Voice 2.0 Startup Watch: Fonolo
Fonolo, a Canadian company that essentially provides a search engine for IVR menus, has recently showcased an interesting new market application. We think it’s a good example of innovation in ‘Voice 2.0’. Skype Journal reports that visitors to Sirius Satellite Radio’s website can now search for the person they need to speak to, and then click-to-dial them directly without going through the IVR.
Not just that, they can click-to-schedule the call at some time in the future. And they don’t need to wait on hold, either. Everyone who has ever struggled to navigate the IVRs, kicked their heels in the call queue, taken down the same details from the same customer three times in a row, or experienced the amazing inefficiency of call centres in general should be able to appreciate this.
Person-to-organisation voice is a product that should deeply embarrass our industry. Information doesn’t follow calls, far too much of it has to be read out and retyped, callers are expected to wait on the line at their own expense, metrics are poor, and it’s normal for one side of the conversation to have to trust the other’s record of it sight unseen. Very large percentages of calls are abandoned. Call centres are notoriously horrible working environments. As much as half of some call centres’ traffic is made up of so-called failure demand, repeated calls that didn’t work the first time. This is equivalent to rework in manufacturing - notoriously the biggest killer of productivity.
Business voice service is stuck in the mindset of General Motors in the 1970s - obsessed with throughput come what may at the expense of quality, sapped by constant rework, unable to turn around new products, trapped by its over-specialised tools, and apparently unable to think of any better ideas than trying to keep the workers from unionising and running the line ever faster.
And, amazingly, most companies rely on this service for sales.
Now, the equivalent to the quick-reconfigurable general-purpose machine tools that the European and Japanese car makers used to beat GM is already here - it’s the Toolkit of Voice 2.0. But where are the people who will make use of them? Who will build the Toyota of service?
We first reported on Fonolo in this post from April, 2008. Back then, their product was a pure-play search engine like the ones we’re all familiar with on the Web. In fact, it worked exactly like that - they had a bank of Asterisk servers automatically dialing into corporate call centres and trying each option on each IVR menu in turn, mapping all the possible routes through the wilderness of menus in the same way that the Googlebot follows links through the Web.
All this information was made available on a Web site, so you could search through it, and they used a click-to-call package, so you could dial direct to the person you needed from the search result. They also wanted to let the users log what happened during the call, providing a handy record of what you said to them last time and also a source of crowdsourced reviews and comments on the service. Their CEO, Shai Berger, appeared in virtual form at that autumn’s Telco 2.0 event.
At the time, they were expecting to monetise the service through advertising. We suggested, however, that a more direct form of two-sided business model might be appropriate. Essentially, Fonolo was offering to do a lot of companies a favour by getting rid of some of the major sources of failure in their call centres - misrouted and misdialed calls, calls that hang up in the queue, stress and aggravation for customers and employees - and significantly improving their lead generation, sales, and operational KPIs like first-call resolution. So why shouldn’t those companies simply pay for it?
Further, they had a problem in that the indexing process might be perceived as an attack, and their range of NPA-NXX phone numbers might be blocked. Also, there was a risk that their search engine would send a firehose of calls direct to a company desk that didn’t have the resources to handle them. Changing the relationship with the upstream customers - the call centres - from an adversarial relationship to a customer relationship would resolve that. It would also let the upstreams cooperate with Fonolo pro-actively, by providing it with information about their phone system, rather than just waiting for the robot spiders to attack.
Of course, this idea isn’t new - on the Web, it’s almost as old as search engines themselves. A parallel would be Google’s Webmaster Tools product, which basically provides a structured way of pushing information about your website into Google pro-actively, rather than waiting for the spiders to show up and hoping that Google’s automated judgement gets you right. Another, very basic one is robots.txt, the text file most Web sites use to tell search engines not to index certain pages.
Another example would be Internet routing - networks announce the prefixes they route into the global routing table, and update them when this changes or a link goes up or down. One advantage of this is that it permits traffic-engineering - you can influence how your traffic is routed, for example to use cheaper links preferentially, to balance out the load equally between multiple sites, to maintain a production network and a hot-standby network, or to ensure that requests are served by the closest possible machine.
By letting the upstreams actively contribute information about how they would like to be reached, Fonolo gets rid of the problem that the search engine might drop thousands of calls onto a hitherto obscure and under-resourced desk. It also permits its upstream customers to engineer their call flow to suit their own needs without forcing their customers through a tortuous IVR menu tree.
So, Fonolo is now out to sign up major call centres as upstream customers. In return, they get to integrate Fonolo’s service into their own Web site, as well as pushing their site-map into the system. What else?
Well, there’s something we didn’t make much of back then - Fonolo’s click-to-call element has a callback architecture like Jajah’s. You click, the Asterisk starts dialling, and when the call is connected, it rings you up and bridges the calls together. This is necessary to deal with the problem that not everyone uses desktop VoIP and most telcos still don’t have an API for voice. But it has some useful features in itself.
For example, you can eliminate waiting on hold. There’s no reason why, of course, the Asterisk server has to place the call at once - you could send it an AMI event specifying that the call was to be dialled at some point in the future. So you could click-to-schedule a call at some time when you’re going to be free. It’s better than that, though. Because it’s a call-back, there is no need to wait on hold - it’s as if you sent a robot to wait in the queue and summon you when it reached the front.
Of course, good programmers have known for decades that high performance requires you to replace polling patterns (repeatedly checking to see if something has happened) with push-notification and other event-driven ones (being informed proactively when it does), and to replace blocking calls (where the process that calls some function has to wait for it to return before continuing with the program) with callbacks (where the function reports back to the first process, which can get on with some work in the meantime). This goes double or even triple for telecoms. If our voice switches were reliant on constantly checking back on their input queues or waiting on the line for an answer before doing anything else, we’d never have scaled beyond the systems of the 1920s.
But it’s only now that these insights are being applied to customers’ needs. And it’s still a startup, not a telco, that’s doing it. There’s much more to be done here; why should Canadians have all the fun, for a start? The crowdsourcing/logging element of the original Fonolo plan really needs work - it’s part of the whole vision of VRM. Also, we’re still not so good at sending along more information with the call, and therefore eliminating the whole tiresome business of reading stuff off one screen so someone else can type it back into theirs. Metaswitch’s new Thrutu product is a crack at that. And there’s surely a lot to do in changing the call centres themselves - both their software and their organisation - to take full advantage. (Hint: Toyota put a big red STOP button at each workstation and encouraged people to press them and stop the line if they noticed anything dangerous or sub-standard. It stayed stopped until the problem was fixed. What would be the equivalent?)
So there are plenty of opportunities out there. But have the operators simply lost interest?