Analytics to drive impact
Analytics to drive impact
2
DN has been growing at 12% per month to ~90 000 digital subscribers
SOURCE : Qlikview; Analytics team
40
60
20
50
90
30
10
80
0
70
46
67
Aug
72
Sep
5
NovMayJan Feb SepMar Apr OctJulJun NovJul MayAug
56
Oct
85
34
Dec
76
37
Jan
39
JanDecFeb
40
Mar
42
Apr Feb
50
Jun
61
7987
+12% p.m.
Subscribers# thousands
2015 20172016
3
However – with rapid growth, churn also started to become an issue
SOURCE : Qlikview; Analytics team
14.5
13.5
0
13.0
15.0
12.0
12.5
14.0
Oct
11.5%
12.5%
Sep
13.9%
Dec
11.8%
Feb Mar
14.8%14.4%
Nov Jan
13.1%
Churn, %
2016 2017
4
60
0
100
3020
80
10
50
70
90
40
Nov
Dec
Jun
Mar
Jul
Oct
May
Aug
+12% p.m.
90-100
Nov
Oct
Apr
Feb
Dec
Jan
0-1% p.m.
Sep
Dec
Nov
Aug
Sep
Jan
May
Jun
Apr
Apr
Jan
Sep
Feb
Oct
Jul
May
Feb
Mar
Jun
Mar
Aug
Jul
At these churn rates – digital stock would have leveled at ~90 -100 000
2016 2017
Churn levels after march Churn keeps increasing untill aprilChurn unchanged
Subscribers# thousands
2015
5
First - we decided to see what could be done in 1 week by combining data with machine learning and a cross functional team
FUNCTIONS
WORLD KNOWLEDGE
AUTOMATEDRESEARCH ENGINE
Open-Source
repositories
Extended custom
functions
Curated libraries
Billing data
Subscription data
Web data
Contact history
Subscription historyMisc.sources
Customer demographics
Status and subscription
Misc.sources
DN data
HACKATHON
6
And it turns out we could predict and explain churn pretty well
True positive rate
True negative rate0.5 1.0
HACKATHON
We now know with 86% accuracy who will churn
2.01x
1.72x
1.63x1.4x1.34x1.32x
… +200 features!
Lift Up/Down of churn probability6 examples of features from model
Monthly reminders (paper / digital invoice)
Longer subscription periods (students)
Autogiro / EfakturaIphones and IpadsActive >1.5 minute last weeksNot telemarketing
We identified main drivers so we could start acting on it
7
Main drivers
We created a real-time dashboard based on the most important churn drivers that we can take action on every day
IMPLEMENTING ANALYTICS
Time as subscriber
Paper invoice or credit card / Auto-pay?
Frequency of visits
Product type (e.g. E-DN)
Acquisition channel
8
Data driven mapping and cross-functional workshop used to quantify potentials and select high-potential initiatives
… and much more
Channels
Pricing /offers
▪ Introduce broader options
▪ Create Autopay flow
▪ Paywall recognition
▪ Migrate stock to Autopay
▪ Subscription optimization
▪ Invoicing & debt
▪ Telemarketing optimization
▪ New sales channelsSALE
Payment
9
To get it done – we set up a cross-functional war room
What is it What does it look like
DRIVING THE WAR ROOM
Truly cross-functional team
15 minute check-in every morning
Weekly 30 minute demo
Continuous and regular problem solving
10
2017
Our churn is down significantly since our peak in the beginning of the year
SOURCE : Qlikview; Analytics team
15141312
11109
0Jul
9.8%
Jun
8.9%
May SepAug
11.1%
Oct
11.8%
9.8%
Apr
12.3%
Mar
14.8%
Feb
14.4%
Jan
13.9%
Dec
13.1%
Nov
12.5% -28%
10.6%
Nov
10.8%
Okt
11.5%
Monthly churn, %
2016
11SOURCE : Qlikview; Analytics team
80
120
110
100
90
70
600
130
72
Dec OktAug
108
Jul
91
Mar
113
92
Feb
87
Jan Sep
8579
Nov Nov
103
76
98
Maj
92
AprSep
67
Jun
61
Jul
56
Oct
118122
Aug
Our base of digital customers is now increasing steadily Number of subscriptions per month, in thousands
20172016
Start of ’turnthe churn’
Transformation period
Reaping the benefits
12
400’?Subscribers
Dec 2014
~260’Digital
subscribers
Printsubscribers
Busy dying
Nov2017
330’