As I wrote last month, EU Parliament recently voted for a resolution to get the largest traffic generators to pay into network upgrades and expansions. The purpose would be to help telcos keep pace with growing traffic demands and to meet the EU’s 2030 Digital Decade targets.
At first glance, this proposal sounds reasonably straightforward and simple. But how service providers would measure, allocate, and bill the largest content providers to recoup some of the billions they invest is not so simple. Would they divide by traffic volumes, content types, subscribers served, revenue gained, or other metrics?
To come up with a practical, transparent, and democratic process, there are many variables to consider: who pays; what is charged; who gets paid; how to measure usage (upstream and downstream); what to charge; OTT-related Opex and Capex; how OTTs would pay.
Below I look at these factors, and then make some rudimentary recommendations for how this could possibly work.
Question 1: Who pays?
We have to start with “Who.” Who will pay, and who will get paid? Neither question is easy since so many parties are involved.
In one model, any party sending a packet to any other party on the internet could pay. Of course, that would not be practical, simple, useful or democratic. So let’s strike that option from the list right away.
In another model, the OTT players making the most money from their content would pay. The problem is that revenues don’t always correlate to traffic volumes or network costs. There are OTTs whose millions of users have almost no impact on network resources, and others whose few users trigger massive network usage.
In considering who pays, it’s also important to consider the different ways in which OTTs generate revenues. Some profit from advertising, others from subscriptions, and yet others from both. And then there’s the issue of non-profits, like Wikipedia – should they pay or not?
These considerations make this option less tenable as well.
So, let’s think about charging by usage – is that an option? It would address the fact that just a few of the world’s largest OTTs use the lion’s share of network resources, driving much of the demand for bigger, faster (and hence, more expensive) networks. But how would you measure usage? It can be measured by tonnage (total volume over time), downstream or upstream or total tonnage, or upstream and/or downstream throughput. The de facto standard, however, is to use 95th percentile of five-minute average throughputs.
It's also important to consider splitting upstream and downstream. Downstream is by far the dominating direction for most networks (not all, of course, as that would be mathematically impossible). Radio networks typically dedicate 90% or so of their frequency domain to the downstream direction. Some access types, like ETTH or FTTH, are typically symmetrical, but other access types have a significantly more expensive and limited upstream. For that reason, you could make a case for counting and charging for upstream and downstream separately.
Splitting by direction makes further sense if you think about the applications themselves. Some apps are upstream-biased, like Dropbox and other cloud-based backup services. Others are symmetrical, like Facetime, Skype, or other communication services. And yet others are heavy on the downstream, like Netflix and other streaming services.
The 95th percentile method has the big advantage of ignoring the worst peaks, as well as ignoring the times when the network is idle. This matters because you have to build capacity for the peak periods, so the OTTs and apps that are most active during the peak period are what will drive your overall cost.
If there were such a thing as an “off-peak app” that generated lots of volume during the non-busy hours of the day, that app would not drive any cost at all, or at least not CAPEX. Every packet and byte transferred on a network carries OPEX costs – e.g., electricity, licenses, labor, etc.
Conclusion for who pays: After considering all the factors I describe above, it seems the top “sources” on the downstream, and the top “destinations” on the upstream should pay for their 95th percentile consumption/production.
Non-profits are not among the top OTT providers for any network that I know, but if they were, I’d assume no telco would feel right taking money from a non-commercial organization.
Question 2: What is charged?
In order to know what to charge, do you first consider the source? If it’s Alphabet, do you separate Google Mail from Google Maps? If it’s Netflix, do you distinguish download from streaming? What about Netflix gaming, or the act of browsing the titles?
If we ask ourselves whether Google, Microsoft, and Apple should be punished for being big, and for being behind the names behind so many services, the answer would be “no,” as that wouldn't feel right.
If Microsoft buys Wikipedia, should Wikipedia traffic be included in the cost model just because it falls under Microsoft? That honestly doesn’t seem to make sense to me.
