Most
out-of-the-box statistics software packages have the capability of analyzing sales trends using techniques
such as time-series analysis, multiple regression, and other various curve-fitting algorithms. They excell
at quickly answering critical business questions such as, "Are sales increasing or decreasing,
and at what rate?"
You may have noticed we
track weather forecasts (In fact, we collect weather forecasts
from over
700 locations across the country, up to six times per day). We also include
an extensive list of holidays that show up as "variables"
in our
free online multiple regression tool.
We aren't secretly researching global
dimming, nor are we interested in when New Year's eve will next occur on a Friday
night (December 31st, 2010, FYI). And there
is a good reason we examine a "Mothers Day x Passage of Time" interaction; knowing how many days pass between the New Year and the second Sunday in May has a strong influence on the number of sales
one of our clients makes.
If you analyze
sales figures and dates, your results offer insight into
how sales change over time. If you combine sales figures with
environmental and psychological factors that may have influenced
customer
behavior on each date-of-sale, your results can offer insight into
the
behavioral patterns and
preferences of your customers.
A subset of the variables we investigate during analysis (and consider as possible
predictors when
developing behavioral models) is listed below:
|
Simple Effects: |
Low and High Temperatures |
|
Passage of Time (General Trend)
|
Daytime Temperature Range
|
|
Weekend vs. Weekday
|
Fog, Cloudcover, Thunderstorms, etc.
|
|
Many Holidays
|
Wind Speed - Night/Day
|
|
Days surrounding Holidays
|
Hours of Precipitation - Night/Day
|
|
Precipitation - Night/Day
|
Hours of Snow - Night/Day
|
|
Snow - Night/Day
|
Wind Gusts - Night/Day
|
|
Hours of Sunlight/Darkness
|
Wind Direction - Night/Day
|
|
24-Hour Temperature Range
|
Temperature deviation from Average
|
|
Several 2-way Interactions: |
Holiday x Time-of-year |
|
Day of Week x Time of Year
|
Weekend/Weekday x Weather Events
|
|
Precipitation x Day of Week
|
Precipitation x Temperatures
|
|
Precip x Temperature Range
|
Precip x Time-of-year
|
|
Windspeed x Wind Gusts
|
Holidays x Day of Week
|
|
Temperatures x Day of Week
|
Temperatures x Below Freezing
|
|
Temperatures x Wind Speeds
|
Temperatures x Weather Events
|
|
Temperature x Time-of-year
|
Temperature x Holiday
|
|
Freezing x Time-of-year
|
Precip x Monthly Average
|
|
Some 3-way Interactions: |
Time-of-year x Temp Range x Day-of-week |
|
Temp x Holiday x Day-of-week
|
Temp x Precip x Time-of-year
|
|
Wind x Precip x Temperature
|
Temp x Weekend/Weekday x Time-of-year
|
|
Wind x Day-of-week x Precip
|
Holiday x Day-of-week x Weather Events
|
Using certain variables may exclude the use of other variables.
(If we're interested in weekend vs. weekday sales trends over the past year, there
is no need to investigate each day of the week as a separate predictor). Also note
that we use nonlinear modeling, so we consider that the relationship between a variable
and the number of sales made may be best described using a polynomial equation.
Or, we may transform a variable (by considering its inverse, log, etc.) if the transformed
variable retains logical meaning.
(
Back to the home page)