How to test demand factors
If you want to win Global Management Challenge, you need to know not only exclusive tips and bugs, but also you must correctly and accurately forecast sales for each product. This tip describes basic method of testing demand (market growth) factors.
As everyone knows, simulator has 3 types of markets (Nafta, EU, Internet), and 3 types of products 1, 2 and 3. The result is a 3x3 matrix. Sales in each cell of the matrix depends on particular set of factors. One part of factors affects on the whole matrix, second part of factors affects on the particular type of market or product, remaining affects strictly on one of the cells. For an accurate demand forecast is necessary to determine the dependence of each of the matrix cells from a common set of factors.
For example, corporate advertisment effects on particular type of market, assembly time effects on particular type of product, direct advertisment effects on its sole cell.
How to test factors
When you set the influence factors of demand for each of the cells (3x3 matrix), you can continue to refine the power of influence. Now we start to test it. Tests can be two types - perfect tests and other. Perfect tests are the most valuable because they give absolutely accurate assessment of the impact factor on demand. Let see how factor can be tested on corporte advertisment example from the past topic:
You shoud have two or more teams in one group. The main thing that the teams must be in the same group, otherwise the test will not be perfect (and if the influence of factor is weak, it would be meaningless). It is impossible to compare test results of two teams from different groups correctly, because strong influence on the results will provide competitors in the group.
Prepare an identical general decision for both teams, but in one of the teams change a decision on the tested factor. For example, corporte advertisment test was made with the values of 30 (basic team) and 80 (test team). In cases when victory in the group is not nessery (for example, if you have about 200+ teams :) or you play in demo round, choose for your teams general decision from the 5 report in history. Part of factors has a strong residual effect from past periods and this allows us to neutralize their influence.
1 period - you have posted decisions and get reports from your teams from the same group. Team with the base decision (corporate advertisment 30) has sales 900-450-225, test team (image advertisment 80) 900-450-225. Conclusion that corporate advertisment has no effect on sales in the current period.
2 period - basic team (30) has sales 1000-500-250, sales of the test team (80) 1070-55-268. Lets take a look on product 1. If we do not have a team with a basic solution in the group, we might think that investments in corporate advertisment increased sales from 900 (1 period) to 1070 (2 period), growth is 1070 / 900 = +19%, but this is a mistake. Seasonal fluctuations in demand and other competitors change sales in the group. To neutralize these changes you need basic team. Compared with sales of product 1 between basic team and test team, growth is 1070 / 1000 = +7%, ie much less. This is called a perfect test because allows you to find the true impact of the tested factor.
3, 4 and 5 period - continue test with values of 30 and 80, we estimate the cumulative impact of corporate advertisment. 3 period basic team (30) sales 1100-550-270, test team (80) sales 1212-606-298. 1 product growth is 1212 / 1100 = +10.2%, the difference increased, manifested cumulative effect. To assess the cumulative effect, continue test with the same parameters to 5 period.
In the same time perform several tests with values of corporate advertisment 30-0, 30-45, 30-60, 30-80, 30-99. Obtain all results for the entire range 0-99 of the corporate advertisment and arranges them on the graph. It is important to test the full range to determine the type of connection between demand and factor for further modeling of the market - directly proportional, quadratic, etc.
Assessing the impact of factors on sales
As has been previously determined, the influence on sales by corporate advertisment is directly proportional, for the evaluation single test is enough. In excel make a table with the results. To estimate use relative change in sales, otherwise you will not be able to apply this formula to predict demand in the other groups.
Product 1 (relative sales)
Product 1 (absolute sale)
Add a trend line and a formula on the chart. Seeking coefficient is 0.0014 - elasticity of the factor, it means that increasing corporate advertisment by 1 unit gives growth in sales for 0.14%. It is easy to calculate that 99 corporate advertisment increase sales by (99 - 30) * 0.0014 = 0.0966, ie +9.66%
To evaluate the residual effect it is necessary to compare the results of several periods and choose the factor that takes impact on sales of investments from previous periods.It can be found general general formula:
Δ sales (5 period) = Δ corporate advertisment (4 period) * K + Δ corporate advertisment (3 period) * K * X + Δ corporate advertisment (2 period) * K * X² + Δ corporate advertisment (1 time) * K * X³, where K = 0.0014, and we know all Δ from the test. We find that X = 0.6 - 60% of the investments from previous period continue to have an impact on demand in the next period.