data files?
Image quality software checks 15 parameters. But only 3 (too light, too dark, piggy-back) actually affect pay/no-pay decision. 90% of suspect items are false positives. Thus the risk tool introduces a new level of risk as operators viewing thousands of images that are actually okay are more likely to accidentally clear/approve images with genuine issues.
Perry Bailey-Kopp, SunTrust spoke next:
- 4.3 million items per day
- 82% in clearings as image (36% growth in 2007)
- 60% transit items cleared as image (39% growth in 2007)
- 2.8 million items average daily exchange volume
- 1st bank to implement image exchange with the Fed
- 1st bank not blocking any RT for exchange
- 1st bank to receive controlled disbursement via image
- In house for ICL and paper capture
- 5.3 million items via online deposit (remote deposit capture)
Banks are selling distributed capture products. Uncertain of quality of images and data-image integrity from remote deposit end points. Rely on Viewpointe to do the quality review for SunTrust. Leave images at Viewpointe and access the data only. Access images for Day 2 processing, as necessary.
Fed submitting images, too. Same challenges with remote capture sources. Quality not as good as Viewpointe, although Fed is working on it. They have a lot of end-points feeding into their system, too.
Your archive is as critical as your posting applications. If archive is corrupt (mis-matches) then your DDA data is corrupt, too. Must be very careful.
Is it necessary to subject all items to IQA/IIA (Image Quality assessment, Image Integrity assessment testing) each day? NO. 3890 reader sorter does a good job. When there is a jam, run IQA/IIA on items 30 items before, 30 items after. Run test against all rejects. Thus, only 2% of items at SunTrust subject to IQA/IIA. So far, only 7 mismatched items – all due to operator responsible for image quality review (same operator!).
Image Cash Letter (ICL) – customer scans checks and generates their own image deposit file. SunTrust runs all ICL files they receive through IQA/IIA because have no idea of imaging, process control within their customers’ operations.
Distributed Capture – SunTrust is moving toward a distributed model to replace their proof operation. [Note – this is the strategy that I worked on for Wells Fargo nearly ten years ago – way ahead of its time; alas it wasn’t implemented due to the merger with NorWest.] Distributed capture utilizes smaller scale equipment and CAR/LAR, replacing painful Image POD and its inherent problems ("I got my butt kicked by Image POD" – Perry). Power-encode turned out to be difficult: keying worked well, balancing worked well, but could not encode fast enough and get paper out of the bank. But this time around, not having to move the paper (due to Check 21), distributed capture is the way to go. Per SunTrust, if we need a paper copy we can print an IRD.
In 2 years, transportation will be the highest cost factor of check processing. You don't want to be the last guy processing paper!
Image Quality and Image Integrity metrics for SunTrust:
- Transit items = 0.19% suspect rate; 5,500 average daily
- In clearings = 2.23% suspect rate; 24,000 daily (includes account number edits for on us items)
Administrative Returns (NCIs)
- Sent back to Exchange Partners by STI = 0.0140% average daily volume 65
(90% for image integrity issues)
- Received by STI from Exchange Partners = 0.0049% average daily volume 54
(99.9% source document issues)
Return item – wrong RTN. When this happens, someone has to find the check. Then transport it. For a $12 check?
Do bad check images (due to puppies and flowers on the check) really matter? If you can read account number, RTN, and dollar amount go ahead and process even if image is a mess. Day 2 is always looking for reason to reject – there might be a problem if we pay this. Per Perry, just pay it and find out! The sooner you do so, the more likely you are to collect your funds.
Image Quality is Out
- Customers are looking for reasons to pay, not to reject
- Majority of FSTC standards, not utilized for decision making: only too light, too dark, piggyback are material.
But if quality is out, Integrity is In
So many players along the way as items are passed from bank to correspondent bank and via vendors. Fed Receipt has so many end-points that there are more quality/integrity issues. Financial Institutions have confidence in their internal capture environment. But what about corporates using their own process/equipment and data from downstream banks, remote capture?
Mitigate risk: Capture bank has presentment risk, and paying bank has privacy issue. Lack of familiarity with trading/exchange partners.
Presentment compliance – ECHHO – presentment on the image (not the data). But systems are set to process on data.
Image integrity will impact future growth of image exchange; particularly as more and more end-points are added to the network.
How often does integrity issue occur? Hearing stats that range from 1 in 5000 to 1 in 50.000. Privacy breach remediation costs anywhere from $50 to $250,000 but that doesn't take into account the PR, operational impact due to damage control. No one wants to show up on the front page of the WSJ for violating customer privacy.
Banks and vendors need tool that is robust enough to catch integrity issues. Yet, impractical and expensive to evaluate every single item. How to target IQA/IIA tools to high risk items only. Systems need to have the flexibility to target a subset of items, in various locations (including at corporate customer sites).
Parameters for image integrity are much more straightforward than image quality. Difficulty is in determining how to interrogate the data and the image. But not possible to evaluate 100% of MICR data vs image. Need to optimize matching routine and logic. What subset of MICR line is enough (balance cost effectiveness vs. introducing add'l risk)? Matching and logic is not transferable from software platform to software platform. Optical character interrogation dependent on software. So it is difficult to develop a comprehensive solution.