Conclusion for what to charge: I land in a place that says it’s the “App” that counts. Netflix is clearly an App, as is Google Maps, as is Apple’s AppStore. The “big movers” of throughput during peak periods is what costs the operators the most, and what’s fair is to have traffic generated from and to those apps become a part of the costing.
Question 3: Who gets paid?
The concept of who gets to benefit from a “fair share” model is also non-trivial. There are many types of access networks out there: satellite, cable, DSL, Fiber, mobile, etc., all of which have different usage burdens and cost structures. As you’ll see in the table from the Carmel Group’s 2021 Fixed Wireless and Hybrid ISP Industry Report, there are different cost structures to consider (though these are U.S. based, and used here only for illustrative purposes):
Some of these networks are built primarily for enterprises, others for consumers or government. Some are for the last mile connecting to the consumer’s home, and yet others for interconnecting continents, or interconnecting telcos.
OTT traffic could conceivably impact all of these networks, and the companies that run the networks could possibly own more than one type.
Considering OTT-related Opex and Capex
The Opex and Capex costs associated with the OTT traffic for each network varies greatly. Some have a large Capex component, like digging a fiber trench for miles to connect a household, or launching a satellite into geostationary orbit. Others have a bigger Opex cost associated, such as electricity for powering a mobile network and pumping radio waves into the ether.
There’s no rational way these costs will ever be fair to the network operator. For those “unlucky” enough to run an expensive network, there will have to be acceptance that they're going to get paid less. If it helps, it can be argued that it’s less expensive and better for the environment to optimize for less power-hungry and resource-intensive networks, like FTTH.
The other dimension is to examine where the majority of CAPEX investment goes. It’s not going to interconnects, sea cables, packet cores, or distribution centers. Rather, the majority of the cost for “running the internet” goes into “last-mile” connectivity to consumers. For mobile networks, that means cell towers and radio networks; for cable networks that means fiber or coaxial cables connecting buildings.
Conclusion for who gets paid: When we home in on the subset of operators that should get the majority of the OTT contributions, it’s those who have subscribers – enterprises or consumers. If you have subscribers, then you have “last-mile” costs.
The scope of this point could be further narrowed to “consumer networks," as enterprise networks have a very different usage profile in that it is not normally dominated by OTT traffic.
To look further into the OTT costs to EU operators, Deutsche Telekom, Orange, Telefónica, and Vodafone asked Frontier Economics to do a detailed analysis. The result was “Estimating OTT Traffic-Related Costs On European Telecommunications Networks,” and its data suggests the total costs attributed to OTTs in Europe are as follows:
Because of these incremental costs for supporting OTT traffic, the EU Parliament voted to support the European operators’ argument that OTTs (specifically large traffic generators) should participate in funding network infrastructure, which, as I stated early on, is essential to achieving 2030 Digital Compass connectivity targets.
Question 4: How much should OTTs really pay, and how?
For the sake of discussion, let’s make some assumptions. For example, say Netflix is 13.7% of fixed networks, and 2.4% of mobile networks (source: Sandvine 2023 Global Internet Phenomena Report). Using costs from the table above, that would mean 340 to 530M euros per year in incremental fees for Netflix. Whether the OTT bears all of the cost and through what mechanism is yet to be decided.
Since all networks are different, there’s no way to tell who the top OTT providers are on each network without continual measurement.
Also, there is no way for the OTT to know where it ranks on each network. Each network operator would need to implement a method to count the 95th percentile apps for downstream and upstream, using that information to support the billing process, which will have to be defined by regulators.
It is possible that aggregate app provider-based Usage Detail Records can be generated, with the app owner validating the send/receive information through its own records.
With so much to think about and consider, it’s critical that regulators consult with industry experts. This will better the chances they choose the right mechanisms. It will also better the chances they build a process that is simple, fair, transparent, democratic, and useful. Action on this must be taken soon, to ensure the sustainability of the networks that are increasingly central to all facets of life in Europe and globally.
If you're a service provider currently thinking about these questions and possible solutions, feel free to reach out to me on LinkedIn or schedule an appointment with me at Digital Transformation World in September.