Subex Recommends 75% of Analysts Can Be Made Redundant

Congratulations to Subex, the Indian revenue assurance giant, the first vendor to come clean and clarify that software is an alternative to paying people to do jobs, and not an excuse to create even more jobs. However, I am not sure they realized what they were saying…

This story begins by observing a change in Subex’s marketing approach, following the change in Subex’s management team. Subex’s old management team used to issue press releases announcing sales made to real customers, many of whom they could name. Now under new leadership, Subex seems to have copied cVidya’s marketing style, no longer telling us about sales, and instead issuing a press release saying one of their products is really very very good. Honest. Oh yes it is. It really is. It really is very good. And it must be, because the press release says so. Here is a snippet of the new puff release for an old product:

Zen is the industry’s first solution that sifts through millions of records and narrows-in on the problem area in a matter of minutes. It is a powerful magnet that pulls out the needle from the haystack.

Kudos to Ashwin Menon of Subex, who stood in the line of fire and answered a lot of tough questions about Zen from talkRA commenters. Many were sceptical about what Zen actually does. Ashwin’s post generated 20 comments in total. This was a record for talkRA, and double the total number of comments that Gadi Solotorevsky has received in the thousand years since he assumed leadership of the TMF’s RA team. (Note to people who like data analytics: some might think this is evidence that nobody actually reads Gadi’s blog. Or that they suffer a boredom-induced brain haemorrhage whenever they do.) In answering the sceptics, Ashwin appeased those who felt ‘a magnet that pulls needles from haystacks’ was an overly metaphorical way to describe software functionality. They demanded literal explanations of why they should use Zen; Ashwin had the integrity to provide them. However, eight months later, Subex continue to exhibit an excessive love of flowery language. However, they do give us some hard numbers about how well Zen pulls needles from haystacks. And here are those numbers, summarized in one place:

  • “increases productivity by over 75%”
  • “reduces the time to find root causes, typically from 30 hours to 30 minutes”
  • “ideally provides a substantial boost to analyst productivity by an order of magnitude of 10-20”
  • “Tier #1 African operator success story… improvement in time saving… 66.67%”

Yup. They sure offered a lot of numbers saying how good Zen is. Note to people who like data analytics: the numbers are all very different. Some of the improvements are four-fold. Some are sixty-fold. Some ten-fold, some twenty-fold, some three-fold. Which means a lot of them are wrong. Possibly all of them.

Let me spell out what Subex is saying in their headline factoid: analyst productivity will improve by over 75%. What they really mean is that, with the help of Zen, analysts will complete jobs in a quarter of the time that they used to take. (Note to people who like data analytics: this is actually a 300% improvement in productivity – 1 man can do the same work that used to be accomplished by 4 men.) Well, that sounds very good. It sounds unlikely, given Zen is just a big magnet and operators rarely suffer from chronic haystacks, but it does sound good. So, RA analysts, next time you sit down alongside your three colleagues who are also RA analysts, take a good look at their faces. Because you will not be seeing them much in future. Unless you meet them down the unemployment centre, looking for a job.

It is usually at this point in the argument that somebody blusters how you really need those other three analysts to do some other really important work. Maybe they can go somewhere and reduce a previously unknown risk by a few thousand orders of magnitude. Back in the world of real business, where data analytics is about using real data to make real decisions, either you have a job that needs four people to do it, or you do not. If it is cheaper to sack three guys and use software, then you sack three guys who you do not need any more. And if you just fancy taking the three guys and giving them another job, then you had better be giving them a real job which delivers real benefits, not just keeping them on the payroll for old times’ sake. In which case, you should have already employed three other guys to do that other job, and you did not have to wait for Subex, Zen, or any other vendor or product to come along and tell you it is time to rewrite the job descriptions for three quarters of your team.

The bottom line is that each necessary task is a genuine task, whether performed by man or machine. You would have to be pretty stupid to believe that the interests of people who sell machines are perfectly aligned to the interests of people who want jobs. Automation exists to lower the burden on people. And companies lower the burden on people not because they are charities, but because they want to employ fewer people in order to reduce costs and improve profits which they pass on to shareholders. Simple. So this amateurish press release from Subex is the first really strong sign that the cosy consensus between RA employees and RA vendors is breaking down. In a greenfield, buying some software can also create jobs, because you employ somebody to push the buttons. In a mature business, if you can get the same results with fewer buttons being pushed, you employ fewer people to push the buttons.

