Facebook | Ads Targeting Signals + Constraints Vision
Designed with Project Management, UX Research and Engineering Functional Partners
Key takeaways
Project Breakdown
Project Type: Strategic Blueprint and Visioning
Contributions + Role: Research Planning, Research Synthesis, Roadmapping, Framework Development and Concept Discovery
Principles
Product experience should match advertiser objectives and knowledge
All inputs and options must provide value
Our inputs should not prevent our system from delivering more value to our users
Summary of impact
Discovery and explorations led to identifying focus gaps in team structure. Namely, developing charters that focus on first-party signals (e.g. data that Facebook has, etc) and third-party signals (e.g. the ingestion of what information business have and advertisers understand, etc).
Helped form a strategic focus and rallying cry for the team in the subsequent years; “Remove unnecessary constraints preventing our system from delivering value, and collect as much relevant signals as possible to help guide our system.”
Context
Audience Manager is a repository of huge value as our largest advertisers and users of Custom Audiences and Lookalikes have thousands, sometimes tens of thousands, of audiences. However, the product does not offer a scalable method of organizing, understanding use, or management. There are audiences which are at any given moment powering millions of dollar in revenue.
Currently inefficiencies that are inherent in Audience Manager mean that clients are wasting time and resources on repetitive tasks like recreating and finding the audiences they want to use. This represents both a productivity loss, but also a massive opportunity cost as these are people whose time would be better spent working out how to increase the performance of their campaigns through more effective targeting.
Problems
product experience invites the wrong inputs
The language we allow advertisers to use in our system doesn’t set them up for success largely due to the fact the current interface invites advertisers to make unnecessary inputs that reducing optionality
product doesn’t capture nuance
The current system captures primitives, not semantics, and this negatively impacts optional value (e.g. the ability for the system, ML, to explore, etc)
Current State juxtaposed to conceptual exploration
Exploration state variations highlighting varying data source types; (left) catalog (e.g. third-party structured data, etc), (right) Facebook first-party and pixel tracking online signals
Goals
Provide businesses with the best way to reach the right customer audience
Reposition Targeting inputs to be better informed by advertiser knowledge
Customer Insights
Marketing Goal
Business KPI
Business Knowledge
Vertical
True Constraints
Make distinction between advertiser intent and true constraints
Measuring success
To make impact we will need to:
Reduce the use of unnecessary constraints in our system achieving campaign goals
Improve performance for lower funnel ad campaigns
Increase the amount advertiser spend on the ad platform — this signals adoption
Process + Execution
First, lets establish a framework to build upon
Understand and synthesize data structure to high-level advertiser use cases
Develop model of how advertisers interpret targeting inputs
Synthesize advertiser feedback and observations to inform framework for collecting advertiser knowledge and reconciling business constraints
Hypothesis
More opinionated in our products and guide advertisers to towards the most optionality for targeting bringing advertisers more value
Making a cut on “True business constraints” compared to expressions made in campaign creation we will reduce the cases where advertisers are over-constraining
Giving advertiser what information we are using or what is at the disposal of our platform we will increase advertiser trust, increasing liquidity
Providing advertisers a way to provide information that is predictive of their desired outcome, we will support better advertisers behaviors on our platform
Initial hypothesis on structuring advertiser inputs, and business knowledge
Evolved thinking on data structure and true constraints after a series of contextual inquiry sessions with advertisers
Research Sessions in Toronto, ON, Canada
with new understandings from advertisers, lets Explores objective-informed setup
Advertisers wanted us to present the right controls or setup informed by their stated objectives. Now that we have signal on how advertisers plan their campaigns, decide what to put into our system, and measure success on or off platform; How might we build the best UI to adapt to stated objectives? Key topics addressed as informed by advertisers;
[Advertiser Intent] How advertisers think about reaching new customers and bringing them back
[Advertiser Intent] How advertisers think about their funnel
[Advertiser Knowledge] What expectations advertisers have if our system asked about their business or industry vertical
[Decision Making] How advertisers allocate budget, and communicate across teams or managers
[In-Product Feedback] What advertisers need to know in context to adopting an automated task that was formerly manual
High-level design plan abstracting current state, end state, and design plan to achieve success.
Setting expectations and communicating plan for component-level explorations
Informed by advertiser feedback and newly formed strategy of “More signal, and less constraints” in context to advertiser inputs, the following step was to seek out near-term opportunities to reach the desired end state, and a plan to support rationales.
Updated Principles
Product experience should match advertiser objectives and knowledge
All inputs and options must provide value or clearly frame trade-offs
Our inputs should not prevent our system from delivering more value to our users
Product experience should be opinionated and make recommendations when possible
Product experience should display what the system understands about the advertiser and their customers to build trust in automation
Seed selection
Addressing how advertisers understand and select between their structured leveraging first-party data, and leveraging what our system already knows about customers when creating a campaign
Exclusions
Addressing how advertisers inform the system who should no longer receive ads related to a specific campaign.
Automation + feedback exploration
Addressing how advertisers inform the system who should no longer receive ads related to a specific campaign.
Original State of Audience Input
Audience Input UI after implementing new principles
Detail view of Age & Gender Concept variations. Intent is to share what the system can see while content is opinionated in what the system understands is the optimal selection for the campaign.
Results + Imact
Product impact
Opportunities identified in design concept studies carved out understand and execution work focused on capturing advertiser customer knowledge in a way that informed our system without constraining.
organizational/team impact
Results from the effort helped solidify engagement models between product and platform ranking teams to forge new work streams focused on capturing more nuanced information from advertisers that improved advertiser campaign performance
Results from study acted as in input for a partner team research effort focused on understanding advertiser intent