Increasing the value of Shoppar’s Facial Analytics Platform

My experience working with shoppar.ai as a Pivigo Data Science Fellow.
August & September 2018 – London, UK.

 
 

INTRODUCING SHOPPAR

Shoppar is an early start-up retail analytics company, developing real-time data analytics technology and software. Working closely with major brands, including: Samsung, Hugo Boss, Sky, Western Union and Heineken, Shoppar leverages the capability of in-store digital marketing displays to capture customer demographics; for example a rough age and gender, and behaviour; for example mood & attention span. Shoppar’s retail clients can then view their data via a custom built dashboard application, and use this to better understand the effectiveness of their digital marketing campaigns


SHOPPAR’S VALUE
PROPOSITION

Online shopping has come along way since it first appeared in the 90’s, and has evolved rapidly as e-retailers jump on every opportunity to integrate the latest trends in technology. Meanwhile retail stores struggle. They have yet to catch up with the novel technologies that make online shopping appealing…. but used effectively the combination of digital signage and AI could be the game-changer that re-imagines the retail customer experience. Shoppar are uniquely positioned to make a significant impact in this space. Their challenge is helping their broad client base create a more engaging and meaningful branded relationship with consumers, that is truly customer-centric and goes beyond passive advertising displays. Although the retail digital signage industry is large and growing at an annual rate of 6 - 8%, most retail stores display adverts that are not targeted towards customers' demographics, and they are unable to measure the level of customer engagement and response to the content. Shoppar’s goal is to change this.


OUR WORK

Beyond the aggregated stats in these dashboards, Shoppar do not currently leverage their rapidly growing data store. Data science enables Shoppar to unlock deeper insights from their data; identifying trends and extracting meaning that enables them to further develop their product and service. Their end goal is to maximise the effect of marketing strategies and deliver increased value for the clients.

Our work with Shoppar focussed on three key areas:

  • Validating test data and developing ways to better identify systematics and biases in the data;
  • Developing a machine learning framework to advance their capability around customer demographics and engagement; and,
  • Prototyping ideas around content optimisation to inform future development and a more responsive product.

OUTCOMES
& VALUE

 

SUMMARY

We developed a comprehensive data cleaning and analysis pipeline that;

  • Enables Shoppar to better explore individual client data;
  • Helps Shoppar identify biases (and implications) from their facial recognition algorithm (tensor flow modelling);
  • Brings in external weather data specific to each device location, previously shown to affect customer footfall, or traffic.


We created a modelling framework using sophisticated machine learning algorithms;

  • Includes diagnostic visualisations that highlights useful data (features) that can be used to better predict attention span;
  • In future, this will enable Shoppar to leverage their rapidly growing data store and to make more robust predictions.


As proof-of-concept we explored both open-source and proprietary solutions for intelligently
generating new content from existing marketing campaigns;

  • Optimised to match customer attention span and demographics.

 

PROJECT TIMELINE


PROJECT TEAM
& ROLES

Project Management & Client Communication | Shared responsibility
Commercial + Technical Research | Arna Karick, Alex Dunhill
General Programming | Arna Karick, Jack Roberts, Alex Dunhill, Antonello Macchia
Integrating External Data | Alex Dunhill, Jack Roberts
Workflow Design & General Prototyping Arna Karick, Jack Roberts, Alex Dunhill, Antonello Macchia
Data Visualisation | Arna Karick, Jack Roberts, Alex Dunhill, Antonello Macchia
Machine Learning Algorithms | Arna Karick, Jack Roberts, Alex Dunhill, Antonello Macchia
Object Recognition & Scene Detection | Jack Roberts, Alex Dunhill, Antonello Macchia
Commercial + Open Source APIs | Jack Roberts, Alex Dunhill

Mid-session Pitch | Alex Dunhill
Final Presentation + Business Case Study | Arna Karick, Jack Roberts, Alex Dunhill, Antonello Macchia

Shoppar Advisors | Peter Ward (CEO & Founder), Kris Milne (CAO), Akshay Kumar (CTO)
S2DS Mentors Adi Andrei (Technical Mentor), Ben Turner (Business Relationships Mentor)


 

 



S2DS London 2018

My experience as a Pivigo Data Science Fellow.
August & September 2018 – London, UK.

 

Pivigo's Science to Data Science (S2DS) Fellowship program is one of the most competitive data science fellowship programs in the world, and Europe’s largest. Only 90 fellowships are awarded each year.

 

ABOUT THE
FELLOWSHIP

The program is of a similar calibre to the US-based Insight Data Science Fellows Program program and UK-based ASI Fellows Program. All data science fellows have PhDs, with the majority coming highly analytical, data-intensive, scientific research backgrounds (e.g. neuroscience, computational chemistry, mathematics, astrophysics, physics, computational biology, bioinformatics, environmental science, computational linguistics).

Fellows range from newly minted PhDs, to early- and mid- career postdoctoral researchers, to senior (in some cases tenured) professors, to ex-academics who have already made the transition into industry. During the fellowship, the majority of time is spent working on commercial data science problems; either as external consultants or embedded data scientists, in tech startups (e.g. Shoppar), established or high-growth technology companies (e.g. MADE.com), data science consultancies (e.g. HAL24K), small and medium sized enterprises (SMEs), multi-national companies, charities and non-government organisations. Fellows are supported by commercial experts – e.g. CTOs or Heads of Engineering/Analytics, technical mentors – experienced data scientists and technologists from industry, and work as part of a data science team to develop high-impact, cutting-edge, sustainable, and scalable data science solutions. 


BLOG POSTS