Investors and operators have a long tradition of monitoring industry trends. Trends in observed rates provide valuable information for assessment, evaluation and planning. Typically, trend data are presented for rates arising from the population over time.
For example, the benchmarking survey reports published by the Professional Carwashing and Detailing magazine are a popular source for trends of carwash volume, price, revenue, operating expenses and other variables. The nationwide rates described in these reports are believed to be representative of the carwashes responding to the surveys and are available over many years.
Since these rates are considered as true, they are usually presented “as is” in graphs and tables so that comparisons or predictions can be made intuitively. Regardless of what methods are used for analyzing the data, the first step in assessing a trend is to plot the numbers or rates over time.
Graphs are indispensable for becoming familiar with the absolute and relative levels of the data, for understanding the overall direction and shape of the trend, and for presenting the pattern of change over time.
For example, by looking at the data in Table 1, we could conclude that the trend in sales volume for self-serve carwashes is linear and appears to be stationary for wand-bays and increasing for in-bay automatic. Since the data appears to be linear, we could test this by fitting a straight line to the time series using a technique called regression analysis.
The result of the analysis is represented by the straight line that is fitted to the sales volume data points for in-bay automatic shown in Figure 1. The basic purpose of the “best fit” line is to summarize the data trend and evaluate the slope of the line and its significance.
The slope is the most important part of the trend model. It represents the rate at which change occurs over time. If the slope is zero, there is no trend. If the slope is positive, the rate of change is increasing. If the slope is negative, the rate of change is decreasing.
Simple trend analysis may be appropriate for examining rates of change over time but it may result in misleading conclusions if it is used to predict the future. For example, most time series in business are not stationary and exhibit various types of trends, cycles and seasonal patterns.
Trend analysis is also time sensitive and simple methods do not account for variables which may be unobservable but are highly correlated with time such as consumer preferences and technological change.
The regression line in Figure 2 shows why the linear trend would be a poor forecasting model for predicting the expected change in the number of self-service facilities with in-bay automatics. From 1991 to 2000, the amount of distance or “error” that exists between the data (observation) and the points along the trend line (theoretical) is pretty small.
However, from 2000 to 2006, the margin of error is considerably greater and the pattern of the data has become more irregular. It is this deviation or wobble that occurs near the end of the time series that reduces the best fit for the next few periods.
We can use simple trend analysis to test whether the slope value of the regression line is significantly different than zero. If it is, we could conclude that we have a linear trend. If it is not, we must conclude there is no meaningful trend. This can be accomplished with the P-value.
The P-value is the probability of obtaining a result at least as extreme as that obtained, assuming the truth of the theory that the finding was the result of chance alone. As a standard rule of thumb, a P-value of less than 0.05 is statistically significant. In the presence of P < 0.05, you are safe to reject the hypothesis or theory that there is no trend.
To help readers unfamiliar with regression, we developed a template which we used to do simple linear trend analysis for a select number of variables within the self-service segment of the carwash industry.
Data for the 15-year period of 1991 to 2006 was taken from PC&D annual benchmarking survey reports and other industry sources. In the interest of space, all data and graphs are not shown.
Of the variables we examined, 87 percent of the time series exhibited an upward trend and 77 percent of these trends are linear. However, only 40 percent of the linear trends are considered statistically significant. These variables included location age, wand-bay price, operating expense, capital cost, vacuums per site and average revenue per vacuum.
Table 2 shows the findings for the time series that are exhibiting upward linear trends and for which we have an acceptable level of confidence. Table 3 shows the findings for the time series that have linear trends but lack consistency.
We have shown that historical data and simple trend analysis can be used to help describe some of the general characteristics of the carwash industry. We have also shown why investors and carwash operators should not use industry trends to describe the behavior of an individual firm or use simple trend analysis to make predictions of future occurrences.
Bob Roman is a former carwash, express lube and detail shop operator and is president of RJR Enterprises (www.carwashplan.com
), a leading consultant to the carwash industry. Robert is a member of International Carwash Association and PC&D’s Honorary Advisory Board. He can be contacted at firstname.lastname@example.org