Hiya and Transaction Network Services (TNS), two of the big three analytics firms that dominate the labeling of calls in the USA, have exploited a blatant conflict of interest for financial gain, per a new regulatory filing from Numeracle, specialists in telecoms identities and spam. The opening page of Numeracle’s punchy submission to the Federal Communications Commission (FCC) illustrates the point by highlighting how Hiya generates revenue by telling comms providers how to label calls and by helping businesses to get the correct labels applied to their outbound calls. Referring to Hiya and their rivals as analytics engines (AE), Numeracle stated:
Hiya sells a branded calling solution to the inaccurate call labeling problem that Hiya and the other AEs have created. The conflict of interest inherent in having the same companies labeling telephone calls as spam and selling a solution to fix those labels could not be clearer.
Numeracle’s filing comes in response to yet another round of FCC consultations about how to deal with unwanted robocalls. For all the FCC’s tough talk, they have been wary of telling comms providers to block calls solely because analytics predict that the call will be spam. They know that any analytics applied at scale will likely result in many false positives, and hence generate a storm of complaints when legitimate calls are blocked. So the FCC has encouraged the more legally defensible position of analytics being used to advise customers that a call may be spam. This advice is typically displayed as a warning on the recipient’s handset. The problem with passing the buck to consumers is that an ordinary person has no more insight than the analytics firms already have, so they will trust the labeling process and reject any calls marked as spam. Warning labels have a similar impact to blocking for the originators of legitimate calls, though neither the FCC nor the call analytics firms take responsibility for the harm caused by false positives.
It is now established that inaccurate spam labels are being displayed on the handsets of Americans receiving calls, according to Numeracle.
The current approach by the AEs that carriers have inserted into the call flow to block and label purportedly illegal and “unwanted” calls has failed. Not only do the AEs fail to block and label huge numbers of illegal calls, they also apply inaccurate labels to legal and wanted calls, causing vast disruption to calls such as prescription pickup reminders, appointment confirmations, customer service call-backs, crime victim notifications, two-factor authentications, and financial transaction calls.
They accuse the analytics firms of profiting from their own mistakes.
Worse yet, the AEs are monetizing their failures and insisting that legal callers pay to reduce their risk of inaccurate labeling. This occurs in two ways: charges for telephone number reputation monitoring by some AEs and “branded calling” charges by all the major AEs to increase contact rates and reduce what they themselves concede is inaccurate labeling.
Numeracle cites a real-life example that appears very close to extortion.
For example, Numeracle previously contracted with a major AE to purchase reputation score monitoring for our customers. After assurances from the AE that not purchasing monitoring would not affect registration and remediation or have a substantive effect on whether a number was labeled as spam, Numeracle chose to do its own monitoring with test calls to actual devices rather than purchase score reports from the AE. The decision not to purchase monitoring services resulted in dramatic increases in “Potential Spam” tagging by the AE and a substantial reduction in success for remediating numbers that were inaccurately labeled as spam. Numeracle chose to resume purchasing monitoring services as doing so benefited our customers with reduced spam labeling. The decision whether or not to purchase monitoring should not affect whether an AE labels a number as “Spam Likely” or “Potential Spam” as the caller’s identity and calling patterns have not changed, but in our experience it does have such an effect.
Another example presents the exploitation even more starkly.
One AE imposes outrageous charges on callers that go to it directly for monitoring and remediation and do not use a company such as Numeracle as an intermediary. In at least one case, this AE charged more than $40,000 per year to register, monitor, and remediate spam labels for a caller with 100 phone numbers… this is to remove the spam labels that the AE itself imposes. “Pay me $40,000 and I’ll remove the spam labels I’ve created” is the essence of this contract reduced to its core.
The US voice ecosystem has three dominant providers of call analytics for labeling: First Orion, Hiya and TNS. A growing number of press releases show how much US firms want to increase sales of call analytics and labeling services in other countries too. Regulators and comms providers around the world should be wary of an evolving situation where these businesses initially gain a foothold in a market by attempting to provide a solution, but actually create a new headache.
