Using Attribute Affinity
Attribute Affinity is a data analysis tool that enables you to find other audiences that are very similar to your audiences.
Attribute Affinity lets you automatically find attributes for a set of individuals that statistically or significantly distinguish that set of individuals against a baseline of arbitrary individuals. This allows you to understand the composition of attributes within a seed dataset compared to a baseline dataset.
The benefits of Attribute Affinity are twofold:
- To identify characteristics that define your seed audience to better understand what makes them relevant to your brand.
- To use these characteristics to find new prospects within a publisher’s dataset to target programmatically and develop new customers.
You can export Attribute Affinity data from the Platform to a CSV file, which can be used in third-party software to create data visualizations of attribute correlations or create segments for audience expansion. Attribute Affinity uses privacy-safe machine learning on distributed data to produce highly accurate results, with no intervention required from the user once results processing starts.
Attribute Affinity task results are returned as an ordered list of attributes and attribute combinations. A score is calculated for each single attribute in isolation. The top 30% of these scores are then combined to show results for pairs of attributes.
Before you begin
Before running Attribute Affinity, you will need to do the following:
- Use Audience Builder to create your seed and/or baseline audiences.
- Use Segment Builder to create your seed and/or baseline segments.
You can connect to second and third-party datasets to enrich the available attributes or increase the match rate.
Important: For Attribute Affinity to work, the seed and baseline must contain common categories (up to a maximum of 50 categories). For this reason, we recommend using a subset of the baseline as your seed.
Running an Attribute Affinity task
Use the steps below to run an Attribute Affinity task to produce a list of attributes ranked by score.
On InfoSum Platform, go to the Attribute Affinity tab under Analyze.
Click New Task and enter your details.
The example screenshots in this article use the following seed and baseline:
- ACME existing customer seed - A seed segment that consists of:
- Audience: the intersection of the ACME and KnowArc tutorial datasets.
- Segment: the audience filter existing customer = Yes.
- KnowArc Baseline - A baseline audience containing only the KnowArc dataset.
You will need to:
- Add a name to identify the Attribute Affinity task.
- Optionally, add a description of the task.
- Specify whether your seed is a Platform audience or a Platform segment. You can compare a segment with an audience.
- Select your seed audience or segment from the drop-down list. Only audiences or segments with status ‘Ready’ will be shown in the list. Ideally the seed should be a subset of the baseline. For details of how to create an audience or segment, see using Audience Builder and using Segment Builder.
- Specify whether your baseline is a Platform audience or a Platform segment. You can compare a segment with an audience.
- Select your baseline audience or segment from the drop-down list. Only audiences or segments with the status ‘Ready’ will be shown in the list.
Click Next where you will be taken to the form below.
This form shows all the categories common to the selected baseline and seed.
Use this form to select the categories that will be used to distinguish the seed from the baseline. You must select at least one category. On this form, you can:
- Select/deselect a category using the tick box to the left of the category, for example, “Gender”.
- Select a representation within the category from the drop-down list. For example, “Binary and Other” or “Extended Gender” within the category “Gender”. You can only select one representation per category - this is to restrict the number of queries run by the Platform.
- Use the Filter categories field to find any categories in the Categories list.
If your seed is a segment, we strongly recommend that you deselect any common categories that have been added as audience filters in Segment Builder, otherwise the highest results will be populated with combinations of these categories.
Click Create Task to run the Attribute Affinity task. This page shows details for any running or completed tasks. Your newly-created task appears at the top of the list.
The Status field shows the status of the task. Click Refresh to update the status and progress of your Attribute Affinity task.
The Progress field shows the percentage of single attributes and pairs of attributes processed by the Platform. The Platform processes single attributes first and then uses these results to process pairs of attributes. The Platform executes four queries for each attribute in a task.
Use the Action menu to pause or cancel running tasks and resume paused tasks (see below for more details). Completed and paused tasks expire after seven days, at which time they will be removed from the list. Resuming a paused task resets its expiry date.
