Transfer of trade-ins as a part of everyday business
An optimisation model for stock lets a dealership chain know which cars it should transfer from one outlet to another to maximize the expected profit. Thus, optimisation supports the managers of a dealership in everyday decision-making.

Car dealerships are constantly linked into larger chains and this phenomena has offered new possibilities in developing their procedures. It is no longer necessary to sell the trade-ins in the dealership where they were originally purchased, but they can be transferred within the dealership chain in a way that is optimal for the whole chain. This provides many benefits:

- The stock status of a single dealership doesn't govern the price offered to the customer for the trade-in.
- Each dealership has a stock mixture that best corresponds with the demand.
- The average selling time is shorter and the need for storing is reduced.

The systematic applying of Stock Optimisation creates a new need for data to control the transfer of vehicles.

The problem is the decision making in a situation that offers a vast number of possible  placings for vehicles. In the simplest situation between two outlets the problem is still easily manageable

Knowing that three vehicles should be transferred from outlet 1 to outlet 2 is of no use to the decision-maker. To benefit from this data, it's necessary to know as accurately as possible the type and model of the cars that need to be transferred. This data can be processed to the level of individual vehicles, thus the only thing left is for the transport company to pick up the cars from the dealership.

As the number of outlets grows, and possible wholesale market sales come into the mix, the number of variables grows and grows. With five outlets, the situation looks quite complicated. Take into consideration that many of the biggest dealership chains have dozens of outlets. For example with 20 outlets, the number of variables would be 380, and with 40 outlets far over a thousand. In addition, as the stock status changes constantly, the same problem needs to be solved over and over again. It is clear that it gets impossible to find and optimal solution without sophisticated calculation methods.
 
Traditionally, the stock optimisation models have been used to solve problems in industrial engineering and logistics. For example, an industrial enterprise can create production plans for separate production plants and time and measure purchases according to expected demand. Similarly, a transport company can plan its routes in a way that minimizes the journey lengths and has all transports reaching their destinations on time.

Actually, the problems in car dealership chains are not that much different from the above examples. The chain has a particular number of different vehicles in its outlets, with which it aims to meet the expected demand. However, unlike the transport company in the example, the aim is not only to minimise the expenses, but also to maximize the overall profit. The expected selling times in different regions and outlets vary considerably, and the wholesale market naturally has its own price level. The optimisation model has to notice the effect of properties, such as age and mileage, at an outlet level. This ensures that the recommended transfers are aimed to the outlets, that provide the best price for the particular type of car.

The creation of an optimisation model is quite simple, once all the needed information is gathered, in this case meaning the current status of stock, the cost of vehicle transfers and expected selling prices and times of individual vehicles in each outlet. Mathematical optimisation model comes up with the solution that provides the maximal expected profit for the entire chain. It's however clear that the results of the model are highly dependent on the values on which it is based.

The biggest challenge in this respect is in predicting the demand and quantizing the effects of demand. Because the model maximises the profit, it's natural to show the effect of the demand in an amount of money. If a model or type of car is overrepresented in stock compared to the need in predicted demand, the model revises the expected selling price for that particular outlet. This default is essential in reaching a balance between supply and demand, where each dealership has a stock mix corresponding with the expected demand. It's not realistic to expect the same price for a vehicle regardless of the number of stock.

In predicting the demand, the basis can be in data of previous sales, the trends of sales and seasonal variation. On the other hand, the user can have an own view of the target situation, i.e. how many vehicles each outlet's stock should include. To that end, you can feed target parameters that are taken into consideration in calculating the stock need for particular vehicles. Thus, the recommended transfers move the stock sizes towards the target, even in situations, where the stock size for some outlets should be in a different level than it is.

To utilize the results from optimizing it's necessary to pay attention to the way the results are illustrated. The recommended transfers should be reported as simply and clearly as possible, so that the decision-maker knows in detail the number of vehicles of each car type to be transferred and the outlets into which the vehicles are going. Reports can be produced for the entire dealership chain or including the recommended transfers for just one outlet. Thus, people in different positions in the organization get the information that is essential for them. In addition, it is interesting to find out how much the execution of recommended transfers will improve the expected profit for the entire chain. In this way you get a just transfer price for each vehicle, so that both the selling and receiving outlet profit from the transfer. In addition, you can do the transfers that provide the biggest benefits first. However, you should not ignore smaller profit improvements, because they are pure profit to the chain, from which all expenses of the transfers have been deducted.

The use of the optimisation model has given good results both in Finland. The model has proved to be functional in practise, which can be seen in the fact that the results (recommended transfers) are not sensitive to small changes in the source information or parameters. In this way, a vehicle that has been once transferred is very rarely recommended for a transfer again. In addition, the transfer recommendations are reasonable, it's easy for the user to understand why the transfers lead to a better end result.

In this context, we have only focused on vehicles already on stock. Additional benefits are reached with optimising the purchases, where data of stock and demand is noted already in the purchasing stage. In the future, when purchasing a vehicle, the dealership will know in which outlet to sell the vehicle to gain most profit, or whether it's profitable to sell it to wholesale market. Then the price level of purchases can be adjusted according to the needs of the whole chain. In an ideal situation, the stock sizes stay optimal, and the need for transfers declines. At the same time a large chain can benefit from its size, as its outlets can purchase vehicles that have no demand in the particular outlet.

As a conclusion, it must be stated that it's not possible to utilise even the best of mathematical optimisation models without sufficient and correct source information that is based on systematic collecting and processing of data. In addition, mathematical optimisation is a heavy calculation process and demands software that is especially suited for optimisation. Once the technical requirements have been met, it's only a matter of a change in procedures. The transferring of trade-ins will become a natural and routine part of the operations of a car dealership chain.

>> See product info

Esa Heimo
Product manager
Autovista
Last Updated ( Monday, 02 June 2008 )