Pre-paid Services – Time to join the 21st Century?
For some time now, we have all been hearing the convergence story, multi services all billed at a single point with discounting and loyalty points multiplied on the basis of cross-product consumption. However, when we dig a bit deeper, it appears that really this only applies to post-paid services, as the limitations around calculating call/event costs in real-time mean pre-paid services are effectively still stand alone.
There are good business reasons for this, but charging technology and the cost of technology should not be the barrier to enhancing these services.
So how and where does this get resolved? Is this a software or hardware issue? Traditionally the nearer real-time a cost has to be calculated, the less complex the rate plan had to be, as the cost of hardware either as part of an IN or a RADIUS/DIAMETER aware BSS seems to rise exponentially with the volume of calls or the complexity of calculation. This model isn’t sustainable for operators as they aren’t able to pass the cost onto the consumers of their services.
From some investigation we’ve doing, we see new technologies starting to exist, in-memory databases, increased transaction per minute without building server farms, faster processing algorithms amongst them. However it’s not clear which innovation will come to dominate.
I’d be interested to hear other people’s views on this, as it’s always interesting to be on the cusp of changing technologies or business models.
And I haven’t mentioned roaming………………..
Comments
Hi Ian,
Interesting questions. While in-memory database technology has provided some boost to performance, it only helps in the simple case where there is not any pricing complexity such as bundling, discounting or sharing. As soon as these concepts are introduced, traditional IMDBs get bogged down in the same way relational DBs do. Existing real-time technology not only results in expensive per transaction processing costs, there is also no predictability in the performance from one rating scheme to another. The business would be better enabled to roll out bundles, sharing, discounting, etc. to prepaid subscribers if they could predict exactly how the additional complexity was going to impact latency and transaction processing efficiency. Any new technology that dominates this area should not only be much faster, but also more stable and predictable with respect to performance and cost per transaction.
This really seems to be an issue with the software rather than the hardware. Over the past 40 years, hardware technologies (CPUs, memory & disks) have followed Moore’s Law and increased in performance by up to one million fold! Unfortunately, inherent limitations in the traditional OLTP model, which remains unchanged since the 1960’s, have prevented rating and charging solutions from following a similar path. A modern Intel CPU can perform almost 200,000,000 decimal math operations per second, yet most real-time rating engines can only process 15 – 150 events per second on the same CPU – the inefficiency is staggering. To really take advantage of the power of modern hardware and create a charging platform that can scale efficiently with the explosion in mobile data, a fundamental redesign of the underlying OLTP model is required.
Commented by: Dave Labuda | March 23, 2010 1:43 PM