Interoperable data mining

If you only have a single source of data, you will not completely understand why sales are peaking or dropping. In cases where you only have one dimensional sales data, you hence merely have a transactional understanding of the velocity. At the end of the day, that is of course what matters, but to to get there, you leave potentials hanging for grab by not going the extra mile. The analytical future belongs to those who have an interoperable perspective for getting a deeper understanding. The case below is a simple way of increasing knowledge and taking better decisions, with the result of increased sales.

WHY ARE SALES DROPPING?
The sales for client was dropping, and it was their their #1 menu / meal that was not keeping up. The client had the scan data, but did not understand why. The meal was tracked as one unit, and consisted of rice, meat, vegetables and a beverage.

The single first move the client did was to reduce the price to avoid the drop in sales. It was a very expensive move, without giving a proper lift in sales. The whole exercise turned into a loss. The client was still wondering why the sales was dropping?

We asked for the scan data, with the intention of putting a twist to it. We linked the sales to various open data sources; like location, user groups and eight different weather conditions. We started to look for patterns in the sales data, and to our surprise, their best selling item was dropping when foot traffic was increasing. The incidence hence dropped like a bomb. Why?

MULTI-LEVEL DATA MINING
We started by looking at the sales data. We further linked the data to number of customers in their restaurant. They were perfectly negative correlated. Long story short: We then linked the data to eight different weather conditions, and the result was clear. When the weather was turning warmer and more humid, the sales was dropping. It was good to know, but it was still a mystery, as the other menu items, had performed opposite.

FROM MEALS TO INGREDIENTS…
Our next adventure was to start digging into the sales of a la carte items. To our small surprise these were also increasing as the foot traffic increased. So why did the #1 item drop? We searched further, and hence changed our perspective of how to look at the sales data. We could continue to look at it from a top-down angle, or we could start looking at it from a bottom-up perspective. In that way we went away from meals, and started focusing on ingredients.

BOTTOM-UP PERSPECTIVE
When comparing sales - item for item - in the menu and through the a la carte sales, we discovered a tiny difference. The salad in the #1 menu had a ‘winterish’ appearance, compared to the other salads, that had a flavor of sun and summer. So, at a specific temperature level, this menu no longer tempted the customers. Since the salads were made in the restaurants, the solution was simple, near and cheap.

This was by no means rocket science. What made it possible to detect, was the interoperable interface of dealing with sales data and weather data.

MOVING ON WITH TWO APPROACHES
The client had two approaches to the issue. Some restaurants changed the salads, whereas others left the choice of salads up to the customers. This pick & mix option worked better, as the customers felt their preference mattered. Every store countered the previous drop, but the stores that implemented the free choice performed even better.

This simple exercise lifted the total annual sales by 7.3% for the overall restaurant chain.