It is possible that the old consensus will temporarily break down, only for a new form of consensus to emerge between telco staff and vendors. Both the employees and the vendors may try to recreate the original greenfield expansion stimulus, by simultaneously extending their scopes to cover new areas. That way they can try to recreate the convenient win-win of the early days of revenue assurance, where software was sold and jobs were created at the same time. I expect we will see some of those messages from firms like Subex, just to appease the people they will have inadvertently upset with this press release. I am sure we will hear it from Gadi Solotorevsky of cVidya, because he wants to be everyone’s friend all of the time. After all, he has maintained his thousand-year peaceful reign precisely because he is a dead-eyed, cold-hearted, data-driven ruthless terminator of telco waste, whilst also being a simple, kind, jolly fellow who hands out alms to the meek and is beloved by small children and furry animals. (Note to lovers of data analytics: yes, there really is some kind of contradiction in there. As well as sarcasm.) I am sure we will get these familiar contradictions being repeated. However, if bad decisions are compounded for short-term gain, then employees will suffer sooner or later.

Pushing buttons is the wrong way to create jobs. Pushing buttons is a low-skill activity, and will always get squeezed by improvements in automation. Too many RA jobs were defined by RA vendors. They wanted employees to push the buttons on their software and fill in the gaps in its functionality (whilst not drawing attention to those gaps). As such, some jobs were not defined to address the real needs of telcos, and little thought was given to creating a career path for the people who filled those button-pushing jobs. The right approach, which RA could have taken from the outset, would have involved minimizing the number of low-skill button-pushing jobs. The emphasis should have been on the human skills that will never be automated. If that had been the case, then Zen would contain nothing new. (Note: I make no assumptions about whether Zen contains anything new.) The right approach would have involved giving the RA department, and hence its people, a job that went beyond crunching data to find anomalies. Those jobs would have been high-skill, and would have needed to deliver high value-add in order to justify them. They would require the kinds of skills we expect from real professionals – which includes seeing through marketing bluster and ignoring nonsense numbers.

It might be too late to reverse the mistakes made in this pool we know as ‘revenue assurance’, in which case many employees will have to sink, whilst automation sails on. Of course, the people with most to gain might offer some employees a berth on a lifeboat, perhaps named ‘revenue intelligence’ or one of those other phrases they so love to coin. But the destination will be much the same. Or we can learn from mistakes, and create jobs that make the most of human minds and restrict software to the role it really deserves: quickly executing a lot of simple, brainless, tedious calculations. That would be a radical change of approach. It would necessitate that telco employees assume the role of leaders, and their software suppliers would revert to being the followers. The tasks performed by software will have been defined for the developers, not defined by the developers. Now what do you think the current ‘leaders’ of revenue assurance might say to that? Perhaps they would disapprove. Everyone claims to like the forward march of progress, but nobody likes progress when it threatens to make them redundant.

Eric Priezkalns
Eric Priezkalns
Eric is a recognized expert on communications risk and assurance. He was Director of Risk Management for Qatar Telecom and has worked with Cable & Wireless, T‑Mobile, Sky, Worldcom and others.

Eric was lead author of Revenue Assurance: Expert Opinions for Communications Providers, published by CRC Press. He was a founding member of Qatar's National Committee for Internet Safety and the first leader of the TM Forum's Enterprise Risk Management team. Eric currently sits on the committee of the Risk & Assurance Group, and is an editorial advisor to Black Swan. He is a qualified chartered accountant, with degrees in information systems, and in mathematics and philosophy.

Commsrisk is edited by Eric. Look here for more about Eric's history as editor.
  • Dearest Eric,

    Over the last month or so, there’s been plenty of fire and brimstone in the U.S. presidential election. Thankfully, all the squabbling will end on Election Day, November 6th.

    But perhaps the bickering is contagious. I’m not a doctor, but if I didn’t know any better I’d guess, Eric, that you’ve come down with a rare infection, a kind of cross between Romnesia and Obamania.

    And it seems there’s a Stephen Spielberg in you, too. I successfully navigated the minefield you laid in first paragraph only to be singed by the tongues of flame that spurting out of paragraph two. I was knocked down by the hurricane winds you blew at revenue assurance vendors in paragraphs three and four.

    But I think its in the last two paragraphs of your column, where you talk about the idiocy of simply “pushing buttons”, where you really give us a takeaway.

