AI-Powered Recommendation
OVERVIEW
Campaign optimisation is one of the most frequent tasks handled by programmatic specialists on a daily basis.
A key challenge many programmatic specialists encounter is the labor-intensive manual workflow, which often leads to campaigns being under-optimised. This situation poses significant risks and potential setbacks for Adform as a media advertising platform, especially when competing with larger players like Google DV360 and The Trade Desk.
To address this issue, Adform has introduced a new strategy to develop an AI-powered recommendation engine designed to automatically optimise campaigns toward desired goals with greater efficiency.
A key challenge many programmatic specialists encounter is the labor-intensive manual workflow, which often leads to campaigns being under-optimised. This situation poses significant risks and potential setbacks for Adform as a media advertising platform, especially when competing with larger players like Google DV360 and The Trade Desk.
To address this issue, Adform has introduced a new strategy to develop an AI-powered recommendation engine designed to automatically optimise campaigns toward desired goals with greater efficiency.
TASK
1. Design a workflow that generates actionable recommendations for campaign optimisation, enabling programmatic specialists to achieve their objectives
2. Perform discovery research to analyse and understand user behaviour patterns during the process of optimising digital campaigns
CLIENT
Adform, Lithuania
ROLE
UX Designer, Researcher
TIME
1 Year
(2024-2025)
RESEARCH
How does a Programatic Specialist accomplish their goals?
A Programmatic Specialist is a highly skilled professional, well-versed in navigating and leveraging the latest technological advancements within the digital marketing industry.
They create, optimise, and analyse digital campaigns tailored to their client’s needs to achieve various advertising goals, such as raising brand awareness among new customers, reaching more target audiences, and promoting seasonal advertisements.
A Programmatic Specialist is a highly skilled professional, well-versed in navigating and leveraging the latest technological advancements within the digital marketing industry.
They create, optimise, and analyse digital campaigns tailored to their client’s needs to achieve various advertising goals, such as raising brand awareness among new customers, reaching more target audiences, and promoting seasonal advertisements.
By conducting discovery research studies with more than 10 programatic specialist around the world, here is a typical routine of their experience on a daily basis:
Synthesising the data
Conducting discovery research enables me to gain a comprehensive understanding of a specialist’s daily operational workflows while providing valuable insights into their specific needs and preferences.
Often times, campaigns are optimised across various dimensions. The table below outlines user goals and corresponding actions within each dimension.
Often times, campaigns are optimised across various dimensions. The table below outlines user goals and corresponding actions within each dimension.
IDEATION
A Programmatic Specialist's workflow in Adform
Below shows the comparison between the old and the newly proposed workflow:
Below shows the comparison between the old and the newly proposed workflow:
With the old workflow, optimising a campaign in Adform required a significant amount of user effort. Many clients, particularly newcomers to Adform struggled to navigate the complexities of such an extensive process.
In contrast, the new workflow aims to streamline the process by delegating tedious and time-consuming manual tasks to the system. Leveraging the intelligent AI forecasting engine that predicts recommendations for optimal results, the workflow can be shortened significantly.
For programmatic specialists, the task is straightforward. They only need to review the insights and apply it. This approach not only simplifies the workflow but also empowers users to focus on strategic execution rather than manual effort.
In contrast, the new workflow aims to streamline the process by delegating tedious and time-consuming manual tasks to the system. Leveraging the intelligent AI forecasting engine that predicts recommendations for optimal results, the workflow can be shortened significantly.
For programmatic specialists, the task is straightforward. They only need to review the insights and apply it. This approach not only simplifies the workflow but also empowers users to focus on strategic execution rather than manual effort.
Time and Result Sensitivity Approach
Determining when a recommendation should be suggested is a crucial criterion. Through research, we discovered that poorly performing campaigns are occasionally left unoptimised in the beginning. This is because an experienced programmatic specialists know the optimal timing for adjustments.
For instance, a newly launched campaign needs more time to achieve its intended results; therefore, optimising it too early could hinder its potential.
The table below demonstrates the level of urgency to optimise a campaign with the aspect of time versus result:
As shown, the farther away the end date, the less urgency there is to optimise; as the end date approaches, the urgency to optimise increases.
Meanwhile, the better the performance, the less need there is to optimise; the worse the performance, the greater the need to optimise.
Fixing / Enhancing Strategy
Another essential criterion for determining the effectiveness of a campaign optimisation is the severity of the recommended adjustments.
Specialists observing lower performance will implement more significant changes, while those noticing higher performance will only require subtle changes.
Specialists observing lower performance will implement more significant changes, while those noticing higher performance will only require subtle changes.
For instance, a recommendation strategy advising specialists to remove more than 30% of domains is considered a drastic measure and should only be recommended for campaigns performing below average.
Meanwhile, advising specialists to increase bid multipliers on well-performing time slots is a subtle change that can enhance overall results. Therefore, it can be applied to campaigns that are already performing well.
If a campaign has already exceeded its target, then no further recommendations should be offered. While optimisation can often lead to improvement in performance, there are times when it’s better to leave things unchanged, as those adjustments could have unintended negative consequences to the campaigns.
Meanwhile, advising specialists to increase bid multipliers on well-performing time slots is a subtle change that can enhance overall results. Therefore, it can be applied to campaigns that are already performing well.
If a campaign has already exceeded its target, then no further recommendations should be offered. While optimisation can often lead to improvement in performance, there are times when it’s better to leave things unchanged, as those adjustments could have unintended negative consequences to the campaigns.
AI Prediction Tool
The core of this project relies on the sophisticated algorithms and machine learning methods to predict future trends and outcomes using historical data.
This data-driven approach empowers decision-makers with actionable insights, enabling them to anticipate market shifts, optimise strategies, and allocate resources more effectively.
Recommending on strategic pages
The platform's existing features create opportunities to introduce the functionality. Incorporating multiple entry points is a strategic approach that comes with benefits such as:
Seamless User Experience
The integration of AI recommendations feels more natural and unobtrusive, increasing the likelihood of adoption and encourages users to explore the benefits of AI-driven insights
Contextual Relevance
By embedding recommendations within relevant features, the suggestions are more likely to align with the user's current tasks
Scalable Personalisation
As users interact with the recommendations across different entry points, the AI system gathers more data, enabling it to refine and improve its suggestions over time. This creates a feedback loop that enhances the accuracy and relevance of future recommendations.
VISUAL DESIGN
Designing an AI-powered recommendation
Note: The design has been modified from the original version to maintain confidentiality.
Task-integrated Recommendation
Recommendations are woven into a programmatic specialist’s daily tasks, from crafting a media planner to optimising a campaign.
Recommendations are woven into a programmatic specialist’s daily tasks, from crafting a media planner to optimising a campaign.
This integration ensures that AI-driven recommendations are connected to every stage of the workflow to enhance efficiency without disrupting familiar processes.
Seamless Recommendation Access
A collapsible drawer allows users to access the recommendation feature from any page they are on and navigates to the next card easily.
A collapsible drawer allows users to access the recommendation feature from any page they are on and navigates to the next card easily.
It strikes a perfect balance by ensuring the recommendations are readily available without interrupting other workflow. This keeps the feature within easy reach while maintaining a clean and unobtrusive user experience.
