Why is this work relevant for Innovation?

Lumin(ai)re is an SUTD & NTU start-up founded in Singapore, funded by Enterprise Singapore under the StartupSG grant, with co-investment by Spaze Ventures as Accredited Mentor Partner (AMP). Since November 2018, we have partnered with A*STAR I2R (Agency for Science, Technology & Research) for R&D relating to GAN potential in next-generation vision computing. Using a multivariate ad testing model to determine creative tastes based on regional, cultural and product category variations, we are building an advertising ecosystem of the future, where each piece of advertising content is generated by AI in real-time.

Background

With abysmal engagement rates in digital online ads across categories, it is ever more important that engagement rates be boosted for both agencies and clients. Ad spend ROI is adversely affected when ads don't perform well with low consumer engagement, making agencies look bad and clients walking away unsatisfied. Through a new approach called "Data-driven Design", ads can be generated in real-time by an AI system. An AI system is able to design and publish ads in real-time, creating 1000s of variations that are tested in the real world. Using the power of visuals and text, the art direction and copywriting processes presently carried out manually by ad creatives can be replicated by an AI system that can generate 1000s of data-driven ads 24/7, 7 days a week - with a focus on increasing engagement rates through combining the most optimal creative attributes.

Describe the idea

As digital integration takes over most urban city centres, through highly connected mobile connectivity and digital out-of-home assets, there is opportunity to adapt AI technology in serving customised ad designs to consumers. Focusing on trackable metrics on online ads, over a period of time, AI systems can absorb big data and use it to make meaningful decisions on design attributes to achieve optimal engagement rates for audiences. This automation assist SMBs who are seeking to achieve low cost on high churn ad work output, with adaptiveness that ensures consumer engagement does not dip. Conversely, it also helps publishers to maintain a standard quality of engagement when selling ad spaces to clients.

What were the key dates in the development process?

14 August 2018 - company incorporated, raised pre-seed fund from EnterpriseSG and Spaze Ventures November 2018 - partnered with A*STAR Singapore under the T-UP scheme Feb 2019 - launched proof-of-concept (POC) May 2019 - awarded Echelon Asia Summit 2019 TOP100 APAC Startup July 2019 - achieved 5 paying beta customers 1 August 2019 - Lumin(ai)re wins the Vikingcubator 2019 pitch competition in Singapore 8 August 2019 - begun media partnership with the NHST Media Group Asia 15 August 2019 - Lumin(ai)re is awarded Brands for Good 2019 honouree under 2 categories 16 August 2019 -

Describe the innovation/technology

Our current prototype generates creative online banner ads in response to ad performance data, specifically click-through rate % scores. The automated platform can generate around 15 online banner ad variations within 3.5 seconds. The next phase of platform development would involve connecting with publishers, enabling advertisers to automatically generate and publish live ads within a single platform. Once ads are published, AI deep learning algorithms capture performance data and correlate patterns between ad design and market performance. Over a period of time, the platform dynamically adjusts ad design automatically to consistently maintain engagement rates > 1.0%.

Describe the expectations/outcome

Implementing the data-driven design approach, over a longer time frame and larger dataset, would provide results that are even more accurate in prediction of consumer taste. With advertisers concerned about return-on-ad-spend (ROAS), data-driven design as a new approach to ad design can provide them with better assurance on performance. Over larger scales of datasets, current advertising agencies would benefit from implementing technology stacks not unlike technology companies. Tech stacks would include integration of technology such as data classification algorithms, AI predictive analytics and automated processes for repetitive high churn design work such as online banner ads. Preliminary testing has shown positive result, with generated ads outperforming industry averages. Thus far, Lumin(ai)re has investments from Spaze Ventures and Enterprise Singapore, as well as a partnership with the Agency of Science, Technology and Research (A*STAR). We are presently in process of raising our seed round.