Is it time to subscribe to a printer service from HP?

Ever since my dad brought home an...

What’s the best way of buying a phone today?

How did you buy your latest phone?...

MWC: What device highlights did you miss?

So, early last week I predicted that...

Can SpinVox spin their way out of this?

I’m sure you’ve heard of SpinVox – the UK-based firm that turns your voicemails into texts. And I’m pretty certain you’ve probably used them at some point – or know someone who has.

There’s been rumours flying around since SpinVox first surfaced that the clever technology that recognises speech and turns it into a pretty text wasn’t as smart as the company claimed it was. As their website says:

“So D2’s pretty smart. It’s bound to be, as D2’s a combination of artificial intelligence, voice recognition and natural linguistics. But it also knows what it doesn’t know and is able to call on human experts for assistance”

But how often are these ‘human experts’ called? And who are they?

Rory Cellan-Jones, the BBC’s Technology Correspondent, has done a little digging – and come up with some shocking facts (link).

Claims have been made to the BBC that the majority of messages are manually transcribed by low paid staff in battery farmed call centres in South Africa and the Philippines.

And former call centre worker Mohammed Mustafa told the BBC “The machine doesn’t understand anything. You have to start typing when you hear the message.”

SpinVox have declined to tell the BBC how many messages are manually transcribed, claiming that information is ‘highly confidential and sensitive data’.

So do SpinVox treat your voice messages as ‘highly confidential and sensitive?’ Apparently not. The BBC took a look through SpinVox’s registration information under the Data Protection Act, and found the company claims it doesn’t transfer any data outside the European Economic Area. Which is fine, because obviously South Africa and the Philippines had a meeting overnight when I wasn’t looking and suddenly became part of Europe.. didn’t they?

Meanwhile, there’s the whole question over money. Where precisely has all that investment – the BBC claims over £120m – gone? Apparently not on voice recognition technology. That leads on to another question – if its mostly gone on paying for transcriptions at call centres, then surely the business model just doesn’t add up? If the cost of an agent transcribing a message in a call centre is more expensive than the cost SpinVox charges to the customer, then what’s the point?

And whilst we’re on the subject of money, The Register reported last week that SpinVox had asked employees to take share options instead of wages. The BBC has a quote from ISP ANLX – who host some of SpinVox’s servers – saying that they’ve removed SpinVox’s access as they haven’t paid the bills. Plus there’s also rumours floating around that ex MIR contributor James Whatley – who heads up “Digital & Social Media” and is seen by many as the ‘voice’ of SpinVox on various social media sites – is no longer employed by the company on a full time basis.

Now to balance things up a bit. I quite like SpinVox – I’ve used it on and off for a number of years. It sort of does what it says on the tin – the accuracy is sometimes horrendous (Guinness came out as ‘Gurmish’ once) and I’ve had to dial in occasionally to actually listen to the message. Did it save me time listening to voicemails? Yes. Was I aware that they used humans? Yup. Did I mind? No, but I assumed – probably like everyone else – that it wasn’t one big battery farming dictation operation with apparently not a single bit of technology involved.

I’m going to cancel my subscription to SpinVox today – and I’m sure other people will be. Something just doesn’t feel right here. I feel like we – as the mobile industry – have been misled about just how wonderful and technologically advanced the service really was. And if I was an investor specialising in funding hi-tech ventures, right now I’d be on the phone to my lawyers clarifying what I’d really invested in.

27 COMMENTS

  1. Bit harsh Ewan, no? AFAIK, one can honestly declare that they do not move data outside the EEA provided the data is not hosted outside the EEA (i.e. the servers are in the EEA). A thin client outsite of the EEA would not be in breach of this declaration.

    As to the mix of automatic speech to text via tech vs via human interaction, it would be interesting to know that mix. They say only some go via call centres, I would be inclined to give them the benefit of the doubt for now.

    Finally, what difference does it really make from a consumer perspective. I wouldn't use such a service for delicate or highly confidential communication anyway. As to what the investors have got – I would hope they did their due dilligence (and would be very surprised if they had not).

  2. @Barry-Jon.. It was a piece by Alex (welcome back Alex)..

    Anyone who's worked in voice/ivr etc knows how hard speech recognition is.. don't they?

    Cinema – “Please state the name of your nearest town”
    Me – “RICHMOND”
    Cinema – “Did you say 'WOLVERHAMPTON/TIMBUKTU”
    Me – “TAXI”

    Don't think I was ever under any illusion they weren't using farmed transcribers (any different to Apple sweat shops in China?), of course the interesting part is how much of it is actually automated. I bet they got too busy too soon, and have not had time to do anything other than keep majority transcribing. Do wonder where all that funding went mind you.. wow. I believe they have some cool in country on carrier deals in places like Spain.. Maybe that's a bigger play than D2C for them.

    One question – where is the competition?! I'd say it'd be very hard to get many carriers to allow further connections of this type..

  3. Thanks njar, nice to be back. Know exactly what you mean about the cinema IVR.

    I guess we all had a hunch about SpinVox and the human element about their transcriptions – but probably gave them the benefit of the doubt and thought it was some sort of magic QA/learning thing. You know, like machine recognised version goes to operator along with original audio, operator listens and corrects words/fills in the blanks.

    Dig a bit deeper and there appears to be a third world labour/human rights story brewing here. Mike Butcher at TechCrunch has already uncovered the tip of the iceberg with the discovery of the Facebook groups. Run some of those through Google translate and it's not a pretty picture.

    As for competition, I'm giving Vox Sciences a whirl. OK so I've only had a couple of voicemails since this morning when I set it up, but so far it seems to work.

  4. I know I haven't got all the facts in front of me, and I haven't gone through the archives to see what James wrote. However..

