1 Reply Latest reply: Jun 8, 2010 8:43 PM by Oleg Troyansky RSS

    Advice Needed: Visualization of Buying Patterns


      Hi, I'm looking for some advice on a new set of dashboards I need to build to analyze buying patterns. The companies will belong to (or not belong to) various sets. For example:

      1. Companies that have bought (or not bought) within the past month/3 months/6 months/12 months (or 2/3/4 quarters). This would be driven by filters. For example, the company may have bought ProductLineX, but not ProductLineY. Also, there will be a minimum purchase amount. For example, if a company only bought $100 or $1k, we may want to ignore them.

      2. Companies that have consistently made (or not made) purchases in 5 of the past 6 months.

      3. Companies that are growing (or shrinking) relative to themselves or the rest of the market. For example, the company may be growing in general, but shrinking in ProductLineX.

      I've been racking my brain on the best way to visualize this. I've viewed the various example QVW's out there, but haven't found anything close. Is there an example of a dashboard that is similar to this?

      The other concern is performance. You can see that this application will rely heavily on set analysis. I expect about 2 years of historical POS data to drive these dashboards, so the transaction set will likely be in the millions. Is QlikView a good tool for this type of application?

      Any pointers you can give would be greatly appreciated!

        • Advice Needed: Visualization of Buying Patterns
          Oleg Troyansky

          This is sort of an open question, very difficult to answer in a brief forum post... Some scattered thoughts:

          • This is perhaps the wrong forum for the question "Is QlikView a good tool for this type of application?" We do everything in QlikView here...
          • Detailed POS data tends to be very heavy, especially in its detailed raw format. You should be ready with some good hardware - 64 bit server, with lots of RAM and CPUs. Hard to estimate without knowing your data size.
          • Set Analysis, from the performance perspective, is actually a big help - it's the best performing technique there is, because it works via reducing the data set that is going to be aggregated (as opposed to testing every row for a specific condition).
          • I think you need to separate visualization questions from the conditions, at least in your thinking process. Whatever your conditions are, visualization will depend on the type of analysis that you need to perform. When you say "Companies that purchased ...", what are you after - the number of such companies? The list of them? The comparison between their Sales numbers? % of total? Each one of those questions and answers calls for a different type of a visualization.
          • I could see a screen with the number of charts, showing a "standard" set of visual pictures - perhaps a pie with the market share based on a certain Dimension, and a scatter chart with the "Magic Quadrant" of all Companies plotted over 2 major KPIs, and a histogram showing how many Companies fall into certain buckets, and a chart showing % of total. Along that, I'd offer a dynamic selection of "Filters" - Companies that "bought" or "not bought" within certain period etc... For each"filter" option, you'd define the Set Analysis condition that needs to be used in all Chart Expressions. You could even develop "smart" formulas that could combine multiple filter conditions into a single Set Analysis expression.
          • So, bottom line - yes, I think QlikView IS the right tool for the job. No other tool (that I know) can do Set Analysis as well as QlikView does...


          good luck!