I’ve been thinking a bit about operations science and quality assurance for a few weeks now.  I think there could be some parallels to performing loss prevention.

If no one counted the cash at the end of the day, how much would go missing?  The answer: you can’t tell if you’re not measuring it.

So one must at least collect some information.  Great, now you have an idea of how much might be missing.  If you collect that data over time, and compare it, you may also find out if you’re doing better or worse.  But one must go further to actually get to the root cause and determine accountability.  This is where video is an important complement to one’s point of sale data.

If one can consistently demonstrate the ability to detect and correct a problem, then the corrective action provided is not simply a momentary act of containment.  In operations science it’s most desirable to create an “irreversible corrective action” (ICA).  The goal of an ICA is to create a solution to a problem that eliminates the problem in such a way that it cannot reoccur.  However an “ICA” can be invalidated when change occurs.  So when dealing with change, (such as new employees, new sales promotions, new application features, new processes and so on) the corrective action may need to be modified or be re-introduced.  Quality control and loss prevention are a continuous process.

If you just install cameras and a POS monitoring system, but never look at it,  at best you won’t know if you’re reducing “shrink”.  At worst, a reduction in shrink is likely only temporary as people figure out that it’s not being used.    If you want continuous prevention, your employees need to know the system exists and see it being used.  You need a easy, quick, repeatable process you can continously apply.  One that tells you what to look for and where, so you don’t have to spend too much time searching.

If there’s one thing I need in the office, it’s headphones. Headphones piping music straight to my cerebral cortex. One reason is because it can get pretty noisy and when you’re heads down on a problem, trying to tune out a conversation can get distracting. Having said that, I wear headphones even when I’m working from home and it’s dead quiet.. so perhaps that’s just what I tell my boss and I’m really just a slacker.

Sometime in 2008, I discovered a music service that you may already know about called Pandora. It’s a free service and they have a really nice iPhone client, which is key. The beauty of Pandora is that based on your preferences the system plays music it thinks you’ll enjoy.. and it’s a beautiful thing. I’ve been turned on to a ton of new music using the service. To me, it’s magic.  Some people have speculated that somewhere down the road this kind of application know-how will lead to this:

Image via Wikipedia.

Or this:

Image via Wikipedia.

Personally, I think it’ll be more like this:

Image via Wikipedia.

We know it’s not all magic, that there are humans categorizing and tagging Artists/Albums/Songs all day long, that there’s metadata ‘the system’ uses to make recommendations.

Having ‘the system’ make recomendations to you based preferences is something that makes a lot of sense. Using MVaaS and getting a point of sale feed is a great way to do it because it doesn’t involve humans tagging files. You’ve got your POS terminals feeding ‘the system’ metadata about your video.  This is all you need to get started alerting your organization to things you might like to know about.

If you’re a retail operator selling a new widget wouldn’t it be cool to know when you’ve sold X number of widgets? Even better, wouldn’t it be cool to see the video so you know who’s buying what?  On the other end of the spectrum, perhaps you’d also like to be notified when a store executes a particular number of transaction voids.. oh, and here’s the video..  and hey, it’s the same guy and there’s some slight of hand going on. Envysion’s reporting and alerting is something we’ve been hard at work on and we think you’re going to like it. Find some interesting video you want to share? Check out our groups. I’ll get to that in another post.. down the road.

In the mean time, we’ll leave busting the bad guys to the humans.

Fall is my favorite time of year; the leaves are changing, it’s cooling down and most importantly it’s football season.  For the past few years I’ve anticipated the start of the season with much more enthusiasm.  The reason for this new found enthusiasm has been my involvement in a fantasy football league.  Fantasy football, takes something that I already have a passion for and makes it even more exciting by giving me a reason to be involved in several games in order to track my team’s performance.  It also gives me the opportunity to put on my GM hat and make decisions that can help or hinder my team’s performance. 

Effectively managing a team requires time, effort and an understanding of information (e.g. trades, injuries, trends) that is crucial to a successful season.   A GM has a lot of information to dissect each week in order to give their team the best chance to win.   The first couple of seasons I found it difficult to keep up with all of the trades, added/dropped players, injuries and the progress of free agents, which one can add to their team if someone on their team has a bye week or isn’t performing.  Finally, I activated a feature on the Fantasy Football web site that I had never used before, “Alerts”.  The second that I enabled this setting I became a much better GM and could make decisions that would affect the outcome of the game days in advance.  The alerting function of this website takes information that is being gathered about each NFL player and proactively alerts (sends and email) each GM based upon criteria they have designated as important.  A GM can be alerted to set their line up, player updates, league trades, and scores.  This information can be vital to the planning of your starting lineup each week and if acted upon can give your team a competitive advantage.

MVaaS acts in much the same way as the Fantasy Football website.  Through the integration of software based business applications e.g. POS systems, access control, temperature sensors etc. an owner or manager can set up reports with specific business rules, that if broken, can generate an alert that proactively prompts a user to log in and review video from a specific location.  For example if I have an MVaaS solution integrated with my POS system I can define a rule that if a site has more than ten voids in a day I want to be alerted.  This type of capability can come in very handy when making tough decision about your Fantasy Football team, and more importantly your business.  I find it to be a great asset to be able to make decisions based upon up to date information.

