Different results when selecting from Periodic Bar vs Straight Table

[Akio]

I have a Qlik Sense dashboard with:

. 100% stacked bar chart

. Straight table

The source Excel file has 2 sheets:

. Monthly Data

. Periodic Data

Bar Chart Logic

Monthly bars are built from Monthly Data.

Periodic bars (Quarterly / 1st Half / 2nd Half / Yearly) are built from Periodic Data.

Each bar shows the percentage distribution of 〇 / △ / × / ★ categories, and the total at the top is the

distinct product count.

For periodic bars, the calculation uses the latest month in the selected period (for example, September for

1st Half).

Issue

Scenario 1 - Selection from the Periodic Bar

. Select 1st Half

. Select x

The straight table shows records for the latest month (September).

Scenario 2 - Selection from the Straight Table

. Clear selections

. Select x in the straight table

. Select 1st Half in the bar chart

The straight table now shows records for the entire period (April-September), resulting in a different count.

Important Observation

The latest month data in Periodic Data is not identical to the latest month data in Monthly Data.

For the same month, the distribution of 〇 / △ / × / ★ differs between the two sheets.

My Question

Is this behavior expected because Qlik Sense is filtering based on the currently selected dataset in the

associative model?

Since the periodic bars come from Periodic Data and the straight table comes from Monthly Data, and the

underlying category distributions differ, is it correct that different selection paths can legitimately produce

different results?

What I want to confirm from the community

. Whether this is expected associative behavior in Qlik Sense.

. Whether the inconsistent source data is the primary reason for the different results.

[marcus_sommer]

It sounds that the associations of your data aren't not suitable for the wanted views/usability. This is often caused from a multi-facts scheme instead of using a star-scheme data-model with appropriate harmonized data.

Transferring the data into such a star-scheme may not be trivial but it's much simpler and easier as handling any alternatives.

[Akio]

Just to confirm, is the root cause here the fact that:

1. The bar chart is using the Periodic Data fact table.
2. The straight table is using the Monthly Data fact table.
3. The judgment (〇 / △ / × / ★) values are not harmonized between these two datasets.

Therefore, depending on which visualization the selection originates from, Qlik is filtering different facts, so different results are expected. Is my understanding correct?

[marcus_sommer · Accepted solution]

It's quite likely that using two different data-sets are leading to different results which may caused from:

• differences within the facts itself (number of records, equally periods, granularity, data-quality, ...)
• differences by their dimensions (same key-fields as well as the same (missing) key-values to all sides)
• inter-actions respectively any association between both data-sets (it's not purely the filtering else also all associations between the model-tables as well as those associations which were created between the n dimensions/measures within the used visuals)

From this point of view it's a likely and expected behaviour. IMO it should be avoided because it leads too easily to confusions. Either designing it as a single data-set or distributing it as two applications.