The paper examines the various challenges that are encountered by e-commerce retailers in measuring the effectiveness of Media. The retail industry has adopted the Adstock Model to estimate Digital Media ROI. Adstock Model is an adopted model from TV. The new rich media and multiple media asset types (Image / Audio / Text / Video etc) and multiple channels (Facebook / Twitter / Instagram / Blogs etc) and the platforms (PC / Laptop / Mobile / I-Pad’s / Tab) that are entering into the market makes it complex to assess media effectiveness. Automation of media effectiveness is even more complex. As the size and variety of media data has increased without the application of AI and ML it is impossible to accurately estimate the effectiveness of Media. CatmanAi solution uses innovative application of AI & ML to precisely assess the media impact and effectiveness in an automated decisioning solution. A large CPG/F&B market leader with UK operations adopted CatmanAi an AI-ML solution to validate if their current retailer media spends are at sub-optimal level, optimal level or at saturation level for effective media planning, in addition CatmanAi created saturation curves for each media channel. The findings of the study is of great importance to current e-commerce retailers. However more research is required specially on social media, this is because social media ROI is not just dependent on Investments in social media it also depends on reviews on social media outside of e-commerce channel.
E-commerce Channel is a cluttered space. Media plays a strategic role in driving e-commerce sales. However, it is a challenge to estimate the short term and long-term impact of digital media on e-commerce sales. Most of the CPG Marketers rely on third party tools and media agencies to accurately measure the effectiveness of Media. According to Boston Consulting Group, while few companies have adopted media measurement system; only a fewer are able to measure and optimise. A lack of homogenous data can also stand in a way of achieving the Goal. One of the largest Food Brand of a CPG/F&B market leaders with UK operations was facing the challenge in measuring the impact of Digital Media (SEO, Ads and A+ Content) on sales and market share. Client wanted to quantify the impact of media on market share and sales, create saturation curves and find out if their current retailer media spends are at sub-optimal / optimal or saturation level which could eventually help them prioritise their marketing initiatives.
As per Nielsen 2023 annual marketing report media clutter and channel overload is hurting accurate estimation of Media ROI. 70% of the Global Marketers claim that it is easy to estimate the aggregate (Total Media ROI) but on the flipside the confidence on the ROI Measurement at the individual media channel level is much lower.
2023 Nielsen Annual Marketing Report
The Corona pandemic has increased the e-commerce sales to new heights, as people embraced social distancing they turned into online-shopping, what is expected going forward is that a choice between e-commerce and physical stores won’t be the answer – instead offering a compelling omni-channel experience will be a requirement for the survival for every retailer. This has reinforced the need for stronger ROMI Initiatives (ROMI = Return on Marketing Investments). It is important to measure both short and long term impact of Digital Campaigns to generate long lasting brand equity. What is even more important is automating the entire ROMI Initiative.
Globally only a few companies have a standard practice to Measure ; only a fewer were able to both measure and optimise. The key obstacles to adopting up-to-date media measurement processes are data integrity (cited by 89% of marketers, according to IPSOS MMA Global) and organization buy-in (cited by 69% of marketers). Navigating uncertainty requires having agile processes, access to real-time data, and the latest tech capabilities which continues to be a challenge in ROMI Initiatives.
CatmanAi has implemented a Light Gradient Machine Learning algorithm to study the relationship between media and sales. The power of CatmanAi is that it can help both measure and optimise the media investments. Alongside it can also automate the entire ROMI initiative.
CatmanAi can run models, generate response curves for each media channel, measure ROI and optimise ROI by each media channel separately. It can also provide business insights about Media Assets. Ex. What worked better Image or Video? Light GBM is a gradient-boosting framework which has the potential to increase the efficiency in running the models with a reduced memory usage. The algorithm is flexible to incorporate as many as media assets / media channels and platforms without increasing the processing cost required to train the model.
The Model incorporates the TV Adopted learning – Adstock. The Adstock model hinges from the theory that advertising awareness is built over a period of time and the customers will retain the advertisement in their mindset for some time which is called as Carryover effect in statistical terms. When the model accounts for carryover effect it is possible to estimate the long-time ROI of any Media Channel. Media Saturation curves and diminishing return point was estimated for each media.
The Easy-to-use Simulation tool helped client in moving from "measure" to "measure, optimise and chose future".
Key Model Recommendations
Media Optimization for Asda Retailer has resulted in an incremental uplift in ROI by 11.6% from a base ROI of 3.02 to optimised ROI of 3.37 with an increase in spends by 20%
There is a possibility to Increase Asda spends up to 50% as the current spends is at a below optimal level.
Simulation indicated that YouTube Digital Video and Twitter Digital Social for Asda were operating at Sub optimal level. There is a possibility to further increase spends for YouTube and Twitter.
Project successfully developed and implemented CatmanAi solution for a popular FMCG Food Brand in UK and machine learning media models were built for 6 largest ecommerce retailers across UK. Simulation and Optimization was implemented for each media channel at individual level. Saturation curves were generated for all media vehicles.