Marketechnics Report: In-store vs. Host-based Systems
By Bill Bittner
The typical supermarket technical environment has been an in-store processor integrated with a point of sale system that managed the customer checkout process. The in-store processor was a general purpose computer that ran applications like labor scheduling, inventory management, backdoor receiving, time and attendance, etc. The point of sale system was most often an IBM 4690 operating system-based application that supported dual processors in order to provide total redundancy for the front end. No one wanted to be unable to checkout customers.
This environment has had its drawbacks. Locally available processors provide quick response times and give stores a level of independence if something happens with host communications, but they also require the management of software releases and the availability of hardware support personnel when equipment fails.
As seen at this year’s FMI Marketechnics conference, the Internet and the widespread implementation of high-speed data communication have opened up an alternative to in-store processors. Host-based applications operate with only “thin clients” in the store. Much as you use an Internet browser to access Internet applications, stores can access applications that have been designed to run off of a host processor. This leaves them vulnerable to a communications outage, but they gain all the advantages of not having to maintain software and equipment in the store.
Moderator’s Comment: We understand that all applications are not “mission critical,” but how far can the hosting revolution go? What applications are
you willing to relegate to a host system, which must be in constant communication? Would you ever give up the POS function?
Personally, I feel that, as the store becomes more dependent on the computer for all aspects of its operation, the question is not, “how do we keep the
front end operating during a hardware failure,” but rather “how do we keep the store operating?” It is no longer just the POS applications that are mission critical. Everything
in the store must be available all the time. This means that there must be either locally available hardware or host communications that are failsafe. With today’s technology,
I believe that either is possible and, although it is an emotionally challenging choice, I believe sufficient communications backup can be provided to run even the POS applications
from a host based system. –
Bill Bittner – Moderator
The interim step between store autonomy and reliable, centralized Internet hosting will be “host nodes,” with each node hosting 20 stores. Host nodes will be counter-intuitive in a non-regional way, with each node serving stores from every geographic zone. That way, a communication catastrophe affecting a single node will not impact an entire geographic cluster of stores, but individual stores in each geographical zone. Thus, the chain remains in business in most of the stores in each zone.
I think that I can agree with both Mr. Bittner and Ms. Murray. I would also point out that due to the tremendous amount of capability that is being pushed down to especially the POS systems, they are becoming increasingly more difficult to update and upgrade. However, simply because it can be done (hosting), doesn’t mean that it is practical or the best solution as of yet. Let’s just say, it’s heading there.
There are still a tremendous amount of regional concerns regarding the level of communications capability and continuous connectivity. That being the case, maintenance of two types of platforms would be even more difficult. So, in the meantime, as the regional infrastructure catches up, the in-store solution remains the ‘not best’ but reality solution.
Sales_Pro’s recommendation is probably ideal, but it is likely the most complex to maintain. The hosted solution will have significant advantages in deployment and maintenance costs, and if architected properly can provide the same local autonomy as a standalone system. Done right, the system should be more reliable than a local one, as less can go wrong in the local environment where it is less likely there will be someone who can fix it immediately.
That said, stores should always have a well-defined backup plan for continuing operations in case of system failure. A documented “disaster recovery” plan can be as simple as a binder with detailed instructions for how the store should be run during an outage, and all managers and staff should receive some training on it.
I would strongly vote for a store-based system updated via internet. With this type of system, pricing can be controlled by brand/category managers but overridden (with permission) by a local manager. This would allow for actual marketing at a central office but would allow local competitive defensive efforts. With this type of system, my local grocer would have been able to reduce the price of the (normally $12.99 doz) roses he offered for sale for $36.99 doz on Valentine’s Day to a competitive price of $19.99 doz. Instead, the grocer had to suffer the humiliation of having the most expensive “roses” in the market on Valentine’s Day.
Internet delivery will win in the end. It takes a massive amount of manpower to keep individual sites running, upgrades made, software supported. It may not seem so, because it happens site by site. But applications distributed over the internet have been gaining ground year after year after year because of the sheer efficiency of distributing updates and upgrades.
That said, there is still a fair amount of resistance. Companies fear possible virus attacks the more they open their systems up to the internet.
I’m not convinced about the “communications outage” concern. We’re all completely dependent on telephones and cell phones, and we don’t stay up at night worrying about a communications outage there.
The hosting revolution has just begun and those caught betting against the internet have not fared well. Even IBM, with the most to lose, has moved to web services and would recommend any ‘greenfield’ installation be web-services centric and as lightweight as practicable. In as few as 3 years heavy POS will be dinosaurs, as both merchants and shoppers will benefit from r/t discounts, couponing, replenishment and rewards.
Host based systems can be extremely helpful for release management which is a pain area for most retailers today. We are even witnessing the transition to host based systems in the pharmacy world (which is fairly primitive when it comes to technology adoption). There is talk of T1 lines connecting stores providing redundancy of similar bandwidth at major retailers which would make host systems realistic. The advantages of centralized data can also be leveraged easily.