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 discuss Dean Bubley's appearance at a specific event, contact information AT disruptive-analysis DOT com

Friday, November 11, 2011

The smoking gun - I think O2 UK has FALLING mobile data usage

Following on from my earlier post about whether mobile data usage is flattening off, I've done a quick bit of forensic analysis and modelling about Telefonica O2 UK's mobile data traffic statistics:

From Q1 2011 report: "Data traffic from mobile broadband accesses however continued to grow with total volume increasing 45% year-on-year. Despite the removal of further heavy users to ensure the best network experience to all customers, both data usage per customer and mobile broadband penetration continued growing"

From Q2 2011 report: "Data traffic from mobile broadband accesses also continued to grow with total volume increasing 31% year-on-year in the first half."

From Q3 2011 report: "Data traffic also continued to grow with total volume increasing 22% year-on-year in the first nine months. Growth in the third quarter was lower at 7.9% year-on-year following the removal of heavy data users, with all consumer data contract base sequentially growing usage."

The model is slightly complex because we don't know exact QoQ growths for 2010, and we have to reverse them out of the 6/9 month YoY stats and make estimates.

But whichever way I set the spreadsheet up, it looks like O2 UK has recorded falling absolute levels of traffic in Q3 of 2011, and perhaps Q2 as well.

EDIT - one of the commenters has found a set of numbers that doesn't have a fall, but has 3 quarters essentially static at +/- 1% . Might be a one-off readjustment, but it's surprisingly "elegant" if that narrow possibility holds true

We know that Q1 2011 = 45% more than Q1 2010.

So let's put Q1 2010 = 100 and Q1 2011 = 145  . Interpolating, we could have Q1/2/3/4 of 2010 as perhaps 100/115/125/135 or maybe 100/120/130/140 or similar. But then we can determine Q2 and Q3 2011 from the other growth rates reported.

So for example, we get:

Q1 2010 100
Q2 2010 115
Q3 2010 125
Q4 2010 135
Q1 2011 145
Q2 2011 137
Q3 2011 135


Q1 2010 100
Q2 2010 120
Q3 2010 130
Q4 2010 140
Q1 2011 145
Q2 2011 143
Q3 2011 140

Unless you put really silly numbers into the model (eg Q2 2010 = 250, up 2.5x QoQ) I can't see how you get anything but an absolute decline in recent months.

Now this is just one operator in one country. And it's possible that O2 (the original iPhone exclusive operator in the UK) has seen a lot of churn to Vodafone, EE and 3UK (which still has full flatrate dataplans and has been reporting strong data growth). And O2 has repeatedly stated that it has "removed heavy data users" and has apparently also blocked adult content by default, as well as encouraging WiFi. So this could be an isolated example, or a one-time blip before growth resumes.

O2 Germany's stats reported in the same financial statements do not demonstrate an absolute fall, but also suggest a slowing rate of growth. I'm estimating that it's down to maybe 8-10% QoQ growth, based on statements such as "Mobile data traffic also increased significantly (+51% year-on-year in the first nine months; +47% in the third quarter)"

I'm going to be looking very closely for other data points, and more importantly trends over time in growth rates - any other contributions or comments most welcome.



Ram Krishnan said...


Michael Mace had written about this a while back.


Basically, his point is that because of economics and user psychology, he thinks we're headed for a slowdown in the growth of mobile data. The unlimited, exponential growth forecasts are wrong. He believes the ultimate mobile data market will be smaller, and much more segmented, than most people expect.

I believe that tiered plans and WiFi offload will ultimately result in a more measured growth in mobile data - it will be interesting to see if O2 is a special situation or not.

Dean Bubley said...


Yes, I'd forgotten about Michael's articles. He's called Overcapacity before me.

I think I commented on one of them.

He's spot-on about most of the reasons, rationale and implications - as well as many of the core responses.

HOWEVER, I think he's wrong about the idea of toll-free / zero-rated apps. Can't and won't work, for the reason that apps evolve and change much faster than a billing and rating plan ever can. Plus mashups and the nature of the web mean you can't really assign traffic flows to an "application" as perceived by the end-user


Ram Krishnan said...


Actually, growth in overall mobile traffic is less important. What is important is if traffic during peak hours (9pm to 11pm), which drives network capex and TM decisions, is growing in an exponential fashion. What is important is the impact of this peak traffic growth on sector/cell congestion,

Verizon Wireless is implementing a well-thought out (and very well explained, I may add) policy where they penalize heavy users during sector congestion.

