Speaking Engagements & Private Workshops - Get Dean Bubley to present or chair your event

Need an experienced, provocative & influential telecoms keynote speaker, moderator/chair or workshop facilitator?
To see recent presentations, and discuss Dean Bubley's appearance at a specific event, click here

Wednesday, December 22, 2010

Why application-specific policy and charging will (almost) always fail

Over the last couple of days, the technology blog and Twitter world has reacted with righteous anger to this slide and the underlying sentiments around Net Neutrality, especially given the FCC's announcements recently.

I don't think Openet and Allot have done themselves any favours with that webinar chart. Not only was it guaranteed to cause a storm of controversy, it's all made 100x worse by the fact that that sort of app-centric policy / charging mechanism simply will not work, either technically, commercially or (probably) legally. There are so many workarounds, pitfalls, gotchas and risks that any carrier would be crazy to even try it, especially in developed countries.

On the other hand, it's the sort of slide that will make some mobile operator marketing people salivate, especially the more gullible and technically-ignorant among them, so I guess I can cut the two vendors some slack. In all probability, there are some operators which will try to do this sort of thing anyway, so if they're going to waste their money, then I can understand the vendors thinking "well, they might as well waste it with us rather than the other guys".

[Sidenote - if you work for an operator and want to get a detailed view of why this is a *really bad* idea, please get in touch with me to arrange a workshop]

I'm supposed to be on holiday & am waiting for a rebooked flight, so I'm not going to give the long list of problems. But just to get the debate going, let's start with:

  • Customer dissatisfaction & churn
  • Likely regulator disapproval & censure
  • Stuff inside VPNs
  • Mashups (what happens to the YouTube video embedded in a Facebook page?)
  • Who pays for *uploads*?
  • Acceptable level of "false positives" and the legal liability associated with that
  • Need for network ops guys to *really* understand applications & how they evolve on a daily basis
  • Lack of anybody with the job title "policy manager" who grasps and has authority over all the necessary bits of the puzzle - radio, core network, app trends, devices, user behaviour, legal issues, PR, partnership etc
  • Multiple fraud possibilities
  • Fit with CDN architecture & difficulty of charging Akamai
  • Probability of key Internet companies turning the tables on the operator and say "no, you pay *us* or else we're going to block access to your subscribers, suggest they churn, and give free adverts to your competitors"
  • It gives implicit permission for fixed operators to charge your customers to use your femtocells on their DSL/cable
  • Doesn't reflect signalling load in the RAN
  • Difficulty in predicting future direction of applications. Is Facebook's new email service going to be classified as Facebook, or email? What happens if Facebook buys Skype?
  • Tons of arbitrage opportunities
  • Lack of decently flexible upstream billing & reconciliation systems
  • Need to have perfect coverage before charging for incremental services
And plenty more.

Now most of the anger is around the general philosophy of Net Neutrality and whether operators should be "allowed" to do this sort of thing. I'm actually fairly, er, neutral on that. It's simply that this sort of approach is woefully flawed, and in any half-competitive market will lead to immense customer dissatisfaction and churn.

It's not wrong or evil. It's just naive.

Now... that doesn't mean that *all* application-based policy use cases are going to fail. There are some which I can see having value, or which are less prone to outright failure. I can see "zero-rating" of specific partner apps being valuable, because that will encourage the provider to collaborate with the telco. Although it's probably easier to just put those in a separate APN rather than sift through the main data-stream. I can also see the arguments of rate-limiting of P2P traffic for congested cells, especially uplink. In the fixed world, I can certainly see the role of this for specialised services which are use the broadband, but which are not Internet-based (eg IPTV or home security or specific cloud apps).

But the idea that an end-user will pay extra per-application, for major Internet services like Skype or Facebook, on their broadband access? Utter nonsense. As I wrote yesterday in my 2011 predictions earlier this week, there's a huge opportunity on the horizon for Policy and DPI. But the slide on Engadget is 100% the wrong way to attempt to monetise it.


Martin Geddes said...

Yota disagrees with you: http://www.techdigest.tv/2010/12/free_internet_f.html

I think there are a multiple 'sender party pays' models that are possible, some of which work, and some that don't for the reasons you list.

Dean Bubley said...

