• Kohki Yamaguchi

    A technical note that the second peak/cluster is actually a collection of peaks, it only appears bimodal on a log scale. The pattern is in fact a daily one, so there are as many local maxima as there are days in the look-back window.

  • Luca Naso

    Another example of how Data Anlytics can help to burst conclusions based on common sense.

    It makes sense that buying more expensive products requires longer conversion lag. If this was the case, then in the second plot one should see a cloud of orders in the upper-right part of the diagram, say $1000 and 10 days.

    However the second plot in this example shows that this is not the case.
    Large conversion lags (e.g. 10 days) are seen for both low- and high- price orders ($30 and $400), and actually the price-orders range looks the same for both the short and the long time lag peaks.