Use Cases and Features

What are the primary use cases for the Audience Builder?

• Interest-based audience segments for targeting
• Custom audience creation for targeting
• Look-alike modeling based on past and current customers
• Look-alike modeling based on prior or current campaign response 
• Real-time audience estimation
• Turnkey audience deployment

What features are included in the Audience Builder?

• Audience management tools
• Ability to upload a list of addresses or lat/longs using the store management system in order to create a targetable geoset. 
• Ability to define frequency of visit
• Ability to define multiple parameters for audiences including geosets, days of week, interests, lookback window
• Real-time audience estimation 
• Look-alike Modeling
• Custom audience refresh rate
• Turnkey activation

Feature Questions

How often is day of the week calculated?

Weekly on Sunday. 

Cuebiq receives data from our partners in real-time, however audiences are calculated and uploaded into the audience builder weekly to allow for real-time audience sizing estimates. 

If you are interested in more recent data (1-3 days), please contact your Cuebiq account representative and a custom workflow can be established to deliver you data with an SLA of 24 hours or greater.

What if I want an audience with a look-back longer than 120 days?

Cuebiq can create audiences with a look-back longer than 120 days but must do so using a custom process that adheres to our privacy policy. In order for Cuebiq to build an audience of users who have taken an action more than 120 days ago, we must see that user in our system sometime in the last 120 days. This ensures we are only delivering active IDs for targeting and that no users who have opted-out are included in an audience. 

Please contact your Cuebiq account representative to create an audience looking back up to 2 year ago. Please note, scale becomes limited the farther back an audience is requested. 

How often is “specific date in the past” calculated? 

Weekly on Sunday. 

If you are interested in more recent data (1-3 days), please contact your Cuebiq account representative and a custom workflow can be established to deliver you data with an SLA of 24 hours or greater.

What data is extracted when I select “1 Week” as my look-back window”?

The data extracted will be from Sunday to the previous 7 days. 

For example, if I select a 1 week look-back on Tuesday June 30th, I will receive data from 6/28-6/22.

How does look-back window relate to refresh cadence?

As a best practice, it is recommended to have the look-back window and refresh cadence be the same. 

Example: 

Today is Tuesday June 30th and I want to reach users who visited a KFC in the last week.

I select the KFC geoset and then set the look-back window for 1 week and the refresh to 1 month. This means the audience builder will deliver back users who visited a KFC in the last week starting on Sunday 6/28 (6/22-6/28). If I set my refresh cadence to 1 month, that means that on 8/28, this audience will re-run and pull users seen in the last week (7/22-7/28).

Therefore, matching the look-back and refresh cadence will give you continuous coverage throughout the month.

What do the different frequency bands mean?

We have grouped visitation into quantiles for each geoset, meaning “Low” frequency is the bottom 33% of visitors by frequency, “Medium” is the middle 33% and “High” is the top 33%. This allows you to capture loyalty and frequency of visit while still maintaining real-time size-estimation capabilities. 

Monthly average visitation for each Frequency level

How long does it take an audience extraction to run and when should a client expect an audience to be delivered to their seat

Once a client hits “Save and Extract” and fill in their DSP and Seat information, two workflows are kicked off:

1. The audience extraction process begins and typically takes 3 hours. Once the extraction is complete, it is automatically sent to LiveRamp to begin the approval process.
2. Simultaneously, a Jira ticket is created for someone on the ops team. Once the audience has been approved and uploaded into LiveRamp, this person then manually pushes the audience to its final destination in the desired DSP seat. 

The reason this is in two steps is that LiveRamp has not automated the second workflow, forcing us to complete this task manually. 

What is the maximum number of geosets per expression and why?

The maximum number of geosets in a given expression is 6. This is a technical limitation, not a design choice. With more than 6 geosets, the tool is unable to process a real-time audience size at a low cost. 

Store Management System Questions

Does the audience builder use dwell time or radius?

No, since the Audience Builder is reading off of Device Location tables (as do standard audiences), dwell time and radius cannot be read by the tool. The reason is that Device Location does not contain dwell times and radius is a feature that was intentionally left out due to computational restrictions. 

How long does it take for a custom geosets to be uploaded into Audience Builder?

Once a custom set of lat/longs or addresses are uploaded into the Store Management System, it takes 24 hours for that geoset to populate in the Audience Builder. Once it does populate, you can search for it by name in the geosets area. 

How long does it take for a standard geoset to be uploaded into Audience Builder?

Once a set of lat/longs or addresses are uploaded into Krakatoa, it takes 24 hours for that geoset to populate in the Audience Builder. Once it does populate, you can search for it by name in the geosets area. 

Estimation & Modeling

Modeling methodology

1. We take a seed audience (which is a list of device IDs) and look at 2 weeks of visitation from the devices seen in that seed to develop a profile.
2. We then compare that profile of visitation against our loyalty index which contains 2+ years of visitation information on over 600 types of US locations (ex. Brands, Entertainment/recreation locations, etc.)
3. From there we run an algorithm that scores and ranks the 20MM IDs that we have in our database to understand which devices have visitation patterns most similar to the seed
4. From there we cut that model based on how much extension is desired by the client and deliver the modeled IDs along with the seed (or excluding the seed, based on client requirements)

What happens when an audience is less than 10k IDs? Can the audience still be extracted?

The privacy team has mandated that 10k is our bottom limit on audience size for privacy reasons. However, this is also the point where the error rate on estimations becomes too high.

Audience Delivery

How will I know my audience has been received and is being worked on? How will I know it has been delivered?

The moment you complete the audience build, the segment creation is kicked off and a member of the Cuebiq team is notified of it's creation. The Cuebiq representative will then facilitate and confirm the segment has been successfully delivered. There are three statuses within the builder which denote different statuses of the segment creation process:

• Idle: the audience is pending in queue to begin processing
• Running: the segment is in the midst of being created.
• Success: the segment has been successfully created and a Cuebiq representative is working through delivery completion.

You'll be notified by a Cuebiq representative via email with confirmation that the segment has successfully delivered.

What leads an audience to fail?

An audience fails when the underlying AWS job fails. We are currently working on increasing our capacity on AWS, which should lead to less failures.