Smart Tile Grouping Values

[jdemeyer]

I'm using the Smart Tile to review 90 days worth of information.  The results provide groupings up to 3 STD in each direction around an average as I expect.  Is it possible to retrieve the values of each grouping?

For example, when I receive a record returned from Smart Tile that contains "Extremely High (Above 5600)", how might I retrieve 5600 from Smart Tile?  Also, how might I received 5600 from Smart Tile even when there is no record that contains it?

Thanks!

Joe

[apathetichell]

unique tool first - regex tokenize - voila.

[jdemeyer]

I appreciate the effort and agree that solution extracts the cutoff values.

My report requires that I show the cutoff values for "Extremely High" and "Extremely Low"; your solution shows both.  Excellent.

Now, Smart Tile does not always provide a record for "Extremely High" since the data may not contain values for that bucket.  I would still like to have the cutoff value for "Extremely High".  

Thoughts?

Joe

[apathetichell · Accepted solution]

How about:

run the figures you are looking at through a summarize tool to get the average and the standard deviation.

If they are normalized to log 10 - keep that in mind. the numbers I'm running are so the examples below will be.

pow(10,[Avg_LOG10(Sale)]-(2.5*[StdDev_LOG10(Sale)])) should match your extremely low

 and

pow(10,[Avg_LOG10(Sale)]+(2.5*[StdDev_LOG10(Sale)])) should match your extremely high threshold.

If your numbers aren't log10-ed then obviously don't bother with the pow() function. To test your data try running the same formulas with 1 and 1.5 stdev to see if it matches the Alteryx results.

[jdemeyer]

Neat!

I've read about the log 10 thing.  This looks like an intriguing solution and I look forward to giving it a try.  Give me a day or so!

Thanks!

Joe

[jdemeyer]

Good Morning!

Here are my results.  The table provides a comparison of Smart Tile cutoffs to calculated cutoffs.

Name                    Smart Tile     Calculation

Extremely High      5600              5622

Extremely Low       2600              2583

These seem reasonably close.  Perhaps the Smart Tile has some rounding logic for assigning the cutoff.

Thoughts?

Joe

[apathetichell]

Yes - I think there is some rounding for visual clarity - and glad to hear it worked.  Thanks for marking as a solution - there have been a slew of questions on Tile math in the past so hopefully people can find this as a solution.

[jdemeyer]

I have a follow up on our rounding comments.

I created a macro to do this calculation.  I needed to have the cutoff values for several states.  I completed the macro and it worked well.  I thought I'd try some rounding so I tired ROUND and CEIL.  The values were very close to those found in the SmartTile_Name field (this is an optional field generated by SmartTile).  After some research on other rounding possibilities, I found SMARTROUND.

I place SMARTROUND into the macro and the values of the cutoffs matched those found in the SmartTile_Name field.

Thanks again!

Joe