DanG said...
This comment has been removed by the author.
DanG said...

Sorry to be a mathematical pedant, but your numbers don't seem to be right to me. I take it you calculated the Q3 2011 number by applying the 7.9% growth to Q3 2010? When I use your numbers and the year-to-date growth numbers, I get a different answer for Q3 2011.

I have come up a variation which fits both the YTD and YoY by quarter figures:
2010 Q1 - 100
2010 Q2 - 123
2010 Q3 - 142
2010 Q4 - 143
2010 Q1 - 145
2010 Q2 - 147.1
2010 Q3 - 153.2

You'll be disappointed to see that this doesn't show any sequential decline! The growth at the end of 2010 is pretty anaemic, which seems to fit with the introduction of O2's data caps (though all previous smartphone customers still had unlimited, customers upgrading the iPhone 4 would be subject to the caps and it makes sense to me that many of the "power users" driving the growth would be the type of people eager to get the latest device ASAP).

The other factor though, that we should probably factor in is that if it is indeed a "seasonal variation" as they claim then some variation should be seen in Q3 2010. Neither your solutions nor mine show that variation. So I'm also sceptical about that.

Of course, if someone spots a problem with my numbers, then this all goes out the window!

Dean Bubley said...

Hi DanG

Yes, it looks like you might have solved the simultaneous equations to get there without a sequential decline.

Although having absolutely flatlined traffic for three quarters +/- 1% seems to be pretty unlikely!

I think we need to have one more data point to solve this precisely. (Or perhaps something anecdotal)

I've also managed to find 100 / 120 / 137 / 144 / 145 / 143 / 148 which gives a 1% decline in Q2 but reverts to 3% rise in Q3.

Either way... it doesn't look "crunched".

There's also this quote from June 2010 "At O2, we’re seeing a doubling of data traffic on our networks every four months" just to add to the maths complexity


jason @ Voip said...

I've just read somewhere that 97% of usage on phones is now for data transfer?

Does that fit into the figures you suggest or is that just pie-in-the-sky?

Dean Bubley said...

Hi Jason

Depends what you mean by "usage" (time? bytes? value? something else?) and "data transfer". And which types of phones, and where?

Certainly the amount of time spend talking on smartphones is diminishing as a % of total time they're being actively used


Mark Heath said...

I think there's a risk looking at O2 UK and drawing more general conclusions.As you'll be aware, O2 has made significant reductions in its data allowances, as I've previously discussed here.


With LTE nowhere in sight, O2 UK does not have the capacity to support high usage. I'd argue that usage is essentially being artificially constrained.

The downside of this is that growth in the penetration of mobile broadband in the UK won't be as high as it could be. Mobile broadband penetration in the UK has definitely been slowing, as discussed here.


Anonymous said...

> The unlimited, exponential growth forecasts are wrong.

Unlimited, exponential growth forecasts are always wrong and anybody who makes them is an idiot.

My job at [an operator that is not o2 uk] involves load monitoring and forecasting for [an indirect proxy measure of mobile data traffic]. What I think I'm seeing in the numbers for Q3 is the shoulder of the s-curve arriving about a year earlier and at a signficantly lower level than I epxected.

I'm not yet convinced whether this is a real indication of demand growth levelling off, or an artefact of some kind of temporary constraint such as network capacity or the iphone product cycle.

Dean Bubley said...

Anonymous - many thanks for your comment "from the sharp end" of things.

I'm also uncertain - there are various reasons why we might see a double s-curve shape (eg massive ramp up of tablet-based video in 2012-13).

There's also probably a couple of one-time effects like the end of flatrate, or the "USB dongle boom" falling off (partly because of more free public WiFi, perhaps).

Either way, I definitely think that a lot of the forecasts need downgrading, either in terms of absolute volume, or at least pushing out the timeline to the right by a year or two.

Thanks again for your input.

Anonymous said...

Fair Usage Policy - data throughput down because people are throttled if they exceed their data allowance. Simple really, needs no investigation.

Dean Bubley said...


Yes - you could be right. But if so, that points to the fact that very basic policy (ie fair-use caps) is all that's needed to solve the data "explosion".

No need for rushing to LTE, or fancy DPI use cases & all the regulatory headaches of Net Neutrality. Operators can improve revenue by doing clever stuff on billing / charging side such as loyalty schemes and so forth.