    I'd imagine James did know exactly how it worked. But why would he let the cat out the bag? In his role for SpinVox he'd be not only under NDA, but I'd imagine have his balls chopped off if he was found to be writing negative things about his (ex-) employer.

    As for duping the MIR audience, there's a simple choice: Either clearly write a disclaimer stating your relationship with the company, or don't write the article at all. There's been a few times over the past month or so when someone has asked me if I could write about them for MIR, and I've had to pass it over to Ewan as I feel there's a conflict of interest due to actual or potential commercial relationships I have in place elsewhere.

  5. With Google already proving that they can index audio http://labs.google.com/gaudi, I see a short future for mass call centre transcription. 2010 = no more spinvox. Voice to text will probably be just a function of Google Waves and open source, so you will tend to use waves for your voicemail

  6. Probably the worst-kept secret in the mobile industry. Stephen Fry blogged about it last year and plenty of people put him right. It's always reminded me of the talking computer in “Shooting Fish”.

  7. It is a darn shame but it does explain why world domination did not happen – biz does not scale maybe? The problem is voicemail provided by the operators is crap and this was a viable alternative. Anyway I hunted around and found these guys HulloMail, they don't have transcription, although they claim on twitter to be adding it. Pretty cool so far especially with gmail accounts where they sync up with both email and contacts…

  8. Bit harsh Ewan, no? AFAIK, one can honestly declare that they do not move data outside the EEA provided the data is not hosted outside the EEA (i.e. the servers are in the EEA). A thin client outsite of the EEA would not be in breach of this declaration.

    As to the mix of automatic speech to text via tech vs via human interaction, it would be interesting to know that mix. They say only some go via call centres, I would be inclined to give them the benefit of the doubt for now.

    Finally, what difference does it really make from a consumer perspective. I wouldn't use such a service for delicate or highly confidential communication anyway. As to what the investors have got – I would hope they did their due dilligence (and would be very surprised if they had not).

  9. Spinvox was often mentioned and praised by James' contribution on MIR and Spinvox really did seem to be a global success story… I thought Spinvox would have been paying MIR handsomely for all the coverage and attention and MIR happily took the money at the time. How much did James' know? If he was aware then surely he was duping the MIR audience?

  10. @Barry-Jon.. It was a piece by Alex (welcome back Alex)..

    Anyone who's worked in voice/ivr etc knows how hard speech recognition is.. don't they?

    Cinema – “Please state the name of your nearest town”
    Me – “RICHMOND”
    Cinema – “Did you say 'WOLVERHAMPTON/TIMBUKTU”
    Me – “TAXI”

    Don't think I was ever under any illusion they weren't using farmed transcribers (any different to Apple sweat shops in China?), of course the interesting part is how much of it is actually automated. I bet they got too busy too soon, and have not had time to do anything other than keep majority transcribing. Do wonder where all that funding went mind you.. wow. I believe they have some cool in country on carrier deals in places like Spain.. Maybe that's a bigger play than D2C for them.

    One question – where is the competition?! I'd say it'd be very hard to get many carriers to allow further connections of this type..

  11. Thanks njar, nice to be back. Know exactly what you mean about the cinema IVR.

    I guess we all had a hunch about SpinVox and the human element about their transcriptions – but probably gave them the benefit of the doubt and thought it was some sort of magic QA/learning thing. You know, like machine recognised version goes to operator along with original audio, operator listens and corrects words/fills in the blanks.

    Dig a bit deeper and there appears to be a third world labour/human rights story brewing here. Mike Butcher at TechCrunch has already uncovered the tip of the iceberg with the discovery of the Facebook groups. Run some of those through Google translate and it's not a pretty picture.

    As for competition, I'm giving Vox Sciences a whirl. OK so I've only had a couple of voicemails since this morning when I set it up, but so far it seems to work.

  12. I know I haven't got all the facts in front of me, and I haven't gone through the archives to see what James wrote. However..

    I'd imagine James did know exactly how it worked. But why would he let the cat out the bag? In his role for SpinVox he'd be not only under NDA, but I'd imagine have his balls chopped off if he was found to be writing negative things about his (ex-) employer.

    As for duping the MIR audience, there's a simple choice: Either clearly write a disclaimer stating your relationship with the company, or don't write the article at all. There's been a few times over the past month or so when someone has asked me if I could write about them for MIR, and I've had to pass it over to Ewan as I feel there's a conflict of interest due to actual or potential commercial relationships I have in place elsewhere.

  13. With Google already proving that they can index audio http://labs.google.com/gaudi, I see a short future for mass call centre transcription. 2010 = no more spinvox. Voice to text will probably be just a function of Google Waves and open source, so you will tend to use waves for your voicemail

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Recently Published

Is it time to subscribe to a printer service from HP?

Ever since my dad brought home an HP LaserJet printer (version 3, if memory serves), I have been printing with an HP. Over the...

What’s the best way of buying a phone today?

How did you buy your latest phone? I'm asking because I'm thinking about what I should be doing. When I was living in Oman, I...

MWC: What device highlights did you miss?

So, early last week I predicted that next to nothing from Mobile World Congress would break through into the mainstream media. I was right,...

How Wireless Will Pave the Path to Neobank Profitability

I'm delighted to bring you an opinion piece from Rafa Plantier at Gigs.com. I think it's particularly relevant given the recent eSIM news from...

An end of an era: Vodafone UK turns off 3G services

I thought it was worthwhile highlighting this one from the Vodafone UK team. For so long - for what feels like years, seeing the...

Mobile World Congress: Did the mainstream media notice?

I resolved this year to make sure I wrote something - anything - about Mobile World Congress, the huge mobile industry trade show taking...