    Does anybody remember SOA or Software Oriented Architectures? A few short years ago, people in IT were convinced that the computing world would soon be consumed by web services that would define all the processes of a business. I attended a conference session at the time where a couple Brazilian guys gave a presentation explaining how “in theory” they had devised an approach that would enable them to run their entire telco operation at the push of a button. They were perfectly serious. But isn’t that just like the telecom industry? It’s a place where mad scientists are treated with respect :- )

    Writing the code for my Black Swan website brought the “push button” fantasy into sharper focus for me. I succeeded in creating some templates to automate the merging of story text, images, author biography, related article links, etc into an HTML page.

    Fine, except for one small problem: I could never automate myself out of the picture. And trying to document the process or train someone to do a lot of these tasks is far tougher than you first imagine. Lesson learned: you still need an intelligent man (or woman) to do the dozens of little tasks that set up a “just push the button” automated process. In fact, the more complex your automation, the more knowledgeable a person you need to guide it.

    Another example that perhaps everyone can relate to: using Microsoft Word or Excel. If you’re played with that software long enough, the functions become almost second nature and it would take a awfully long time to train a person from scratch to achieve the same level of expertise.

    Could Shakespeare ever explain how he creates his masterpieces?

    I think what you’re saying, Eric, is that somewhere along the line, the close-to-the-process RA genius was marginalized so that somebody could make a software sale. And the challenge today is how to insert that genius back into the software. I’d be interested to hear any ideas others have to help make that happen.

    • @ Dan, thanks for the great comment. Though metaphors can be unhelpful, let me suggest another one to elucidate my point. What is a chauffeur? Well, he’s not exactly a man who just pushes buttons. But the details of what he does in his job – the key skills he must master – are mostly defined by the machine he operates. He doesn’t organize, manage or arrange travel and transport in general. He doesn’t have a choice about what tools to use to get somewhere, and if there is a place that cannot be reached by car, then he won’t be going there. He operates a specific machine, the motor car. I’d rather see RA people being given an open remit to get the company from A to Z, rather then being told to drive a specific vehicle. Getting from A to Z might very well involve driving a car, so RA people are more capable if they know how to do that as well. But it shouldn’t define their job. Cars are wonderful things, but some places can only be reached by more expensive and complicated technology, and there are other places where the route to get there involves putting our boots on and doing the legwork. I want to free people so that the tools of their trade are nothing more than their tools. They should be picked up and used when appropriate, not machines that define the work of their operators. There’s no easy route to that – it’s easier to write a manual on how to use a machine than it is to teach someone a range of skills and the way to decide which is most appropriate. And that’s why I get angry at manipulators like Alon Aginsky or Gadi Solotorevksy. I have nothing against businessmen making and selling machines, but I have a serious grudge against anyone who tries to manipulate a skillset so everyone is reduced to being an operator of machines.

  • Subex’s 2 step strategy appears to have merit.

    Step 1) sell a COTS solution that still requires MANY RA analysts, but a fewer than before. (see Cathy Veranos posts on blackswan great stuff-thanks)

    Step 2) Sell improved RAC/ZEN and 75% labour savings when CFO’s start drilling down into these RA empires @ $100k per person/yr cost (US/Canada, benefits in).

    My pre-lim research shows that cVidya RA solutions seem more labour optimized from the start. Of course if you realize that Subex was developed in a country with 1.2Billion workers and cVidya Moneymap comes from a country with 7.8million – which culture values labour optimization?
    (that’s a low blow, but since I grew up near Mysore, I earned the right to be a blogging bigot)

    A quick employee sorting on LinkedIN into Verizon vs Bell for RA, fraud Analysts etc. reveals some insight- easy to find numerous analysts at the Subex accounts! My 16,000,000 connections make this very easy.

    In fact I just asked this question on GRAPA Group LinkedIN – rightsizing the RA department. Erik’s book has a small section, but seems industry needs to quantify this before the CFO finds out.

    How many RA/fraud analysts do I need per 1MM triple play subs??

    As to the quality of the fewer RA zombie assembly line folks at the cVidya accounts- they seem pretty smart to me – when I can find them.

