Call Quality

 

By Simon Woodhead

TLDR: We’re taking proactive steps to combat low quality traffic amongst our customers. This blog explains what, why, and hopefully encourages you to respond positively if we get in touch.

We’ve talked before about how the world of voice telecoms is shrinking, with pretty much every country locking down its borders in one way or another. This takes the form of blocking calls “originating” overseas but with their own country-code on, [dirty] origin surcharges and regulatory racism. 

At one level, this is about protecting consumers from spam and scam calls – a very worthy goal we not only support, but lead on. At another level (and often the accidental result of initiatives nominally to help consumers), the goal is profit – eliminate weaker players, consolidate the market, charge more, rinse and repeat. We rile against this, as you’ll see with our stance on dirty origin surcharges, which I maintain do nothing to reduce spam and scam calls nor reduce call costs to consumers. In fact, they mandate more fakery and encourage more harm as criminals, and our competitors, attempt to bypass them. For example, no longer does that call from “Microsoft” to tell you you have a virus obviously come from India, nor does it show as withheld. No, it comes from a spoofed number in your local area code which instils trust in the victim and avoids a surcharge.

Our friends in the US have, characteristically, been very enthusiastic about all of this and gone bigger and louder than anyone. They have a huge problem, and they have a huge prize for the operators left standing. The market is consolidating at a crazy-fast rate as operators are forced out of business one way or another (we’ll return to this), and the spam/scam traffic is shopping for a home on an industrial scale. We’ve done some analysis and can predict where we’re going to get sales enquiries from based on new entries in the various databases required to operate there now. These guys are spinning up hundreds of companies at a time and have a veritable production line, such that when one of them gets fined or booted off enough networks, they have others clean and ready to go. The scale and fervour is shocking.

One odd thing in telecoms is that crime is proportional to cost and progress. If a call costs little or nothing, and connects quickly, they’re likely to be made on a far more industrial scale than if they’re expensive, but that is going to cost somebody down the line. This gives operators like us a challenge in that our pitch to customers and peer carriers is that we run a clean network; we don’t tolerate spam/scam traffic and if you take a Simwood number for your global HQ you can be confident it wasn’t taking fast food orders 10 minutes before. This feeds through to our rates and combined that makes us a honeypot for this dodgy traffic. We’ve talked before about our KYC procedures (which reject many many more accounts than we open), but we still have a problem: you!

If accounts cannot open directly with us, one step removed is just as desirable (maybe more so as you’ll have established history and other legitimate traffic to blend the crap in with). This may present itself to you as a very exciting opportunity, like most things that are too good to be true do. I often tell the story of a customer who, against my advice, accepted a briefcase full of cash in a supermarket car park for one of these no-lose deals; he’d lost tens of thousands before he even got back to his office. If it looks too easy, too lucrative, too sure, it is likely going to cost you.

How can it cost you? The example above was plain and simple arbitrage but today it is likely to cost you your account with us and other credible carriers. Those less credible people will end up with more business, until they themselves find that they have no peers and are extinguished themselves. This is what is happening in the USA and, whilst Europe is way more laid back, it will happen here too.

In all our interconnect agreements in the USA (remember we’re a licensed CLEC in 21 States), they stipulate strict quality standards. Breaches of these quality standards in the first instance lead to eye-watering financial penalties; the second and increasing instances lead to accounts being terminated. No questions asked, no repercussions, no business!

I’m not about to write a how-to guide, and these standards vary by market, but in the USA they are very, very fine-grained to catch gaming. Strict single-digit percentage quotas for specific failure codes, for example – something you might not even measure, let alone monitor. Whether you send traffic directly to an in-country operator like us, or dump it on an international aggregator, at some level these will apply. Nobody wants this traffic, nobody can escape these constraints forever, even if in the short-term there might be specialist middle-men who have accounts about the place to dump it. 

As well as KYC, we’re upping our game substantially in dealing with customers who have unacceptable traffic quality. This will irritate some I’m sure because our competitors don’t bother, but that is a short-term game that will end in tears (rather like not spending anything on their network in a decade). If you hear from our support team, please take it seriously because we’re already on page 6 of a manual that sees you paying substantially more by page 8, and having no account by page 9. To put it bluntly, we won’t risk our other customers’ businesses to be the bridge to a quick buck for you or someone taking advantage of you.

We’re monitoring numerous metrics but those we can tell you about will be of no surprise to those who’ve been in this space for a while:

NER – Network Efficiency Ratio. This is closely related to ASR (Answer Seizure Ratio) which shows the percentage of calls which complete. The difference is that while ASR includes all failures, NER excludes those you can’t control such as the user being “busy”. We monitor ASR too, but NER is very important and quality customers maintain in excess of 90%. Those who have illegitimate traffic will race towards 0 but so will those who routinely retry failed calls numerous times (as will those who only send us calls which have failed elsewhere). These are all things to address if we’ve reported low NER to you.

ACD – Average Call Duration. When calls connect, how long they are, on average, is significant. If you’re scamming people, most will tend to be short as they hang up on you. Legitimate traffic will be longer, with some customers averaging over 15 minutes. 3 minutes is reasonable, 1 minute is acceptable, 10 seconds is neither!

SDR – Short Duration Ratio. This is the percentage of calls which are very short. We’re not going to say exactly how short, but it is single digit seconds. A high percentage of calls which are very short, i.e. an SDR above single-digits, is a red flag. 

Now, one reason for not giving specific figures is that scammers will often quote them (i.e. make them up) to convince sales-people that their traffic is just good enough. Others employ modules which can be programmed to achieve them, e.g. holding short calls open longer carrier-side to avoid penalties. The main reason though is that it comes down to intent and use case and isn’t so black and white, as Pete blogged about recently. We acknowledge that customers in financial services who are calling about debt arrears might have a lower than normal NER, but we would expect them to have decent ACD and negligible SDR. By contrast, our customers sending medical appointment reminders will have a very high NER but low ACD and SDR. They all inter-relate and different use cases will generally trade one for the other. Criminals will be bad on all three!

That’s why we’ve introduced and are reporting on (internally at the moment) what we call the Simwood Quality Index (SQI). It is simply the (NER – SDR) * ACD and gives an index value for the desirability of the traffic. The highest customers are in the high hundreds (decent NER, negligible SDR, high ACD), whilst there are some with negative SQIs (SDR is greater than NER) who are heading rapidly to termination. In between are many where they may look bad, but if we understand the use case, we can live with it. We haven’t got as far as determining that a specific score is the boundary between good and bad, but we know the worst offenders and will be making contact.

For the record, we’re doing this globally across accounts but also by market. Our level 3 support guys have the means to report instantly on your traffic if you feel proactive, but if you’re the reactive sort, please please respond positively if they get in touch. Not only is cleaning up your traffic to us going to affect our relationship positively, but maintaining the quality of traffic leaving Simwood will continue to give us the pick of top-drawer routes at great prices, which benefits all our customers.

I hope that gives some background and clarity but please feel free to reach out, or discuss in our Community Slack.