Signify Project-1
Signify Project-1
New Hue users as measured by bridge activations are down 15% YOY based on 1 st 22 weeks’ data for Project Lead(s) Tushar Kumar
2019 vs. 2018. If the trend continues, then accounting for seasonality, we are likely to hit only
540,000 new users by the end of 2019 – approximately 30% below our target of 760,000. Project Scope (In Scope) BG Home (Hue only), US Mkt., Consumer Channel, New Hue Users
Based on Feb-Apr 2019 vs. 2018 number of units sell-out data comparison, our biggest declines are in: only
• White (W): 11,769 units (-64%) at Best Buy, 7,320 units (-62%) at Amazon, and 2,045 units (-68%) at
Walmart
• White Ambiance (WA): 4,178 units (-91%) at Best Buy and 1,466 units (-93%) at Amazon Project Scope (Out of Scope) Existing Hue users looking to increase their attachments (BMC Hoshin
Project is linked to BMC Hoshin strategic initiative #1 and Horizon initiative #129. strategic initiative #2)
Bridge activations are the cleanest metric pointing to new users for Hue. The goal of this project is two-fold:
For the 1st 22 weeks, average week-over-week growth for Hue bridge activations in 2018 vs. 2017 was • Identify the root-causes for the decline in new Hue users, and address them to achieve a sustainable 20%
+29% as against -14% for 2019 vs. 2018 (Cumulative YTD growth for the same period was +31% in 2018 vs. week-over-week growth by end of Aug 2019
2017 and stands at -15% in 2019 vs. 2018). A % decline trend started in week 36 of 2018 and continues to • Hit 760,000* new users (bridge activations + Bluetooth users) by end of 2019
week 21 of 2019. The gap between current state (average week-over-week % growth in new Hue users, -
14%) and desired state (target, 20%) is 34 p.p.
Financial Impact (Revenue / Profitability / Customer NPS)
Direct impact of $14M in incremental revenue for 2019, based on the 220,000 gap in bridge activations vis-à-vis
target.
Milestones - 2019 Wk25 Wk26 Wk27 Wk28 Wk29 Wk30 Wk31 Wk32 Wk33 Wk34 Wk35 Wk36 Core Team Members
1 Project Definition X
• Tushar Kumar
2 Current Condition X • Michael Deschamps
• Angela Smith
3 Target Condition X
• Jan Pieter Dekker
4 Root Cause Analysis X X X X X • Steven Lie-Atjam
5 Countermeasures X X X X • Fabiano Vivori
• Kaitlin Sullivan (KD)
6 Results X X X • Kristen Povolny
7 Standardize and Sustain X X • Olivia Piao
Project Start Date Jun 20, 2019 Project End Date Aug 30, 2019 Assigned Business Lead(s) Jeroen S and Jeroen W
*The annual target of 760,000 new users is based on a 20% increase in the 636,135 Hue bridges that were activated in 2018; only reaching a week-over-week increase of 20% in new users by Aug
2019 would be insufficient to compensate and get us to the annual target of 760,000.
Project Timeline, Milestones and Owners
Week Ending
Mar Apr May Jun Jul
Project Stages 12th 9th 16th 23rd 30th 7th 14th 21st 28th 4th 11th 18th 25th 2nd 9th 16th 23rd 30th Owner
Data Discovery Jorge/Snehit/Vinay
Project kick off workshop
Understanding of Data and Data Sources
Data Preparation & Pre-Processing Vinay & Shivappa
Data Gathering & Collation
Data Cleansing & Preparation
Classification Model Vinay & Shivappa
Build models to classify 0th hour Complaint
Validation & Selection of best model based on accuracy
Topic Modelling Vinay & Shivappa
Build models to classify the cause of 0th hour Complaint
Validation & Selection of best model based on accuracy
Model Outcome Demo to Stakeholders Feedback & Jorge / Vinay
Modification
Analytics Solution Implementation Snehit
UI requirement Jorge
Business Requirements Signoff Harald
Finalize the Solution Architecture Marco/Snehit
TCT Approval Snehit
Solution Development Snehit
UAT & Production Deployment Jorge/Snehit
On2going Services TBD Snehit
Project Risks and Dependencies
Stage Activity Dependencies on Description Mitigation Plan
Classification & • Data from Nov 2020 – Feb 2021 (For Validation) To check availability with
Model Validation Business
Topic Modelling • Business availability to validate model outcomes Harald
Analytics Solution Business availability for the UAT of the Solution to be To check availability with
Implementation UAT Business Implemented As a result of this it will be moved to Harald
Production
Timing & Quality of Resolution of these Dependencies will have downstream impact on the timelines of the project