Numeracle concedes that the AEs have a difficult — perhaps impossible — task. They began operating in good faith seven or more years ago as the only then-available tool against illegal robocalls… they have limited resources to make accurate determinations, which harms legal callers just trying to get their calls through without interference. Their incentive structure defaults to overblocking and overlabeling as there are currently no downsides to being overzealous in that regard because they are accountable only to their carrier partners. But they have not kept up with evolving technology, standards, and calling patterns. The AEs have thrown up their hands and conceded that inaccurate labeling is inevitable and have monetized the problem.
Numeracle’s submission slams Hiya on its first page, but they are similarly critical of TNS.
Hiya’s competitor TNS has two ways to charge businesses to overcome its inaccurate labeling: branded calling and reputation monitoring. TNS states on its website that “Enterprise Branded Calling controls spam labeling and helps increase answer rates for legitimate callers that follow best practices, enhancing overall enterprise calling reputations and boosting your business.” TNS also concedes that mis-labeling occurs when selling its reputation monitoring services. “TNS can help enterprises avoid mis-labeling, prevent damage to their brands, and improve call answer rates.” TNS sells a solution to its own mis-labeling: TNS Telephone Reputation Monitoring, in which it promises legitimate callers that by paying for a solution, TNS can “[e]nsure your legitimate calls get through and aren’t inadvertently being blocked or tagged as spam.” Who is “inadvertently” blocking and tagging? TNS — the same company selling the solution.
However, Numeracle is kinder in their assessment of First Orion than their two rivals.
Numeracle compared the spam ratings for its customers provided by the three major AEs for the second quarter of 2023. First Orion only labeled a handful of numbers as spam
because it does have an effective registration process for legal callers…
It is important to note that one AE does work cooperatively with legal callers, and its efforts should be recognized. First Orion has created a rebuttable presumption for vetted callers that its calls are not spam… Numeracle and First Orion have worked cooperatively on registration, monitoring, and remediation for almost five years, which results in better information for First Orion’s carrier partners’ customers about who is calling them.
Some limited data gathered by Numeracle indicates Hiya and TNS are much more likely to flag a call as spam than First Orion, but that does not mean Hiya and TNS agree with each other.
Numeracle worked with a number provider to random selected (sic) 100 numbers out of a newly allocated block of 1,000 numbers in the 930 area code. None of the numbers have ever been assigned to a customer. In the test, Hiya rated 32 percent of the numbers as spam and TNS rated 24 percent as potential spam. First Orion did not rate any as spam. Interestingly, there was only an overlap of 10 numbers of the 32 rated as spam by Hiya and 24 rated as spam by TNS.
It is not uncommon for the big US call analytics firms to promise a false positive rate in the order of 0.1 percent, though Numeracle explain why the real rate may be higher. Even if the error rates reported by analytics firms are accurate, they should still not be considered small. One in every thousand businesses is at risk of severe damage because of the incorrect labeling of their calls. I would not want to be the CEO of the telco that gets sued by failed businesses seeking compensation because the telco engaged the services of an analytics business which under-reported the number of mistakes they make. Everyone should also be wary of ‘reputation monitoring’ becoming a pseudonym for a new tax on making calls.
The harm that can be done by mislabeling good calls as spam was highlighted earlier this year when Unified Office CEO Ray Pasquale complained to the FCC about how difficult it is to correct mistakes in the labeling of calls made by restaurants and car dealerships he serves. With other countries following the lead of the USA by making increasing use of call analytics, the global telecoms industry needs to find satisfactory ways of rapidly and inexpensively reversing errors in spam detection and labeling, or else there could be a second explosive wave of complaints from phone users who have lost confidence in their ability to both make and receive calls.
The full text of Numeracle’s submission can be found at the FCC website here. It contains many insights and is well worth reading, especially for anyone who had been hoping analytics would solve the problem of nuisance calls.