I am not an expert in loss prevention.  I have only been in the video world for the last three years.  Despite my limited experience, however, I have picked up a fair bit about how our customers use exception reporting and video to help aid their loss prevention efforts.  The other day I was having a discussion with a partner of ours about whether real-time alerts generated by an exception reporting system and tied to video would be valuable for loss prevention.

The partner’s hypothesis was that if a customer (or in this case a partner on the customer’s behalf) could get a real-time alert of a specific exception transaction that was likely a theft, the customer/partner could immediately look at video, respond and get the offending cashier off the register, thereby immediately addressing the problem and stopping further losses.  Sounds like a good thing.

The impetus for the conversation was that the partner wanted us to develop real-time alerts, where our current process is analyze business rules for any violations on a regularly scheduled basis, not in real-time.  My initial response to the partner was that if they wanted real-time alerts, we could provide real-time alerts - such is the magic of a SaaS-based business in that we can make changes to our service like this on the fly.  After thinking about it awhile though, I started to question whether this was truly something that would provide the benefit that our partner was expecting to provide.

I started to think about how you would actually use a real-time alert based on POS data.  The first thing I would question as valuable is a real-time alert that is based solely on the occurence of a specific individual type of transaction, such as a line-item cancel.  First, there are likely to be dozens or more instances of any specific transaction type in any given day at every single store, most of which are valid transactions.  An operator or someone monitoring the service would not be able to keep up with all of the line item cancel transactions that happened in a single day if they got a real-time alert every time someone rang up a transaction and canceled an item out.  They’d be overwhelmed and start ignoring all of the alerts, which defeats the purpose.  Second, if there was a transaction type that was such a red flag and happened so infrequently that it demanded an immediate response, do you think anyone who was trying to defraud the system would actually use that type of transaction to steal?  Why would the company even let that transaction type be used in that case?

Exception reporting systems themselves work more with summary data than they do with individual transactions.  At the highest level, they crunch data over some period of time and look for anomolies or violations of specific business rules.  They then enable the user to drill down into specific transactions to investigate.  A mundane example of a business rule might be to look for any cashier that has more than 15 line item cancels in a day.  A loss prevention team might investigate every time this happens.

So could a real-time alert be useful when one of these business rules is violated?  Again, I’m not sure how much.  Assume someone is really stealing from the register by misusing the line item cancel (voiding an item that was paid for in cash by the customer so that they can later take said cash out of register without having a cash discrepancy).  Maybe every third or fourth void they do is bogus, so they are stealing multiple times a day, but not every time they use that transaction type.  The business rule wouldn’t be triggered until the 15th void though.  At that point someone would get a real-time alert and could pull up video of that person and see… nothing related to the offending transaction.  The customer would have left, the employee would be standing behind the register.  The reviewer would then have to go through all 15 transactions (the last one that triggered the alert is no more likely than any of the others to be good or bad) to evaluate whether the employee’s actions were legitimate or whether some portion of them were inappropriate.  The reviewer would then have to make the determination of whether that employee should immediately be confronted or whether further investigation is required.  All of this takes some time and a real-time view of the store is not that relevant.  If you react as quickly as possible to the incident and intervene with the employee, you keep the employee from entering any more fraudulent transactions during that day - although it is likely near the end of their shift anyway unless they went crazy and entered a bunch of bogus transactions early in their shift and you were able to review them all very very quickly.

The question I have is how much value is there in getting that alert in real-time vs checking for those business rule violations multiple times a day or even nightly? Faster response is better - no question.  If the backdoor opens after hours or there is movement in the office when there shouldn’t be - I get the need for a real-time alert as you may prevent a crime in progress.  In the case I just described though, responding in real-time doesn’t buy you that much over a regularly and frequently scheduled review.  In most retail cases responding in an hour or 30 minutes versus instantaneously would be virtually the same.  To me, the bulk of value in exception reporting is identifying and eliminating theft that occurs in small amounts over long periods of time.  If you can stop someone from stealing from you after a single day rather than taking weeks or months to catch them you have saved a tremendous amount.  Given the investigative process required however, I don’t know that real-time alerts (versus just frequent alerts) is a huge value creator in this pursuit.

Having said all of this, I return to my first sentence - I am not an expert in loss prevention.  I could be completely missing the boat.  The good news is that I know people that are experts and that if there is value, it is definitely something we can (and will) provide.

We’ve often wrote about how sharing video and data can be a powerful tool, especially when that information can be trusted and is easy to get at.

Another instance of using information sharing over the Internet is a work in progress in the IETF.  Authored by some folks at Verisign and Entrust, the IETF Internet Draft “Sharing Transaction Fraud Data” proposes a way to share intelligence about fraud incidents over the Internet.

Efforts like this should lead to easier access of sensitive information sharing across organizations .  It’s also interesting because it’s perhaps an indication of increased willingness for organizations to share information in light of procedural, trust and legal considerations that will have to be overcome to make such sharing possible.

1. issue a request to search for video to the MVaaS server. Specify store (e.g., DVR), register (e.g., camera), and time of day. The MVaaS server returns a list of clips it found
2. initiate a video player in your web browser
3. instruct the video player to play the clips that were found in step one.