    I am very new in the RA Busdev area, although I have worked in several telecom sector B2B companies including Avaya, Primus, Rhythms (TELUS partner, defunct clec), TeraGo (WiMax fixed wireless). I have executed B2B leadgen programs with AT&, TELUS, etc. FYI the cost of 1 (ONE) solid B2B lead is about $250-500. Only 50% of sales reps make their MRR quotas. Customer acquisition in USA, Canada is a HUGE, annoying expense in all CSP’s.

    So RA’s gotta provide some lambs to the EBITA Gods.

    2 cents from Canada…

    (ps, great blog, a bit rough, nasty at times for B/OSS nerd types, but being 50% British myself, we all know better to be born a dog than a man in England)


  • Eric,

    I like the chauffeur analogy, and there’s a fantasy about that too. Check out the YouTube video about Google’s “driver-less car”. Works great in a “carefully programmed route” when there’s a person in the driver’s seat as a backup safety measure. I just wonder how fast Google will drop this experiment the first time the car runs over a pedestrian. Alas, the automation fanatics are a fixture of 21st century life.

    To get a better perspective on merging man and machine, I ran across a very interesting article about Knowledge Worker Productivity by the late, great management guru, Peter Drucker.

    Here’s a quote: “The most valuable assets of a 20th-century company were its production equipment. The most valuable asset of a 21st-century institution will be its knowledge workers and their productivity.” (1999)

  • Hi Stuart,

    Glad you liked the interview with Verizon’s Kathy Romano.

    Intrigued by your question: “How many RA/fraud analysts do I need per 1MM triple play subs??”

    But I’m not sure we’re ever going to get a satisfactory answer to that question. The answer will vary due to a few factors:

    1. The maturity/experience of the RA department.
    2. Region of the world (your “labour optimization” point).
    3. Scale of telecom operations. When you merge two large telcos, you should be able to massively consolidate billing, and RA savings would follow.

    Here’s another dynamic: at the smaller operators in North America, RA is often a part-time function where RA expertise is kept fresh by occasionally hiring a consultant who comes in and renews RA processes and the monitoring/analytic systems. See the nice article by Curtis Mills of ProCom Consulting on how this works.

    On customer acquisition costs, I guess the costs to acquire a mobile customer in US/Canada are even higher than the $250 to $500 costs you cited for B2B.

    AT&T put out a bid of $39 billion to acquire T-Mobile USA not long ago. The deal was squashed by the regulators, but the offer at least gives us a fairly solid metric on the value of customer acquisition. Assuming T-Mobile has 34 million subs, the price-per-subscriber AT&T was willing to pay was $1,147.

    Now looking at AT&T’s 2011 annual report and factoring in the costs of network, customer care, and on-going operations, I estimate that marketing and sales costs are about 25% of its revenue.

    By that calculation, AT&T’s investment would have paid for itself in 4 years. But there are other factors to consider as well. For instance, AT&T would enjoy some favorable network consolidation benefits since a great deal of T-Mobile USA’s wireless backhaul infrastructure is leased from AT&T anyway. In addition, by acquiring its only major GSM competitor in the USA, AT&T would probably be able to raise its prices by 10%. So AT&T was probably looking at a 3 year payback.

    So while the cost to acquire seems very high, apparently U.S. mobile subscribers don’t churn that much anymore.

  • Dan,

    Thanks for putting up with my neophyte observations.

    My apologies about Kathy’s name, its ROMANO. I learned a lot about RA from that interview.

    Agreed, imho the OSS, network, RA/fraud management C level policy, RA process maturity + wide range of offerings + people systems are too unique in each CSP to draw broad staffing, productivity conclusions. But, there is empire creep in any big telco.

    Thanks for link to Curtis. I have read all of his postings on Blackswan-still waiting for part 2 of “Magic Market Multiplier”. (in fact I read many other postings there).

    Re cost to acquire a new MRR client for telco using outbound leadgen/telesales. My $250-500 refers to the cost of just securing an A lead in B2B WAN services for SMB’s. ie with an MRR ARPU of approx. $500-2500.

    Telco’s still have to get a sales rep to follow up and close that lead = another $500- $1000, if they have to visit client or close it on line via call center.

    So, yes easy to get to 25% level, and more, which is why RA/fraud revenue recovery is so valuable and having some extra “minds” finding money is a “good thing”.

    Valuation is another story as your example explains; Here’s some info about NPV values investors paid from 2011.

    And At&T, Verizon have raised cel prices and gained market share it seems.



  • Pingback: A Kind of Magic - Analytics Needs data()