STEM newsletter

Modelling distributed car hire

30 October 2007

Although its default icons look like telephones and servers, STEM’s graphical interface can represent the processes of virtually any business, allowing users to quickly grasp the fundamental questions facing a proposed investment and examine the impact of operational choices on its viability. This has been demonstrated at the last two STEM User Group meetings, where participants have used STEM’s icons to represent airplanes, airports, cars and engineers.

In 2006, the interactive modelling session determined the lowest (profitable) fare on a budget airline. Following the popularity of that exercise, the 2007 meeting offered two more non-telecoms modelling tasks: modelling an online conference provider and investigating the revenue and cost drivers of a car hire company. This article describes the car hire session, in which some innovative thinking about unfamiliar topics led to a final model that was quite different from what the session organisers had in mind.

How does a car hire company work?

The modelling team were asked to put themselves in the place of an entrepreneur looking to start up a car rental business. Conventional car rental companies are built on over-the-counter transactions and centralised car storage. That model clearly works well, as such companies have captured the most valuable and high-volume segments of the market, namely business rentals and airport rentals. Catering to casual urban consumers without cars has never been a high priority, particularly with car ownership and overall levels of income rising. Could a car rental business targeted at this segment be successful?

The hypothetical company was to begin operations in a sizeable city like Birmingham in the UK, with about one million residents. Some background data was provided (see Annex) which suggested that a suitable target market would be residents in densely populated urban areas (above 25 households per hectare) where parking is a problem, as indicated by low car ownership (see figure below).

Target market (blue area) is communities with a high density of households and low car ownership (Source: Transport Planning Society, 2006)

The key issues the team was asked to address were:

  • How does the business work? How many cars will be required, where will they be located, what tariff structure will be used, and what additional services will be required? These decisions would essentially define the iconic representation of the business in STEM.
  • How will the enterprise be funded, and how much money will it make? The metrics that any startup needs to present in its business plan are NPV, peak funding, time to break even and to pay back the investment. Profit margins will also be important as input to determine how low prices can go without losing money.

Apart from the basic business case brief, and a few suggested numbers to input into the model (included in the annex of this article), the team was given carte blanche in devising a business model for the car hire industry. And the modellers certainly did use that freedom: the model that they built bore little resemblance to the prototype assembled by the organisers. The team even decided that the cars suggested for the exercise were too expensive!

Building the business

The first priority was to decide where the hire cars were to be kept when not in use: all together in car yards (the approach used by conventional car rental companies), or distributed in convenient public parking close to the target communities. In the spirit of putting STEM through its paces and thinking about the value drivers in an innovative business model, it was decided that the latter approach was of more interest. But could a distributed car rental business targeted at the casual urban user without a car be successful?

Using a quickly assembled model, the team evaluated some options for how far apart the cars should be located (for example, in dedicated spaces in public car parks or using on-street parking). The convenience of the car hire service depends on the maximum distance that a customer would have to walk to find their closest car location, and the figure below depicts the impact different distances would have on average utilisation of cars.

Shorter walking distances between cars result in lower average utilisation. The distances on the horizontal axis represent maximum distance to walk to the closest car. The working model assumes customers need walk no further than 800m.

Arriving at a car location, however, doesn’t necessarily guarantee that there will be a car there. Implicit in the figure above is a demand model, which can be one of two types:

  • Peak-driven, where the total number of cars required is determined by how many concurrent car rentals are expected in the “busy hour” (or, more generally, the busy period). In telecoms this is typically used for time-sensitive or circuit-switched services.
  • Volume-driven, where the total number of cars is determined by the average “bandwidth” (the average number of cars out for rental each day) and the contention ratio, which is a rough measure of how intermittent we expect demand to be. The telecoms equivalent of this is “best effort” data services.

The team chose a peak-driven traffic model as car renters are likely to be highly sensitive to the absence of a car after they’ve walked a few hundred metres. The model that was eventually built around this is shown below.

Final version of the distributed car hire business model

Two services are modelled: Car rental event and Car rental hours, which allows fine-grained control over tariff structure: we can charge an initial setup fee, monthly fees, a fee per new rental, as well as fees per hour or day of rental. In addition we have replicated the services to allow us to have three different tariff structures for high, medium and low volume users.

The results of the team’s three-and-a-half hour discussions and model-building are shown below. We managed to make the business profitable, although the results suggested further questions for investigation – in particular, the total number of cars required peaks at over 1000, to be distributed over 97 sites. Apart from the practical feasibility of such a large number of cars per location, there was a general feeling that the traffic model (pun not intended) required more investigation.

Financial results for the distributed car hire model

A new feature in STEM 7.1, the tornado chart (shown at the top left in the figure), came in particularly handy for investigating the sensitivities of the business model to key parameters. As can be seen, the NPV after 10 years of operation is about EUR8 million, and the tornado chart shows the impact of variations of up to 50% in a few key parameters on this NPV. In particular, the company loses money (the NPV is negative) if the cars it is hiring out are too expensive. The chart on the top right also clearly shows that this is a depreciation-dominated company – so finding ways to cut down initial capital outlay, and maximise the residual value of the cars, would have a large influence on profitability.

The UK company Streetcar is actually targeting exactly the market segment in our modelling. It deploys a fleet of cars at over 440 locations in London (plus a few in other UK cities), and a car can be rented for GBP4.95 per hour. Membership of the scheme costs GBP49.50 per year, and that gets you a smartcard which gives access to the cars, plus a login to the company’s Web site, allowing you to book a car.

There are many details in the service provided by Streetcar which were not even touched on in our modelling exercise; but then the founders did spend 18 months on their business planning. It is not entirely clear whether their service includes rotation of cars between locations in order to ensure that cars are almost always available at any spot – as in our model – but this is somewhat balanced by Streetcar’s ability to check car availability and book while on the move. Despite the reservations of our User Group interactive modelling team – who thought that taxis would always be a more convenient and economical option for lower usage car users – Streetcar is riding a wave of public support for traffic- and pollution-reducing measures, and has been growing steadily since its launch in 2004.

Annex: Cost and revenue drivers

On the revenue side the following inputs to the model were suggested:

  • 40% of urban households don’t have a car (according to the UK census of 2001, the proportion varied between 25% and 50%), and the distributed car rental market is 40% of households without a car.
  • UK urban areas have a density of 25 households per hectare.
  • Subscribers to the imagined service pay EUR45 per month membership and EUR35 per full day of rental, but typically take the car for only 10 hours at a time (charged pro-rata), four times a month. The market share declines from 95% (due to uniqueness of service) to 50% in five years.
  • Non-members (pay-as-you-go users) pay EUR35 per full day, and typically take 4 two-day rentals during the year. Market share is 45%.

The following table presents the suggested cost drivers for the car hire business – although cost estimates were revised based on discussions during model building.

Cost Item Cost
Car EUR18k, three year lifetime, 40% residual value
Shop location EUR45k p.a. for 120m²
Car yard EUR15k p.a. per 100m² (10 cars)
Engineers 1 for every 40 cars, EUR35k p.a. (minimum two per yard)
Sales staff 1 for every 20 cars, EUR28k p.a. (minimum two per shop)
Insurance EUR1000 p.a. per car
Spare parts/repairs EUR500 p.a. per car
G&A overhead 15% of revenues
S&M overhead At least EUR20 p.a. per customer

© Implied Logic Limited