About Grouping

The main purpose of grouping is to remove duplicates so search results are easier for users to read.

With grouping, your results only retains a hit (or N hits) for each group. The particular hit that displays for each group is determined by how you sort the hits within the group. Within these groups, you can nest additional grouping clauses.

Important: As with other search-time manipulations like facets and sorting, you can only define group or group sorting expressions on fields that are stored in RAM.

While grouping in Exalead CloudView works similar to the SQL GroupBy in that you can use multiple clauses and create aggregations, keep in mind that it is not identical to database grouping.

For search results, the ultimate purpose to collapse several hits into one (or several) "representative" hit. The following examples demonstrate how grouping works.

This page discusses:

Simple Grouping Example
Multicriteria Grouping Example
Displaying an Aggregation Example

Simple Grouping Example

Say that you are sporting gear retailer. Spring is coming and you want to find out what kind of bike merchandise you have in stock.

Without using any grouping, here are the results for a query on bike:


Hit rank in results	Name	Quantity	Warehouse distance (km)
1	Bike	10	100
2	Bike helmet	100	1
3	Bike seat	50	12
4	Bike shoes	20	28
5	Biking jersey	500	5

You want to reduce this list by grouping the results on which the quantity is higher than 50. To do so, define a group where quantity > 50.

How Many Search Results Do You Have?

Two, because defining a group creates 2 groups:

More than 50: Bike helmet, Biking jersey
50 or fewer (the "other" group): Bike, Bike seat, Bike shoes.

Which hits Do Display for the Group?

This is defined by:

The number of hits to represent the group. Let us keep the default, 1. This means that 2 hits are displayed in the search results, one for each group.
The sorting expression for the group. In our example, you want to know what's close by, so let us sort on the distance field in ascending order. This means the hit with the lowest distance value in the group represents the group.

In the grouped search results, this returns:

One hit from More than 50, Biking helmet (because distance = 1 km, less than for Biking jersey).
One hit from 50 or fewer, Bike seat (because distance = 12 km, less than for Bike or Bike shoes).

Which Group Displays First in the Search Results?

The group sorting determines which hit represents the group. But ultimately it is the representative hit's relevance score before grouping that determines the group’s rank (display order) in the search results.

In this case, Bike seat ranked higher than Biking jersey before grouping. So the grouped results look like this:


Grouped hit rank	Name	Quantity	Warehouse distance (km)
1	Bike seat	50	1 km
2	Biking jersey	500	5 km

Multicriteria Grouping Example

You can add additional clauses to group on multiple criteria.

Without using any grouping, here are the results for your bike query, this time with country information.


Hit ranking	Name	Quantity	Warehouse distance (km)	Country
1	Bike	10	100	Germany
2	Bike helmet	100	1	France
3	Bike seat	50	12	Germany
4	Bike shoes	20	28	France
5	Biking jersey	500	5	France

In addition to grouping by quantity, you also want to group by country.

This means you now have three groups:

More than 50 in France: Bike helmet, Biking jersey.
50 or fewer in France: Bike shoes
50 or fewer in Germany: Bike, Bike seat.

Still displaying only one hit from each group, and sorting in ascending order on distance, this will return three hits:

Bike helmet from More than 50 in France
Bike shoes from 50 or fewer in France
Bike seat from 50 or fewer in Germany

Since CloudView ranks these "representative" hits to their initial rank before being grouped, the hits display in this order:


Grouped hit rank	Name	Quantity	Warehouse distance (km)	Country
1	Bike helmet	100	1	France
2	Bike seat	50	12	Germany
3	Bike shoes	20	28	France

Displaying an Aggregation Example

You can also calculate an aggregation for group and display it in the grouped hit. Note that the aggregation is based on all hits within the group, regardless how many hits display in the grouped hit.

Available aggregations:

Sum
Max
Min
Average
Standard deviation
Concatenation - this lists all the values for a field within the group. Unlike other aggregations, it needs an alphanumeric field.

Say that you want to display the total quantity in stock for all hits within a group. Let us go back to the simple grouping example with only two groups.


Group	Name	Quantity
More than 50	Biking jersey	500
	Bike helmet	100
50 or fewer	Bike seat	50
	Bike	10
	Biking shoes	20

By creating a sum aggregation on the quantity field, you can display the aggregation in your grouped hit:


Grouped hit rank	Name	Quantity	Total quantity for group
1	Bike seat	50	80
2	Biking jersey	500	600