I've spoken to Yota on several occasions, including its CEO, and I don't think we're that far apart on almost all issues. They are about as close to being a "happy pipe" operator as I've met.

There's a big difference between *sponsored* "free" broadband (which is already done by governments for some deprived regions) and paid-per-MB or paid-per-signalling-event services.

Similarly, I can easily see certain apps/services being "zero-rated" or "discounted" against quotas and caps. (This already happens in fixed networks, and on mobile via different APNs)

These are all things I covered in the Broadband Business Models report I published with Telco 2.0 earlier this year: http://disruptive-analysis.com/new_bband_bus_models.htm


Unknown said...

Completely agree. If Facebook was charged for, it would take someone about a minute to come up with a proxy.

Anonymous said...

I believe there is a business case here for Telcos.

Instead of imposing application specific policy/charging and risking subscriber churn, they can turn the tables and give the power to the user instead.

Referring to this tariff plan by TATA-DoCoMo in India:


TATA-DoCoMo has provided users with a choice to subscribe for unlimited browsing plans for Rs 10 a month for specific website/s.

There are special plans for social networking sites as well: http://www.tatadocomo.com/pps-tariff-plans.aspx

This way,they are making money as users are willing to pay a fixed amount for websites they visit often (assured revenue per month per site subscribed for) while not showing preferable treatment to a particular internet service (maintaining net neutrality).



Unknown said...

The application-specific policy & charging is not only for charging the 3rd parties' applications, e.g. Youtube, facebook, etc., it also helps the operators with customized services to end-users and new business model. The mobile broadband business model should be very different from the traditional PSTN broadband because of the always limited resources (wireless bandwidth). DPI & PCC is the most important methods for this policy control.

Martijn Brouns said...

Wrt service dependent charging, I think you have a point. It would be as if the mail man would charge you more if a package would come from Amazon rather than you mom.

Err... on the other hand. If I order 20 books at Amazon I DO pay more than that xmas card I sent to mom right? and if I take a flight to LA, I DO pay more if I bring 4 suit cases. And... don't I pay my cable company more when I watch all premium channels on my PVR?

Okay, okay... the internet is kind of different. The whole concept of mashups is kind of a complex one. But if 5% of users consumes 80% of bandwidth, I'd say it's a fair motive to seek for ways to have people pay for what they actually consume.

If Net Neutrality prevails, I guess we will all need to pay more for mobile broadband. In the end, things that get scarce get more expensive.

What's more, it is not only bandwidth that gets scarce. Also PDP Context (active data contexts available per base station) are limited.

With an ever growing list of devices (such as smartphones, tablets, etc) that setup 3G connections every now and then, you won't always be sure to keep an active Skype session if your 50 neighbors like to watch videos, play games, send BBM texts, etc too.

Perhaps, ultimately, services that eat a lot of bandwidth and make money when using valuable IP access will need to close SLA's to make sure a user gets a decent Quality of Service. In the end... infrastructure is NEVER for free.

Dean Bubley said...


There is a difference.

If you take an extra 4 bags on the plane, it uses more fuel even if the flight isn't full.

But if your video download means a cell runs at 60% capacity (air interface, RNC, backhaul etc) rather than 10%, there is near-zero incremental cost unless you're on a Pacific Island and have to haul data across the ocean.

It's not about "consumption", because the cost of "not consuming" is the same to the operator for an uncongested network.

Yes, you could say that a heavy user "should" pay more than a light user - but by the same argument surely a remote user in the countryside should pay more than a city-dweller?

The ideal would be some form of charging that maps onto the *actual* congestion risk caused by your behaviour.

So signalling-rich email downloads & Facebook IM in a busy Tube station should cost more than a video in the suburbs at 3am.


Martin said...

I'm late to this party, but we spent considerable time thinking about this. I think Dean are right on almost all points. But to me, the opportunity for service based charging is when it is about whitelisting, not blacklisting.

Take a mass market 1mbps mobile broadband plan. Keep youtube, running as-is (which, more often than not, is a so-so experience in many mobile networks). But sell a youtube premium package for $5 a month, that gives Youtube more speed, added features, etc. Can be done with or without Youtube's consent, but better with.

Whenever policy is about punishing or blocking, people will quickly find ways to go around it. But if it is about giving premium lanes, the incentive is smaller (and much easier technically to control).