Serving more than 55 million subscribers in every corner of Bangladesh is not a child’s play. Traditionally, telcos have depended on their distribution network to ensure that their customers have easy access to the retailers to top up their mobile balance.

An army of Direct Sales Representatives (DSRs) from the distributors would normally hit the road on foot to go from retailer to retailer to make sure they are adequately stocked up to serve the customers. There are around hundred thousand retail accounts who keeps the telco business afloat every single day of the year.

Based entirely on manual system, this legacy system has many shortcomings. Due to the manual route planning, the DSR allocation against the routes remain unoptimized. Naturally, the DSRs fail to serve all the retailers in an area/ geographic location. Besides, critical retailers, in terms of their level of impact on the business, are not considered while preparing the route plan for the DSRs. Also, the DSRs don’t have any visibility on the forecasted demand of each retailer.

In short, the three salient pain points of the legacy system are that, it is unable to forecast stock-out situations for retailers, retailer prioritization mechanism is completely absent, and route planning for the DSR is at best sub-optimal.

Through an internal study of Robi’s distribution network, it was found that 15% of retailers are prone to run out of balance on any given day. Within them, around 20% retailers are not in the plan to be served by the DSRs. The study also revealed that 25% of customers are served by those 15% retailers; but more importantly, on average, 35% of revenue is generated by this user base. Therefore, the apprehensions on the legacy system is fully substantiated.

Robi, being keen on digital innovation, resorted to AI to help it plug the gaps in its distribution network. Robi’s Analytics Centre of Excellence team delved deep into the historical data of DSR, and retailers. Following necessary cleansing of the data-sets, it was run through AI models to forecast sales, estimate geo-impact of retailers, and retailer clusters. Retailer prioritization, route plan, and stock-push alert for each retailer are then made available to DSRs to ensure smooth supply minimizing stock-out scenarios for the retailers.

Data (like retail location, DSR visit history, recharge logs, customer usage, etc.) are sourced from various sales and billing systems. Oracle Exadata & Cloudera Hadoop platforms are used for storing data, and Apache Spark platform is used for data processing. While creating the AI models, supervised, semi-supervised learning techniques are deployed using algorithms like- Neural Network based LSTM, time-series forecasting, Geo-spatial clustering, constrained clustering, etc. The outcomes of the AI model is then delivered using data visualization tools, like- Tableau, SAP Business Objects (BO) and Sales app.

In summary, Robi’s innovative solution- Distribution BOT forecasts stock demand, prioritizes retailers, and provides optimized route for DSRs. Using the AI based stock recommendation feature, the solution predicts future stock demand for every retailer and raises early alert accordingly, forecasts sales for different locations/ points, and most importantly recommends data driven amount of stock provision for each retailer.

The geo clustering feature of Distribution BOT prioritizes different points based on geo-location and business impact, and provides distributors with prioritized list. Through the route optimization feature, the solution cluster points based on geo-location and business priority and suggests optimized route for the DSRs.

It is estimated that the Distribution BOT can potentially bring in 1.3 billion taka in additional revenue for Robi. Of which, Stock Push Recommendation feature alone can potentially contribute 950 million taka in additional revenue. The Retailer Prioritization feature can potentially contribute 125 million taka, annually.

Following implementation of the Distribution BOT, we have observed 25% degrowth in low balance scenarios and 26% reduction in easy load failure scenarios due to insufficient balance for the retailers. Though yet to be tried in real life scenario, the Route Optimization feature has the potential to contribute 215 million taka in additional revenue, annually, and facilitate 15% reduction in the number of DSRs engaged by the distributors.

Though it was designed to support Robi’s business, the Distribution BOT can be used by any telco in markets with similar features. Beyond telcos, multinational FMCGs who have distribution chain like telcos can greatly benefit from this innovative solution.

"Robi Distribution Bot" has been recognized as the champion in the 2022 "Leap Now”- an initiative under the LEAP Programme run by Axiata Group Berhad to identify the best digitisation projects among all its operating companies. Earlier, Robi was named champion in AI Maturity for the fourth year in a row in 2022 among all Axiata entities, including digital and ICT business units. These accolades clearly highlight Robi’s position as a leading global mobile operator in enabling AI for digital transformation and creating value for our customers and shareholders.

“Artificial Intelligence will revolutionize supply chain in ways that haven’t even been thought of yet”- said the legendary AI expert, Dave Waters. He also predicted that Artificial intelligence will be rampant in the digital supply chain. On a more sublime note, he said- “Predicting the future isn’t magic, it’s artificial intelligence”. Robi’s Distribution BOT in many ways attests all his wise predictions, which in turn attests Robi’s credence as an innovative digital company in the market.