As some of you might know, our work on nuisance call prevention is a project I’m leading personally. We have such an awesome technology team nowadays, under the watchful eye of Charles (CTO) that it has been many years since I’ve needed (read: been allowed) to tinker with the flashing lights and, to be honest, I’ve missed it. So I love it when I can go back to my roots, upskill and build something cool. That’s what we’re doing with nuisance calling.
Just like our work in VoIP Fraud back 10-15 years ago, we’re following the data and making some interesting discoveries. While others are virtue signalling, yet doing nothing to upset the gravy train, we’re putting our money where our mouth is. While others are lecturing about what they don’t know or understand in an effort to sound relevant, we’re trying to know and understand in an effort to combat this relentless onslaught of shite. Without going into too much detail about what we’re doing (I feel a webinar coming on for that), it is revealing some interesting things that I thought I’d share with you.
Pushing regulatory boundaries
Some of our discoveries have caused us to engage with Ofcom and, I have to say, I’m so far impressed. I don’t say nice things about them often and generally shy away from the talking shops that seek to educate them towards a pre-determined goal (e.g. “we need a central database, because <latch on to problem de jour you know nothing about>”). However, we’ve come across use cases in actuality, and counter-measures in theory (as Pete espoused last week) which needed grown up consideration.
The reality is that if we were to turn on measures we have to fully enforce the rules and guidance we have today, we’d block an awful lot of traffic. That might be good, but many CPs have built businesses blind to some of the rules that exist and now unwittingly rely on nobody else complying with them. Doing so for us is technically trivial (and a button press away) but would frustrate a large number of our customers. Rather like blocking fraud, wholesale customers don’t see the £100k they weren’t scammed, they see the calls that we rejected and they (naively) sent to another carrier who creamed the margin at a much higher price. It is a thought process I don’t profess to understand but can just about reconcile. In today’s terms, blocking all calls from a nursing home because an incumbent sold a service which breaches the General Conditions causes more harm than the rule is meant to prevent. Put simply, if the nursing home was calling one of your customers, it would be no surprise that you’d leave Simwood to go to a slower moving operator who doesn’t care and calls could resume. We are acutely aware of that balancing act.
Before we can turn on a lot of what we’ve built, there needs to be an enforcement of existing rules to ensure the collateral damage is minimised and we look forward to working with our regulator and other relevant stakeholders on some of the quandaries that are resulting. As well as a deep-clean of networks to meet long-standing rules, we hope for greater clarity on some of the terms the existing General Conditions and underlying Guidance rely on which, when you really start defining rules around them, can be interpreted multiple ways.
The good news is we have data. We can see which self-proclaimed “global” operator has lax KYC and whose numbers are used everywhere. We can see which UK carriers are transporting that traffic. We can see patterns of behaviour which are in breach of the numbering plan. We can also see which of the existing rules are responsible for over 90% of blocks, and they’re unambiguous. And way more.
We need to use that data for the wider-good because, frustratingly, other regulatory obligations stand in the way of us treating traffic from a given rangeholder differently to another, even if we know that their traffic is 70% likely to be spam. Or that calls over a certain virtue signalling peer’s interconnect are 130x more likely to be spam on a revenue adjusted basis – although those calls appear to have been re-routed after the last blog mentioned it (but, unbeknown to them, those calls still get blocked, because we use calling behaviour, not the identity of the interconnect to determine action)!
Unlike for Google when they all but killed email spam for Gmail users, the regulatory road is going to be longer than the technical one, but we’re excited to engage with Ofcom with their current and future work streams and we hope our customers will be too.
A little matters
We see about 800,000 unique CLIs on the Simwood network every day. Many of them are good and represent the traffic that feeds our families. Some are bad. With the nuisance filters quite loosely set to avoid false positives (and losing the goodwill of our customers on this journey), we blocked just 100 unique numbers yesterday. However, those 100 CLIs represented 1.93% of the traffic on the network, calls in to many customers, calls out from some. As with fraud, others might see that as 2% revenue lost, we see it as old ladies saved and value delivered to our customers. I should add this is 1.93% over and above the traffic we’re already shunning through measures our peers have yet to get to, e.g. requiring valid CLI.
I don’t know if nearly-2% is an impressive figure or not? If 100 people walk down the street and you save 2 of them from being mugged, that feels worthwhile. What matters ultimately though is how many would be mugged without you? 3, 10, 40 or maybe 2? We need to keep cranking the handle and finding where that natural level is but in the mean-time it doesn’t take a lot to make an impact.
Statistics are irrelevant
In an engineering-led industry, we like statistics and it is fair to say all existing efforts to keep networks clean rely on them. They’re important, but unidimensional and insufficient in 2025.
And this is a problem. In any telco of any size, complex analysis of calling patterns will be problematic. For years it plagued me that I couldn’t do the kind of analysis I wanted against our CDRs, even if we had huge database instances there for little else. If you scoff at this, just remember you have no idea how huge what I want to know is! In 2025 we have the cloud, and with the heft of Google or AWS we can run complex queries over a month’s call records. The query might still take 20 hours of computation to run but the beauty of the cloud is that that 20 hours is available in under 20 seconds due to horizontal scalability.
The result of this is that the answers to questions which transcend simple statistics become possible. You can have a hunch, and follow it, with results arriving before you’ve forgotten what you were asking or, more often in my case, got distracted with something else! That is where the insights arrive.
Legacy telcos like to think they’re too big and complicated for <insert innovation>; I simply say you’re not bigger than Google or Amazon and this approach can scale to anyone in this sector.
I like to think statistics is how many calls a given number has had, or simple metrics such as what % were successful, while data science is the why. Intercepting a call and listening to the media is the brute force way of understanding why, but is intrusive (and is not something we currently do). It is probably ultimately part of a solution, but in my opinion, the bar for applying it has to be set rather high, i.e. you should be pretty certain the caller is a scammer (or nuisance) before going there. The only way to get to that point is to follow the data but doing so has to go way way beyond simple statistics. How do you tell if a service with 5 second calls and a low answer rate is an energy scam or a GP reminding customers of appointments? Those not answering, recognising the call, and already knowing, and those hearing the reminder hanging up quickly? With statistics alone they look the same, but as we’re discovering the data is there.The data paints a picture.
Webinar
We have a lot of data to share here for the empowerment of our customers, and ultimately for the protection of your customers. We won’t be producing a how-to guide, nor will it be as deep as we would give Ofcom, but do want to share some findings amongst close customers. If you’re interested in learning more then please register below and we’ll send you an invite when we’ve set a date.
