Linear Equations

Remember this equation?

Y = mx + b

The dependent variable, Y, responds to a change in the slope (m) times an independent variable (x), plus the intercept point or constant (b).

Typically, we strip out the letters and replace them with words to make it nice and readable.

(Number of Conversions) = (conversionRate)(PageViews) + constant

What’s the constant? The best way I can describe it simply is “left over unexplained stuff”. Ideally, you’d have an equation that intercepts the Y-axis at 0 (the origin), but, that’s fairly rare, because there are other factors involved, and it’s really rare, at least in marketing science, to get an equation that completely explains everything, perfectly. (Though, it’s not impossible, as the above equation demonstrates).

So then, we can start adding on additional variables into our model in an attempt to make it more complete and more accurate in its predictive ability. For instance: (And just a “For Instance”):

(Number of Conversions) = (rateOfUniqueVisitorsWithPositiveBrandImpression)(numberOfUniqueVisitors) + (rateOfReturnUniqueVisitors)(numberofUniqueVisitors) + (rateOfBudgetOptimization)(TotalBudget) + 990

If you can derive these kinds of equations, then you use them as a form of “foresight”, and you can optimize the future.

The regular statistical laws apply here, you wouldn’t have a very robust model with a sample size of 10 unique visitors or anything.