5 strategic questions to begin a use case
For business w/ COE support
For COE
1. What data is available related to this use case ? Is this use case technically feasible
Example of completed questions/answers
1. What strategic question(s) are we trying to answer and by when?
Since January 2018, the company collects details related to customer complaints in C4CS. For 2021, we have a specific interest to understand and reduce what we call zero-
hours-complaints, meaning things that are wrong at the very first moment the customer sees or energizes a product. While C4CS has date specific information it’s by far not
good enough to decide which incident happened at 0-hrs and which happened at a later time. Reading through customer complaints, I do think, the text of the complaint is
sufficiently specific to determine if it actually was a true 0-hr complaint.
2. What are the outcome(s) and final deliverable(s) expected from the CoE as part of the Project?
•I would like to come to a table that contains the following data fields;
a. [Ticket date] - readily available in C4CS
b. [Ticket number] - readily available in C4CS
c. Assessment if the failure is a 0-hrs complaint by looking in the text of the complaint for phrases like; ‘upon receival of the goods’, ‘when opening the box’, ‘missing item’,
‘while initializing’, etcetera – TEXT analysis
d. Determination of the likely root cause of the 0-hrs complaint, like; transportation damage, missing component; wrong product, did not light up, did not connect,
etcetera– TEXT analysis
e. Product involved [Product family] - readily available in C4CS
f. Number of failures that were reported - readily available in C4CS
3. How will we measure success for the said Project? Are there specific criteria?
The output table (see #2) will allow us to make several paretos;
Product families with highest level of complaints and/or 0-hr failures TOP reasons why customers have to get back to us at 0-hrs We’ll also be able to combine these data
with production volumes, which will then lead to reliable ppm levels.
These paretos will allow us to choose effective countermeasures and reduce 0-hr complaint levels. These in turn will reduce CONQ, improve customer satisfaction and
customer NPS and stimulate sales.
4. What top-line or bottom-line impact will the Project have over the next 12 months?
An estimated 25-35% of the roughly 1,000 complaints that Digital Solutions gets every month in the USA and Canada is a 0-hr complaint, hence the anticipated impact on
customer satisfaction is huge, possibly up to 5 NPS point. In 2020, we spent 2.8 M€ in External CONQ. We’ll be addressing 1/3 of the complaints and say we’ll be able to
reduce 1/3 of the associated cost, that would lead to an estimated CONQ savings potential of some 300 k€.
5. Who is the assigned channel or function lead to work with the CoE?
Harald will act as function lead
Market Mix Modelling -Benelux
Data cleaning
Benelux First cut model
Presentation &
ROI & Incremental
revenue Demo
PR/influencer Spend
Missing value/outlier
Final Model
treatment
Retail spend
8
ML Model details
• UCM model will be used for decomposing the sales into marketing component, seasonality & base sales
component. Because of data patterns if model fails to fit then equivalent alternative models will be used
for analysis.
• First version of the model will be developed and discussed with Benelux team, feedback’s from that
discussion will be incorporated into the final version of model for delivery.
• Following few metrics will be calculated to deliver actionable insights for marketing team of Benelux:-
Power, Annualized coefficient, User budget split, Spend ratio, Optimized marketing contribution,
unoptimized marketing contribution & optimized spend.
• Minimization function will be used for optimization of the spends across different marketing channels.
• Model developed will be productionized & results will be enabled for users to access on SAC MMM
dashboard.
9
Outcome : Benelux market will be enabled on MMM SAC dashboard
10
Marketing mix model for Benelux Market – Detailed Activity Plan ~14 weeks
Week Ending
Activities 30-Jan 06-Feb 13-Feb 20-Feb 27-Feb 06-Mar 13-Mar 20-Mar 27-Mar 03-Apr 10-Apr 17-Apr 24-Apr 01-May
Data Discovery - 2 weeks
Project kick off workshop (DS)
DS To be identified 1 8 67 536
23-Jan-23 30-Apr-23
DE To be identified 0.25 8 30 60
27-Feb-23 21-Apr-23
UI To be identified 0.5 8 5 20
23-Apr-23 30-Apr-23
12
Refresh cycle of MMM
PR/influencer Spend
Retail spend
14