January 21, 2019 | 02 min read
“Is the travelling salesman problem an NP-hard or an NP-easy problem?” our CEO asked one of the new R&D employees at Mobisy.
The non-techies, like me, maintained grave, intelligent expressions, pretending we knew what NP-hard meant.
But we actually probably looked like this.
The reason for his question was that being in the business of transforming the supply chain for consumer goods, the ‘travelling salesman’ is truly at the core of the $100Bn CPG Industry in India.
First posed in the 1800s, the ‘travelling salesman’ problem is still a favourite puzzle amongst developers and mathematicians. You can read about it here.
But why am I telling you about this? Because it’s FMCG’s biggest puzzle too which every brand, including yours, has been trying to solve at some level or other.
The Travelling Salesman in CPG
With thousands of outlets in a single city/town where one retail outlet caters to approximately 25 households, it’s always a balancing act to cover as many outlets as possible and service the top 10-20% of your key outlets which gives you up to 60% sales.
So how do you optimize your salesman’s time and effort to ensure that your products are visible and available in as many outlets as possible, ensuring the highest throughput possible.
I sat down with our in-house Data Scientist to get some gyaan on how this works. In between jargon like VRP and the knapsack problem, he managed to simplify it enough for laymen like me to understand and put it in the context of FMCG. So here’s my breakdown.
There are two parts to this problem: Beat Optimization & Delivery Optimization. To keep you from dozing off we’ll stick to just Beat Optimization for now.
Scenario: A city with 30,000 outlets
Quick Math Time
Daily outlets visited per sales rep: 45 (Industry average)
Total outlets covered in a week: 45*6=270
Assuming the brand has a weekly beat system
Total No. of sales reps: 30,000/270= 112. Let’s round it off to a nice-sounding number 125.
By using basic logic, one can prioritize outlets depending on sales and optimize this so that the crucial outlets are visited at least once a week. The rest are serviced fortnightly and monthly and so on.
But this has various limitations as the ASM assigning beats manually just cannot take into account multiple parameters and create the ideal beat plan which covers the most outlets in the shortest time while also achieving targets and not missing any key accounts.
(Spoiler: You can increase productivity by up to 40% by simply letting machines do what they’re good at.)
So here are some of the constraints that an algorithm can take into account to produce the most optimized beat plan.
While all this sounds cool, how do we align this to a brand’s desired outcomes and strategic goals?
Well, depending on what your goals are, the beat plans can change dynamically. So depending on whether outlet growth is a priority, sales targets is a priority, or you’re looking at a mixture of both, the algorithm can churn out a beat plan most suited to your needs and ensure outcomes are met at a route level.
For eg., If you want a minimum of 30 outlets per beat and a minimum order value of Rs 10,000, then you would have a different route plan as opposed to if you wanted to concentrate on just reaching the maximum outlets in a new geography.