Microsoft Invest - Dimensions, Metrics, Filtering, and Grouping

This page outlines how to design your reports so that you can:

  • Reduce the size of the data set by removing data you don't care about.
  • Increase the relevance of the data you do get.
  • Get your results more quickly; requesting smaller data sets often increases reporting performance.

Every report returns a collection of data. In this document, this collection of data is referred to as a data set; these data sets are described in terms of dimensions and metrics.

Similarly, when you are configuring a report, you can choose to filter the data set.

The key recommendations of this document are as follows:

  • Filter out as much data as possible - smaller data is faster data.
  • Choose as few dimensions as possible - the report automatically groups by all selected dimensions. The more dimensions you group by, the larger the data set becomes.
Term Definition
Dimensions Dimensions are things the data set describes, such as advertisers, publishers, line items, or domains. In other words, dimensions are what the data set is about.
Metrics Metrics are numbers that measure the performance of the dimensions you've selected. Examples include:
- How many impressions came through this publisher?
- What is the click-through rate across this campaign's creatives?
Filtering Filtering removes dimensions you don't care about. This reduces the size of the data set. For example, you might want to see only those impressions that occurred on weekends in a certain region of France.

Tip: Filter out as much irrelevant data as possible – smaller data is faster data.
Grouping Reports automatically group by all selected dimensions. Grouping causes the dimensions to be output in a specific order. This can be useful for understanding relationships between dimensions. For example, you can see how the different line items under an advertiser are performing.
Since the report automatically applies grouping to dimensions, selecting many dimensions can generate a very large data set. For every additional dimension you select, the data set becomes larger.

Tip: Select as few dimensions as possible. The more dimensions you select, the larger the data set becomes.

Visualizing the process

The diagram below gives a simplified view of the process that generates the data set that you receive.

In Step 1, unnecessary data is removed. This is important to keep your reports small and relevant. We recommend filtering out as much information as possible.

In Step 2, the data set is automatically grouped by the selected dimensions. This step can generate a large data set if you selected many dimensions. We recommend selecting as few dimensions as possible.

In Step 3, our systems generate the report. You have the option to view the report, have it run in the background or have it exported and emailed to you (in CSV or Excel format). This goes more quickly with smaller data sets.

Diagram of the filtering and grouping process.

A tiny example database

All of the examples below will use the same underlying "database", which contains two dimensions: Advertiser and Line Item; and two metrics: Impressions and Clicks. Here it is in its entirety:

ID Advertiser Line Item Impressions Clicks
1 A ACME Axe Co. 1792 21
2 A ACME Axe Co. 4355 34
3 A Bar-None Ice Cream Cones 78,231 1408
4 B Candles R-US 2843 65
5 B Doggie Accessories 9486 123
6 C Evergreen Synthetic Xmas Trees 2238 46
7 D Frog Hollow Athletic Socks 8198 214
8 D Frog Hollow Athletic Socks 103 12
9 E Best Soap & Candle 3883 41
10 E Best Soap & Candle 1292 183
11 E Best Soap & Candle 902 81
12 E Best Soap & Candle 5352 212
13 F Farmer Snowplows 12,448 256
14 F Farmer Snowplows 23,984 782
15 F Farmer Snowplows 8,764 128
16 G Gretchen's Organic Chocolates 48,996 973
17 H Hap's Go-Kart Track and Petting Zoo 1108 87
18 H Hap's Go-Kart Track and Petting Zoo 1872 116

You can also download the example in CSV format.

Filtering examples

The examples in this section illustrate how data set sizes are affected by filtering. Filtering is useful for several reasons:

  • Filtering the data set reduces its size. A smaller data set is easier for a computer to process more quickly.
  • Filtering the data set reduces the amount of irrelevant information humans have to deal with.

Example 1: Filter by a minimum of 9,000 impressions

In this example, we use a filter to see which of our line items has at least the minimum specified amount of reach. This reduces the size of the data set from 18 rows to 5.

ID Advertiser Line Item Impressions Clicks
3 A Bar-None Ice Cream Cones 78231 1408
5 B Doggie Accessories 9486 123
13 F Farmer Snowplows 12448 256
14 F Farmer Snowplows 23984 782
16 G Gretchen's Organic Chocolates 48996 973

Example 2: Filter by two line items

In this example, we check the performance of two of our line items. This reduces our data set to 6 rows out of a possible 18.

ID Advertiser Line Item Impressions Clicks
1 A ACME Axe Co. 1792 21
2 A ACME Axe Co. 4355 34
9 E Best Soap & Candle 3883 41
10 E Best Soap & Candle 1292 183
11 E Best Soap & Candle 902 81
12 E Best Soap & Candle 5352 212

Example 3: Filter by advertisers B and D

In this example, we filter out all but two advertisers. This reduces our data set size from 18 rows to 4 rows.

ID Advertiser Line Item Impressions Clicks
4 B Candles R-US 2843 65
5 B Doggie Accessories 9486 123
7 D Frog Hollow Athletic Socks 8198 214
8 D Frog Hollow Athletic Socks 103 12

Grouping and dimension selection guidance

Grouping can increase the size of the data set by adding rows (possibly a very large number of rows). Since the Microsoft Advertising reporting system automatically groups by all the dimensions you select, a best practice is to select as few dimensions as possible to get just the information you need. You can further reduce the effects of this automatic grouping of dimensions by applying filtering.