Click on a task to view the task’s description, together with its seed and baseline details.
Select View Results from the Action menu to view results for a task. You can view and download partial results for a running task. You can download the first 10,000 results using the selected filter and sort criteria to a CSV file. You can also view and download results for previous tasks.
By default, task results are sorted by score and ordered from highest to lowest. The following information is shown:
- FIRST ATTRIBUTE shows the category, representation, and value for a single attribute.
- SECOND ATTRIBUTE shows the category, representation, and value for the second attribute in a pair of attributes. If this field is empty, the score is for a single attribute only. If this field shows a second attribute, the score shows the combined score of both attributes.
- SCORE shows the internal score that InfoSum uses to rank single attributes or pairs of attributes. Scores are used to sort the attributes relative to each other in their strength of differentiation to the baseline. The higher an attribute's score, the more the attribute in the seed differs from the baseline. The score indicates the likelihood an attribute can identify your seed audience. This allows you to reach audiences that are not yours, but are very similar to your audience. You can use the highest scoring attributes to create new segments or insights in the Platform. If you need further information, please contact support@infosum.com.
- BASELINE REACH bar shows how many records in the baseline have the attribute or combination of attributes.
- SEED REACH bar shows how many records in the seed have the attribute or combination of attributes.
You can use the Filter By pane on the left to filter attributes in the Task Results page. For example:
Use the tick boxes to filter task results by singles and/or pairs of attributes, and use the drop-down lists to filter by attribute details. A filter box can have more than one option selected (as shown).
Use the sort arrows in any of the column headings to sort the results in ascending or descending order.
If you sort by first or second attribute, attributes are sorted in the following order:
- By category (ascending/descending)
- By representation (ascending/descending)
- By value (ascending/descending)
- By score (descending only)
Downloading Attribute Affinity Task Results to a CSV file
You can download up to 10,000 rows of an Attribute Affinity Task Results to a CSV file.
- On InfoSum Platform, select the Attribute Affinity tab under Analyze.
- Drop down the Action menu and select View Results for the task you want to download.
-
On the Task Results screen, click on Download CSV.
-
Click Download CSV to save your task results using the selected filter and sort criteria locally to a CSV file.
The CSV file gives a detailed view of the attributes and scores, which can be used for example to create data visualizations of attribute correlations or audience expansion in Google Data Studio or Excel.
Note: When you download task results, the reach bars for an attribute appear as “value count” in the CSV file. For example, “SEED REACH” appears as “Seed value count” in the CSV file. The bars are called reach on the Platform because they show the value count taken out of the category count. The reach bars are shown out of the maximum category counts to ensure the bars are relative and comparable.
Pausing an Attribute Affinity task
You can pause an Attribute Affinity task that is running. Pausing a running task allows other tasks to run faster.
- On InfoSum Platform, select the Attribute Affinity tab under Analyze.
-
Drop down the Action menu and select Pause for the selected task.
The status of the task changes to “Paused”.
Resuming an Attribute Affinity task
You can resume a paused Attribute Affinity task.
- On InfoSum Platform, select the Attribute Affinity tab under Analyze.
-
Drop down the ACTION menu and select Resume for the selected task.
The status of the Attribute Affinity task updates to show the task’s progress and the task’s expiry date is reset.
Canceling an Attribute Affinity task
You can end a running or paused Attribute Affinity task and view the partial results.
- On InfoSum Platform, select the Attribute Affinity tab under Analyze.
-
Drop down the Action menu and select Cancel for the selected task.
-
Click Cancel Task to cancel the selected task.
The status of the task changes to “Canceled”.
Select View Results from the Action menu to view or download partial results for a canceled task.
Deleting an Attribute Affinity task
You can delete a canceled or completed Attribute Affinity task from the Attribute Affinity page.
- On InfoSum Platform, select the Attribute Affinity tab under Analyze.
-
Drop down the Action menu and select Delete for the selected task.
Click Delete to delete the selected task.
The task no longer appears on the Attribute Affinity page.