Skip to main content

Case Study

Enabling Multivariate Forecasting and Intelligent Rostering at CBSL

As the inbound call volumes turned into a constant barrage with only 1 dependent factor for planning or tracking them, call center associates faced with a major impact on achieving business KPIs. For one of their biggest clients, they needed more hands on deck to not only answer calls in real-time but also maintain five key LOBs (lines of business), viz. ELT, mass desk, dealer support, buddy desk and win back.

Conneqt Business Solutions Limited (CBSL) is India’s second-largest business process management (BPM) and customer lifecycle management (CLM) player, with more than 22 delivery centers across the country. The Business Process Services team helps establish superior customer experiences with simplified process management and sustainable business value for leading companies including India’s largest DTH operators. Standardization, centralization and automation lie at the core of the company’s solutions targeted toward customer lifecycles across industries.

Conneqt’s WFM team is responsible for accurately forecasting volumes, planning periodic capacities and scheduling rosters for pan-India locations, variable time zones and multiple skill sets. Staying ahead of the curve with so many moving parts required a unified and comprehensive solution offering data-driven insights with much lesser resource consumption.

The Challenges

DTH being a dynamic space, there were challenges of data pouring in from diverse sources and calls from multiple geographies. Despite a large team of agents managing customer services, Conneqt needed a factual method to distribute the call volumes and forecast the staffing needs to handle any unpredictable workload, especially the upsurge during major events or promotional campaigns.

Manual forecasting techniques conducted on day-level data, steered by the headcount in respective locations or LOBs were major deterrents for ensuring accuracy or credibility in the projected numbers. The client provided the forecasted numbers every day, and they divided this data manually at intraday level. Due to 12 locations, 13 languages and five desks catering to five operating times, scheduling and rostering was challenging and led to over/understaffing.

Shift planning with Excel was tedious and headcount evaluation across multiple locations could take several weeks to plan weekly capacity. Multiskilling and multiple locations were major constraints when in-office and remote agents had to be allocated based on hourly intervals. Conneqt also needed granular access to data for the desk, language and interval levels.

The Ask

  • A single source of truth enabled by a centralized data mart.
  • Accuracy in forecasting and scheduling with ML & Business rules-enabled modules.
  • Location-wise scheduling to support different shift patterns for each location.
  • Reduced manual intervention to plan rosters matching the right agents for each desk, location and language.

The ultimate goal was to ensure correct mapping between total call volume and the number of required agents to achieve the service levels (SLs) & 98% answering level (AL) with only 2% scope for errors. All in all, this was a multivariate use case where multiple datasets impact overall business planning, forecasting and scheduling.

Consolidating data in a single repository, the company intended to simplify its operations, assigning the right number of agents for every desk, no matter where they were located. They also needed to generate customized dashboards and detailed reports for accurate scheduling and rostering.

Based on their previously successful engagement, Conneqt approached FLOW with this new use case, seeking a comprehensive solution to tick all the right checkboxes for their customers.

Conneqt broke data silos by connecting all the data using FLOW

FLOW created a data mart to consolidate data across multiple centers with just a few clicks and built a single source of truth. Users could bring all the data into a single repository to analyze it anytime from anywhere, improving workforce management lifecycles for various use cases.

The WFM team could push their raw data via FLOW’s upload utility or populate the data with a customized template at the end source, including any ad hoc updates in the shift timings.

With customizable templates, users can push files from any data source or data structure as needed.

FLOW enhanced the accuracy of forecasting for streamlined capacity planning

The power of AutoML combined with FLOW’s inbuilt, advanced algorithms helped train modules over months of data to forecast call volumes at language and desk levels. With 5 desks handling 13 languages each, the process involved around 70 queues or skill levels for multivariate forecasting.

Adding ML-forecasted outputs for both groups helped determine the overall numbers for at least a week ahead. Automated selection of the best-fit model based on the training start-end date, prediction start-end date, etc., helped generate a daily-to-hourly ratio for forecasted calls.

The ability to import any algorithm using an advanced low code/no code interface further simplified the process for the WFM team. They could choose parameters and methods according to their specific KPIs or even dig into client forecast data by using FLOW’s upload utility for integrating reports.

Multivariate use case with factors like:

  • 12 locations, 5 desks, 13 languages, 70 queues and 30-minute interval data
  • Seasonality,
  • Mandatory holidays,
  • Business campaigns, events
  • External regressors

Since the agents work a 9-hour shift, the team needed overlapping headcount at intervals to achieve the SLA of 90/20 with 98% Answer level. FLOW’s what-if modules allowed the users to create different scenarios using impacting parameters to estimate the accurate staffing needs at a call center for a specific shift, interval, month, week and day. Once all the parameters were selected and call volumes predicted, users could lock the forecast for a given month.

The users created forecasting reports using 30-minute interval data while also factoring in the effects of seasonality, mandatory holidays and business campaigns on call volumes for optimum efficiency in capacity planning.

Scenario management through what-if conditions helped plan staffing requirements across locations and skills more accurately.

FLOW enabled automated workforce scheduling for optimum resource allocation

Conneqt was looking to schedule location-wise shifts based on workload predictions, scheduling rules and pre-set parameters. Earlier, the scheduling was done manually in Excel spreadsheets, increasing the chances of errors leading to over- or under-staffing. FLOW’s ML algorithms and business rules offered more seamless and efficient shift planning that included full-time, part-time and split-time shifts. The teams could allocate the correct headcount to each shift across locations and eliminate redundancies with automated scheduling that matched arrival patterns with staffing needs at each center.

Machine learning and business rules enabled them to achieve best scheduling efficiency and achieve better ROI.

Intelligent rostering to assign the right talent mix at every desk

Using the right mix of multiple skills at different desks and locations, FLOW enabled Conneqt to build intelligent rosters every week, eliminating tedious manual processes. Factors like business rules, separate work timings for women, desk-wise language contributions, weekly leaves, agent tenures and AHTs helped integrate human and social intelligence into the system.

Intelligent, automated rostering with:

  • Self-service, WFH/WFO/hybrid scenario
  • Complexities across different operational timing, 12 different locations, 70 skillsets
  • Automated, on-the-go over- and understaffing alerts

Data from diverse sources was collated to generate rosters and reduce headcount variance. The hybrid nature of their workforce and different operational hours for each location posed some challenges on the way. FLOW mitigated them with automated rostering adjustment where users could upload changed timings to the roster and the entire plan would be adjusted accordingly without any human intervention.

In addition, FLOW offered two dashboards—analytics and real-time queue management (RTQM)—to help compare forecasted and actual call volumes for a specified time range. RTQM dashboard allowed near-real-time access to key SLA metrics like total calls offered, number of calls accepted, abandoned calls, short calls, hold time, occupancy, etc.

Comparing the actual and forecasted numbers helped Conneqt gain valuable insights into overall business performance.

Ushering in a new era of WFM with FLOW

Removing redundant and time-consuming manual processes off their plates helped Conneqt’s WFM team focus more on their key responsibilities. 360-degree visibility into their data backed by AutoML capabilities, ensured improved forecasting and scheduling for a complex staffing structure. With FLOW, they can optimize restaffing costs and enhance employee satisfaction, all while creating a data-driven culture and elevating customer experiences.

To learn more about FLOW, connect with us at

FLOW Capabilities

  • AutoML-based multivariate forecasting for accurate scenario planning
  • AI-powered automation for intelligent rostering
  • ETL to consolidate data across multiple call centers
  • ML algorithms and business rules for seamless and efficient scheduling
  • Analytical dashboards for real-time and historical comparisons
QR Code

Scan the QR Code to download the Case Study.