Here is a non-technical article which I hope you might enjoy reading. It’s mainly about how, combining data in a creative way can help improve a business.
This example is based on something I figured out as I was a fresh data analyst, working at Avis in London-based headquarter.
The challenge Avis was facing at the time
It might feels useless to present Avis, as a car company. Maybe one the most successful in the world, always innovating and focusing on customer satisfaction.
Back in 2007 – 2008, Avis created in London a brand new department, focusing on yield management, combining skills from Engineers, Mathematicians and business analyst, to maximize the return on investment, based on forecasting and pricing adjustments.
This was similar to what has been already in place for decades in the airplane industry, hotels or leasure ticketing.
One of the challenges, which everyone can foresee, is that in order to make a good pricing decision, data needs to be reliable: How many cars exactly have each station at the opening of the day on the car park available for the day, at which time exactly a car was handed back to the stations, what is precisely the required turnaround time from getting back a car to prepare it, clean it, and get it available for the next customer.
When calling the regional network manager, the person in charge of the district, we could help but notice sometimes differences what they saw on the ground and the numbers being displayed in the IT control system.
So in one hand, we had to come with a pricing based on the reading of the activity at a daily level but on the other hand, some gaps and delta could sometimes blur how we had to interpret reality.
How to overcome this? How to align what people saw on the ground vs. the data displayed in the system?
One thing very important to know and to understand when it comes to car rental industry is that the offer is not really fixed.
Hotels, planes and trains have a fixed amount of seat / beds available. Once the hotel is full, the hotel manager can’t go beyond the maximum capacity. He/She will have otherwise to push the walls.
When it comes to car renting, cars can move from one pick-up location to another. Let’s say that demand is not strong withing Paris downtown but very strong at International Charles de Gaulle airport.
What do you do then? You move cars from downtown to the airport to catch the demand!
Why this precision?
Cars are expensive – hence cars were rare! In order to minimize fixed cost, fleet had to be moved between pick-up rental stations: 100 pick-ups for a day for a given rental station didn’t mean having 100 cars on the car park at 6AM when the station opened. Stations had to deal with returns and incoming fleet from other stations movement to satisfy all the customers every day.
Precision was key to success! (precision is almost always key to success)
Data-driven actions undertook
This question of system accuracy was a burning topic, not only on the French market, but EMEA wide. The pricing team was located in London, while managers where local with the car park on sight.
It all started by analyzing the database and the data which was stored in it.
I couldn’t help but notice that the database offered the possibility to track :
- The datetime of the actual the customer returned the car
- The datatime when this return information was entirely closed in the system (you need to check the car and clean it one this has been returned, so during this time, the car is not ready to be rent)
This is something very standard in database management. Every new entry (or new row in a table to put it into more concrete ways) has a log timestamp.
The idea I had was to analyze the gap between these two datetimes. Not for a day, but for all rental stations, per month, since 2 years in order to establish a benchmark between rental stations and get some historical trend.
I built a matrix like, showing, for each station, for each month, what has been the percentage of cars “closed” within the same day the customer returned his car, % at day + 1, % at day + 2 and so on.
You could visualize it on a simple Excel spreadsheet, with multiple stacks and sub-stacks, with the months on the x-axis and a y-axis corresponding to the 100% based distribution of the amount of rental contracts closed on Day 0, Day + 1, Day + 2, … , Day + N.
Which was a simple turn-around time measurement revealed to be a more power new perspective than early thoughts in first place; which led to many concrete measures.
While, for confidential reasons, I can’t comment on the financial aspect of it, this simple report surfaced the true reality of the activity – and actually revealed much more than just what we’ve been initaly looking for.
- Differences between rental station based on their location (Airport stations, Downtown stations, ..) emerged showing that not all the stations had the same processes when it came to fleet management and turn-around time. Airport located rental stations could be compared between themselves, establishing a valid point of comparison, for instance, based on their typology, volume of rental and human ressources.
- Fleet management performance improvements or … degradation per rental station over time as rental location managers could leave, not be replaced, which impacted directly the way employees handled the whole process. Having no managers for 3 months was surfacing directly into this KPI.
One “fun fact” was when I looked at a specific rental location and saw 1 rental actually closed after … almost 30 days after having the car “returned”. I took the phone, rang the region manager to figure out what happened. It could have been a simple bug, this happens – but very unlikely thought. It turned out that the car was on a corporate leasing. Contract ended… but the car stayed on the company’s car park… which gates have been closed the whole month of August, as French people went on holidays and the company had its yearly closure!
This report directly lead to process improvement and concrete process measures to improve the process; which ultimately lead to the system monitoring improvement – what we’ve been seeking in first place when I started looking into the data.
This framework was then scaled and rolled-out accross all EMEA countries.