PUBLIC SAP Innovation Awards 2022 Entry Pitch Deck Arobot - AI Flavors & Fragrances Huabao Group
2PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Company Information
Headquarters
Industry
Web Site
Hong Kong
Consumer Products
http://www.hbglobal.com/en/
Huabao International Holdings Limited was established in 1996 and listed on Hong Kong Stock Exchange (00336.HK)
in 2006, with major business in R&D, production and sales of flavors and fragrances (for tobacco, food and
personal care), aroma raw materials, tobacco raw materials and ingredients (including RTL, expanded stems and
innovative tobacco materials) and condiments. With customer demands as the core and technological innovation as the
motivation, Huabao insists the development strategy of “Concentric Diversification” and has become a modernized and
internationalized multinational conglomerate. The corporate’s vision and mission are:
❖ Establish an innovative organization and management mechanism to accomplish the strategy upgrade of Huabao
group to become the ‘Leader to a Better Life’.
❖ Modernize and digitalize the flavors and fragrances technology to form a future oriented, globalized and diversified
industry cluster and sustainable business centered around flavoring people’s life
❖ Build up technology innovation platform and flavor industry ecosystem to grasp the trend of consumer upgrade
❖ Expand the footprint of aroma raw materials segment and establish a leading production base of aroma raw
materials in Asia and in the world
3PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Challenge Solution Outcome❖ Digitalize scents and trawl flavor and
fragrance knowledge with digital
formulation generation
❖ Apply Machine Learning (ML) and
Artificial Intelligence (AI) to create new
fragrances with minor tweak by skilled
perfumers
❖ Leverage ML and AI to analyze and
predict consumer flavor preferences by
maximizing the olfactory performance,
generating enhanced formulas without any
bias, recommending the new fragrance.
❖ More efficient and optimized
fragrance innovation and making
process - enhance formulas, identify
new ingredients combinations with
fewer human expertise tweaks.
❖ Personalize fragrances for unique
customer experience – more
efficiently understand consumer
preferences to provide customized
scent to improve consumer satisfactory
and experience.
❖ Whole fragrance-making process and new flavor invention heavily depend on the personal experience of perfumers
❖ Dozen years to develop perfumers, requires professional and intelligent tools
❖ Knowledge with perfumers only, lack of extracting knowledge and organizing formulation library
❖ Lack of consumer insights during new flavors and fragrances development
Huabao Group
3 Mins
Consumer2Manufacturing Collected 461 personal flavors and fragrances surveys, predicted and produced customized fragrances during the 4th China International Import Expo
461Knowledge Base Analyzed and extracted 20,235 formulation of perfumes and fragrances by machine learning modelling
20K+
Arobot - AI Flavors & Fragrances
Intelligence and Automation With AI powered and automated formulation to production process, a new fragrance is created and made in 3 minutes
4PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Arobot illustrates the huge potential of leveraging AI in
our digitalization journey, it’s the right way to go in the
future flavors and fragrances making process, and we’ll
broaden and deepen the technology innovation practice
in our products innovation especially for our core
business of food and beverage flavors.
- Lei Shi, VP of R&D, Huabao
5PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Business Challenges and Objectives
Business Challenges Project Objectives
• The status quo fragrance-making and new
flavor invention process heavily depend on the
personal experience of perfumers, which
takes as long as a dozen years to develop,
and the formulation knowledge is saved
nowhere other than with the perfumers.
• The new fragrance or flavor tailoring process
for every enterprise customer is long and
iterative as customers’ requirement or desire
of scents cannot be explicitly described, the
perfumers can only randomly try to look for
what customer may be satisfying with
• This slows them down from new innovations
and understanding customer or consumer
need, not to mention predicting the trend
• To improve efficiency of flavor and fragrance
making process, the breakthrough idea of the
project is to digitalize scents, so that we make it
possible to leverage AI to do the job of trawling
scent knowledge, generating formula and creating
new fragrances
• Automate formulation to production process to
realize digital manufacturing and improve fragrance
making efficiency and reduce cost
• To lead the digitalization and innovation of flavors
and fragrances for a better life, Huabao would like
to better understand customer and end consumer
preference. The project will use machine learning to
predict and customize fragrances that customers
may like.
6PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Project or Use Case Details
❖ Scent Digitalization
▫ Note Circles
▫ Typical Notes
▫ Characteristics
▫ Note Pyramid
❖ Digital Perfumery Master
▫ Perfume Knowledge Base
▫ Fragrance Library
▫ AI Fragrance Creation
❖ Fragrance Formulation
▫ Essence Library
▫ Note Contribution
▫ Formulation Generator
❖ Fragrance Recommendation
▫ Preference Questionnaire
▫ Preference Prediction
▫ Fragrance Customization
Arobot is a tool which deploys machine learning to analyze and predict consumers' preferences, trawl
scent knowledge, generate formulas and recommend new fragrances catered specifically to different
personalities. Behind the scenes, the model digitizes the scent notes, symbolic ingredients, and
formula combinations. It also absorbs and learns from existing knowledge to create a formula library
that will innovate, optimize, and enhance the formulas with minimal human expertise tweaks.
Core functionalities and use cases are:
7PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Business Process Details
Process Before Process After
The current development of new flavor is composed of
product imitation, component analysis, essence ratio,
composition design and sampling experiment.
Basically, the current research process of new
flavor products mostly relies on the personal professional
experience and artistic appreciation of senior perfumers.
The research cycle of new product is quite long, and the
perfumers are still working on office software or even
paper forms manually, lacking digital and intelligent tools.
After the new formula is designed, a lot of manual
experiments are needed to prepare and sample for the
raw materials to make the essences. These experiments
require high precision and cost a lot.
Furthermore, it’s difficult to adopt data-driven approach by
market trends and consumer insights in the research
process of new flavor, which causes considerable risks
when they go to the market.
On the other hand, since there’s no good enough or
digitized domain knowledge base and formula library for
rapid learning and growth, junior perfumers still rely on
senior ones for 5-10 years of long-term counseling and
teaching to get started.
Arobot is changing the status quo.
Based on the digitalization of the notes, perfumers can analyze the characteristic of the structured note, the contribution of raw materials aided by the visualization of the fragrance formulation. By using AI fragrance application, even junior perfumers can fine-tune and modify existing fragrance formulations to generate new ones. Even, AI fragrance can generate new formula by itself based on learning large scale of existing fragrance notes and formula data and then evaluate new products.
Digitalization and intelligence of new product research process can support perfumers effectively produce a large number of formula combinations, conduct trial and error. Meanwhile the perfumers can optimize, and sublime formulas generated by AI.
At the same time, the questionnaires collected from the market research will help to effectively design the market trends and consumer preferences in the process of new product development, and quickly realize the flexible production of personalized fragrances.
8PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Benefits and Outcomes
Business or Social IT* Human Empowerment
* IT benefits are required if you are using SAP Business Technology Platform products
❖ Enable Direct to Consumer
business model, produce
personalized fragrance for
consumers in segment of one,
and integrate with digital
manufacturing to automate the
E2E Consumer2Manufacturing
processes
❖ Close the loop from obtaining
consumer insights, to fragrance
creation, production, and
feedback collection to
accelerate product innovation
and with lower cost
❖ Successfully transform from
the traditional production
process into an innovative,
digitalized, intelligent and
efficient way of developing new
products.
❖ Turns the cumbersome
fragrance making process
which requires massive
human expertise work and
years of development into a
semi-automated efficient one
that only needs a few
minutes with minor human
interactions
❖ Productivity gain of skilled
performers so they can
spend time on more valuable
tasks in fragrance invention
and tweak
❖ Prominent efficiency,
productivity gain and cost
saving with the digitalized
fragrance innovation and
making process
❖ Presents the huge potential
AI could bring to the
manufacturing business,
leading the trend for AI
technology becomes the
norm in developing new
flavors, personalized
cosmetics and other
consumers product.
10PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
SAP® technologies used:
Date
Number of end usersDeployment status
Number of customers
Transaction Volume
Deployment Details 1 of 2
SAP
product
Primary
productDeployment status LIVE or POC [proof of concept]
Contribution
to project
1SAP S/4HANA Consumer
Industries Cloud
POC SAP Consume Industry Cloud offers the foundational platform for this
intelligent application as various business services.
2SAP Business
Technology Platform
POC SAP Business Technology Platform supports the industry cloud by K8S
which is offering capacity of the elastic computing and resources.
3SAP HANA Cloud POC For persisting the data of consumer surveys, digitalized perfumes and
notes, the composition of fragrance.
4
5
2021/10/15
20~30POC
1
100+ Surveys, 60+ Perfumes per day
X
11PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
The following offerings from SAP Services and Support were utilized during the implementation
or deployment phase
SAP MaxAttention™
SAP ActiveAttention™
SAP Advanced Deployment
RISE with SAP for Industries
SAP Innovative Business Solutions
Other:
SAP Value Assurance
SAP Cloud ALM
Deployment Details 2 of 2
SAP Advisory Services
SAP Customer Experience Solutions
SAP Preferred Success
SAP Solution ManagerSAP Innovation Services
SAP Enterprise Support
Contribution to the project
❖ SAP Innovation Services worked together with SAP S/4 HANA Consumer Industry Cloud team in designing and running a Design Thinking workshop together with customer Huabao, in which the idea of Arobot was incubated, i.e. leverage AI and Machine learning to transform and customize the fragrance making process by predicting and capturing consumer preference
❖ It is being planned for SAP Innovative Business Solutions to provide digitalization services to help Huabao in bringing the Arobot POC to productive usage
X
X
12PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
The following advanced technologies were part of the project.
Advanced Technologies (1 of 2)
Technology
or use caseProduct *
Contribution to project and how product
used integrates with SAP products
1 Intelligent technologies Machine Learning Foundation Services of
recommend algorithm, recognize service of
note digitalization, regression and
classification of perfume preference
The core functionalities of Arobot is realized by
machine learning services and algorithms. Note
digitalization, intelligent perfumery and
composition generator. Even more, Arobot
applies the recommend and classification on
personal perfume prediction.
A) Machine learning or
artificial intelligenceConversational AI, AI-based knowledge graph,
AI Business Services, Robotic process
automation
B) Blockchain
C) Internet of things
2 Intelligent data managementMulti-cloud and or hybrid deployment,
data virtualization & governance, privacy
compliance, cloud data lake service
SAP S/4 HANA Consumer Industries Cloud
SAP Business Technical Platform
Arobot is a native cloud application. The
consumer industry cloud which based on SAP
BTP offers the foundation of elastic computing
and resources.
*If this is not an SAP product, explain how it integrates with SAP products.
13PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
The following advanced technologies were part of the project.
Advanced Technologies (2 of 2)
Technology
or use caseProduct *
Contribution to project and how product
used integrates with SAP products
3 Advanced cloud integrationNew business models using API’s, Connecting
business partner(s) with API’s, Integration Advisor,
Digital integration hub architecture, Event Mesh
Open APIs of integrating with
the digital manufacturing
machine
Arobot offers the open APIs for the manufacturing machine to
communicate. The machine access the VPN network created by
SAP BTP and accept the production order of perfume which is
forecasted by personal survey of perfume preference.
4 Advanced and augmented analyticsReal-time and streaming analytics, spatial
analytics, natural language processing, machine
learning to identify trends, patterns, and outliers,
predictive analytics and planning)
Word cloud of the
comment analytics from
consumers
Dashboard of the metrics of
AI perfumery
Arobot collects the comment of consumer from market and
analyse the data as the critical input of the AI perfumery.
Meanwhile some predefined metrics during the production are
monitored by the application.
5 Combined transactions and analytics on single data setReduce data latency and footprint from dedicated
data marts, data warehouses and data lakes (>
1TB)
*If this is not an SAP product, explain how it integrates with SAP products.
14PUBLIC© 2021 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Additional Information
Arobot Debut at 4th China International Import Expo
Customer Traction & Showcase VideoReport by China People’s Daily