Finding the Diamond in the Rough (IIEX.AI 2024 presentation)

Presented at IIEX.AI on 15 October 2024 together De Beers. We introduce an innovative approach to uncovering insights in retailer sales data. By defining hierarchical jewellery categories and leveraging automated analytics, we identify trends and driving forces behind them. Inspirient's integration of ChatGPT synthesizes findings into draft summary reports.

1. Finding the Diamond in the Rough
Surfacing Trends in the JewelleryMarket
Through Automated Analytics
Diana Mitkovand Dr.Guillaume Aimetti
inspirient
cognitive analytics

2. 1
YOUR SPEAKERS
Diana Mitkov
SENIOR INSIGHT MANAGERGuillaume Aimetti
CO-FOUNDER & CTO

3. 2
BUSINESS CONTEXT
Consumer Demand Estimation
Price Optimisation
Demand and Inventory Forecasting
Strategic Decisions
Industry Influence

4. 3
TRANSACTION DATA FROM INDEPENDENT
DIAMOND JEWELLERY RETAILERS…
1147 retailers
5 countries
84% US retailers
Fortnightly data feed
Over 1m transactions a year

5. 4
TARGET SOLUTION
FortnightlyPOS
transaction data
Self-service
dashboard
Monthly executive
summary communication
Consolidation
and analysis

6. 5
THE CHALLENGES
Connecting the raw transaction data to
BI dashboarding tools doesn’t cut it…
1.Too much data for multi-year analysis
2.Non-trivial jewellery classification
3.Data quality issues
4.Hierarchical category processing
5.Detecting and explaining trend
changes across all categories

7. 6
THE SOLUTION
FortnightlyPOS
transaction data
PowerBI
dashboard
Automated executive
summary
Analytical
results
Secure Large
Language
Model
inspirient
cognitive analytics
Consolidation and analysis

8. 7
THE SOLUTION: INSPIRIENT AUTOMATED ANALYTICS ENGINE
Inspirient’s hybrid AI
that autonomously
applies analytical
reasoning
Inspirient
Core
Add-on components
Pattern Detection
Runtime (incl. adaptive caching for in-memory operations and real-time processing)
User Interface // REST API
Custom output
Generation
Prioritization &
User Feedback
Data Model
VisualizationStory Telling
Base Analytics (incl. descriptive statistics)
Semantic UnderstandingData Enrichments
Data ImportData Cleaning &
Transformation
Anomaly
Detection
Predictive
Analytics
Survey
Analytics
Root Cause
AnalysisAdd-on
Internal and external business dataIndividualized output
inspirient
cognitive analytics
…

9. 8






















!
"#$$
%&'
()* !+

!"#$
,'-
".&$$*+$!$)$!# #*'+/ !$)*0 +/ 0 $#$
$$$)123$/4$)$)/($ ) )$)*'1"+/$# )$)*561+/$7#
8 )$)*561+/61 $$/61$)$)
$9:1 2






















!"
#$%
"!& '(')*

!"#$
+%,
-$" ! !"")* '! "!")%*." '!)/*./" "
!" '""0"& 1 2.!""3" "' '!" .( '"' '!)%&!"!*. ' '!)45&*. 6
072' '!)45&*.!5&0" 2.!""5& "' '!"
89&
THE SOLUTION: AUTOMATED ANALYSIS RESULTS
Example custom Excel reportExample generated PowerPoint

10. 9
THE LAST MILE: BEYOND DASHBOARDS
Intuitive Data Interrogation
Dynamic Digests
Timely Communication
Executive Summary

11. 10
THE LAST MILE: USER FEEDBACK
The creation of
a hierarchical structure
for the data has been
essential in understanding
the full picture
Speed of data anomalies
detection has enabled
speedy resolution with the
data provider, including
remapping at their end
Fast automated monthly
summaries provide
‘at a glance’ updates
for decision makers
A great win!

12. 11
THE AUTOMATION JOURNEY SO FAR HAS CONSISTED
OF THE FOLLOWING MAJOR STEPS…
Continuous optimisation
Nov ‘23
Definition of jewellery
categories in a hierarchical
structure, supported by data
Consolidation of sales data,
i.e., category classification
and data cleaning routines
Aggregation logic for all sales
metrics and growth rate
calculations
Dec ‘23
Investigation by data provider
into inconsistencies due to
updated extraction process
Resolution of data
inconsistencies and
re-basing completed
Process in place for
analysing data and
publishing dashboard
In progress…
•End-to-end integration
•Automated deep-dives
•Ad-hoc queries

13. 12
CONCLUSION
Implemented automated data analysis system that can
handle large quantities of transactional data which can
be analysed with non-trivial custom jewellery category
definition.
ACCOMPLISHMENTS
IMPACT
CHALLENGES OVERCOME
LESSONS LEARNED
LOOKING AHEAD
Faster time-to-insight, which frees up time to think, and
more comprehensive findings, which reduces the risk
of missing the obvious.
Third party POS data extraction bug resolved through
thorough data investigation and pro-active stakeholder
management.
Emphasized effective communication, stakeholder
engagement, and ongoing optimization.
Positioned to leverage automation for continued
innovation and value creation.