General enquiries :
+44 (0)20 7602 6000

How BT uses data science to forecast call-centre demand

Monday 3 December 2018 AICustomer ExperienceData Insight & Analytics


John Tansley's picture
By John Tansley

With a presence in 180 countries, and 18 million customers in the UK alone, BT is one of the biggest and most complex telecommunications providers on the planet.

But telecoms can require greater ongoing customer support and advice than many other sectors. This poses a problem: how do you handle and route call centre traffic efficiently, whilst also reducing costs and eliminating manual processes?

To help, BT aspired to forecasting call centre traffic more accurately – enabling the business to better plan its marketing activities, overall operations, and manage its long-term budgets with greater confidence.

Here’s the story of that journey, and how BT used CACI Forecaster to change the way it optimises planning and anticipates demand.


The problem: demand prediction in the dark

The need for forecasting is driven by uncertainty. If there was no uncertainty, planning would be a simple exercise indeed. But by analysing historical data and making best use of expert knowledge, it is possible to usefully predict things like sales and demand – taking into account how they’ll be affected by the time of day or year, public events, or the weather.

However, when your processes are manual and you have limited resources available, this kind of accurate forecasting can seem very difficult to achieve.

When BT first approached CACI, the remit was simple: it needed to simplify its forecasting process, as its manual processes made predicting demand slow and labour-intensive.

The company needed a tool to simplify this process – not just to maintain and improve its performance with new, limited resources, but also to produce valuable insight, and potentially expand into new areas.


The solution: tackle complexity head-on

To deliver accurate forecasting, it is vital to understand the drivers of demand, including the effects of factors like marketing campaigns and caller information – such as where the customer is in the buying cycle and what they’re interested in. If this kind of information is not available, techniques must still be able to produce the most robust forecasts possible.

BT used CACI Forecaster to better understand and forecast its demand, cross-sell opportunity, and conversion rates. Forecaster’s ability to provide information on the likely range of outcomes was crucial in enabling BT to plan effectively for different possible scenarios, rather than simply the most likely scenario.

For example, if Forecaster predicted a range of between 10,000 and 10,500 calls, BT could then use that information to prepare for the range of possible business scenarios.

Using Forecaster, BT was able to understand its forecast performance and manage risk in real-time. It not only allowed quick visualisation of performance versus its forecast, but also helped to highlight any significant deviations to suggest where the model could benefit from additional information to improve the forecast’s accuracy.

BT’s overall forecasting now takes the same amount of time to produce 66 forecasts, as it did to come up with 11 using manual processes.


The outcome: optimised outcomes and unexpected insights

For BT, the benefits of using Forecaster were clear. The software enabled the company to achieve an improved level of service, while also offering on-the-fly adjustments and valuable insights into the drivers of customer demand.

It also enabled managers to spend less money by anticipating demand and enabling better allocation of resources.

In fact, BT’s overall forecasting now takes the same amount of time to produce 66 forecasts as it did to come up with 11 using the previous manual processes. And with new model insights to explain what drives each forecast, the process is both simpler and more transparent.

What’s more, BT discovered insights it didn’t expect. For example, it identified a segment of customers that it didn’t previously believe were interested in sport. As well as providing this new insight, Forecaster was able to use Champions’ League activity on BT Sport as a variable to improve forecast accuracy.


The future: a culture ingrained in test and learn

For an organisation with such a significant presence around the world, the work to optimise and improve never stops.

The plan for BT going forward is to create a more integrated and automated forecasting and planning process across all functions – extremely useful for their demand-led budget process.

As it ventures further down the data science avenue, BT also wants to tackle more difficult planning problems, and discover opportunities to drive further insights and achieve greater forecasting accuracy.

Couple this with some of the more advanced machine learning algorithms in time series forecasting, and it’s safe to say the company’s data planning effectiveness is set to grow from strength to strength.


CACI Forecaster: making the complex simple

Forecaster is a demand forecasting tool which provides advanced data-driven predictions.

BT found that the key to improving their data management was automation. By using Forecaster, that’s exactly what enabled it to gain greater insights, with fewer call centre resources.

Forecaster enables you to:

  • Forecast demand through a simple interface that can be shared with your team
  • Take advantage of machine learning and automated data ingestion
  • Remove the need for manual seasonality modelling

The potential applications go far beyond call centres; from logistics and supply chain systems, to web traffic and email marketing campaigns, countless organisations now use it to forecast a wide range of business outcomes.

For an organisation as large and complex as BT, running an efficient and cost-effective call centre is critical. To support this, the company’s data scientists use forecast modelling to predict the future, boost customer satisfaction, and save money.

How BT uses data science to forecast call-centre demand