IT IS a familiar litany. Your call is important to us. An operator will be with you as soon as possible. Your call may be recorded for quality-control and training purposes. These phrases are familiar because the use of call centres has become so commonplace: in 2007 companies worldwide spent some $280 billion on outsourced call-centre services, according to NASSCOM, a call-centre trade group in Delhi. Costs are rising (they jumped 30% last year in India), customers are becoming more demanding, and competition is increasing. So the companies that operate call centres are adopting new software to monitor and improve performance.

One trick, which can shave seconds off each call, is to take into account “context data”—such as the caller's location, the local weather, the time of day and records of their recent transactions—when routing a call. Cisco, an American communications-equipment provider, recently upgraded its call-centre software to provide this feature, which is popular with financial institutions. The caller can then be put through to an agent who is experienced at handling a particular sort of transaction.

Aneesh Nair, head of the Academia, a call-centre training school in Bangalore, says context data may also include information gleaned from news reports. Storms, sporting events and transport strikes, for example, can affect callers' moods and purchasing dispositions, he says, and can be used to provide tips to agents. “Demographic mapping” features provide additional information about the products callers are most likely to buy based upon their location, time of call, and whether they are calling from a land-line or a mobile phone.

The next step, beyond analysing the context around a call, is to examine the call itself, using “word spotting” software to evaluate the performance of the agent and the reaction of the caller. There is strong demand for quality-control software, says Donna Fluss of DMG Consulting, a specialist consultancy based in West Orange, New Jersey, because call-centre managers are only able to listen to a few calls per agent each month. Word-spotting software allows managers to monitor agents much more closely.

As word-spotting systems spread, call-centre training schools have started coaching students in gaming the system so they look good under the additional scrutiny. JTS Institute, a cramped school in a bustling Bangalore neighbourhood, now shows students how to “prompt” callers with questions and statements that elicit the replies that word-spotting software, and call-centre managers, are looking for, such as “I understand” and “that's great”. In addition, says Basheer Shaik, the head trainer, a good accent is even more important now that employers can see tallies of how many times a particular agent is asked to “repeat that”. Students are encouraged to practise English even “with their parents who hardly understand English”, he says.

The market for word-spotting systems is “gaining a lot of momentum,” says Barak Eilam, general manager of NICE, a provider of word-spotting software based in Ra'anana, Israel. Its latest systems are designed to foil prompting by taking into account the context of high-scoring phrases. A caller who says “wonderful” shortly after the agent has said it, for example, may have been prompted.

As well as being used to evaluate the performance of agents, Mr Eilam says, word-spotting systems can also provide valuable insights about what callers think about a particular product, and how their vocabulary changes in response to advertising campaigns, recalls and publicity stunts. In one case word-spotting technology quickly alerted a refrigerator-maker to a design flaw after the words “tipped”, “top-heavy” and “fell” were uttered to call-centre agents in different countries just half a dozen times.

Much of India's call-centre industry, which employs roughly 300,000 agents, is located outside the ring road that encircles Bangalore, in a string of smart new business parks with tidy lawns and private security. Were it not for the stray dogs, a visitor could be forgiven for mistaking the area for Silicon Valley. Costs, of course, are much lower, but business leaders are aware of a growing threat from call centres in other countries such as the Philippines. Vasudevan Bharathwaj of 24/7 Customer, a call centre in Bangalore where 5,000 agents handle calls for 30 companies, explains that firms like his must innovate and develop intellectual property for sale to other firms.

Listening to the customer

Accordingly, programmers at 24/7 Customer have developed “neurolinguistic” software that does not just spot which words callers use—it tries to provide agents with insights into their psychology. Callers' words and cadences are analysed to create a profile that helps agents adjust their vocabulary and behaviour to improve their rapport. Agents receive on-screen tips on which phrases, sales pitches and conflict-resolution tricks are most likely to resonate. For example, callers who use “kinesthetic” terms such as “digging through the website” will be answered slowly with phrases suggesting body movement: “Please hold while I pull up more information.”

Similarly, Cisco's voice-analysis system monitors parameters including volume, cadence, tone, pitch and inflection, and then sorts callers into six personality types to help agents fine-tune call handling. “It's the bleeding edge,” says Laurent Philonenko, vice-president of Cisco's call-centre business in San Jose, California. Ms Fluss says “tremendous innovation” is under way in this area, and sales of caller-profiling systems will increase by 70% this year compared with 2007.

CUNA Mutual Group, a credit-union insurer in Madison, Wisconsin, installed such a system last year. Rick Roy of CUNA says it provides agents, who receive tips and script changes automatically during calls, with “another set of ears”. The same software can also be used for quality control, says Mr Eilam, by scoring agents depending on the way the caller's emotional state changes over the course of the call. Supervisors at many companies, he says, retain customers by phoning back those who exhibit “negative excitement” at the end of a call.

Some in the industry worry that voice analysis smacks of Big Brother, and prefer to keep quiet about it. Callers are usually told only that their call may be recorded for quality-control purposes. The JTS Institute does not mention the technology to its students. “Too much information”, says Mr Shaik, “is not good for them.”