We all know that a good set of KPIs is composed of lag and lead measures. Lag measures refer to achieved result, lead measures are predictor of future performance.

Financial results (such as sales) are typical lag (or lagging) KPIs, while client satisfaction is a typical lead (or leading KPI): if customer is satisfied today will probably buy again tomorrow (customer satisfaction is the subject of a future post).

Lagging KPIs are someway intangible or abstract. We can figure many of them, but we always keep asking ourselves is that KPI predictor of a lag KPI or measure?

Let's consider a simple example: yuo would like to verify if New Visitors Rate (New Visitors / All Visitors) is a valid lead measure for sales, given the following data:

Here is a simple trick to verify how much a lead indicator – V(i) for period i – is effective in respect to a lag indicator – L(i).

1. Disassemble the KPIs into its component and consider only those that are expected to impact. In the case of New Visitors Rate, we consider only New Visitors. Repeat the following steps for each of them.

2. Guess leading span(s): how many periods lead indicator will precede lag indicator. If you are unsure try with different values repeat the following steps for each guessed value end pick best result. In our case we expect that a new client is converted on about 3 months and guess a leading span s = 3.
   

3. Smooth lag values to compensate leading span. I use this formula with smoothing constant a = 0.7. S(i) is smoothed value if period i and V[i] is lead variable value at period i.

   S[i] = V[i] × a + ( V[i - 1] + V[i + 1] ) × (1 – a) / 2
   

4. Plot a scatter diagram – XY chart – of smoothed lag value after leasing span – S(i + s) on Y axis – as a function of lead measure – L(i) on X axis. In our case:

As you can see dot dispersion fits a straight line and we can consider new visitors (thus their rate) a good predictor for sales 3 months later.