You are on page 1of 39

USING PLANNER AND

CHANGING RESPONSE CURVES


It is better to be roughly right than precisely wrong - John Maynard
Keynes

This artwork was created using Nielsen data.


Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

WELCOME!
IN NEXT 5 MINS WRITE 3 THINGS YOU WANT TO
GET OUT THIS SERIES [IF YOU HAVEN’T DONE IN
PREVIOUS SESSISONS]

2
GROUND RULES
During the training …

• Challenge us, ask questions, share any concerns, this is an extremely


important juncture, and we do not want to miss anything.
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

• Please give your mobile phones and laptops some rest

• Critically think of how would you change behavior when you get back from
the training. The proof is in the pudding….

3
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

LEARN
To UNDERSTAND
UNDERSTAND TO BE ABLE TO EXPLAIN

TEACHERS
Become TRAINERS
4
THE POWER OF 3
3 Things to Learn High-level overview of the tool
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

Marketing
Mix
What happens when Understanding what
we click optimize MPO is saying and
how to recommend

5
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

3) Change
2) Immerse
1) Challenge
MY ASK OF YOU IS SIMPLE

6
OPTIMIZING WHAT DOES
THAT MEANS?

This artwork was created using Nielsen data.


Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

BEFORE GOING INTO DETAILS LET’S UNDERSTAND THE CURVES

8
WHAT ARE RESPONSE CURVES?
In case of media campaigns, the x-axis is referring to the adstocked GRPs – Best
practice to load TRPs, although GRPs can be loaded as well
Technically Speaking:
• A response curve describes how lift over core (represented by the y-axis)
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

changes with changes in adstocked GRPs (represented by the x-axis) “in any
given week”.

• It will also tell you the point on x-axis (adstocked GRPs) at which the lift will
saturate (not produce any additional lift).

9
WHAT DOES IT MEAN TO US

• Core: Any volume which can be assumed to be unaffected in short-term by


Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

changes in marketing strategy – For example – Base Volume, seasonality,


holidays, etc.

• Lift: Additional Volume achieved by an activity as a percent of Core.


Mathematically Lift = Volume from activity/ Core volume

• Adstocked GRPs: Describe the prolonged or lagged effect of advertising on


consumer purchase behavior

10
LET’S LOOK AT AN EXAMPLE
Each dot represents week
1.8

1.6 eek
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

1.4

1.2

0.8

0.6

0.4

0.2

0
0 50 100 150 200 250 300 350 400 450

Lift His

11
LET’S LOOK AT AN EXAMPLE
1.8 Lets fit a line
1.6
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

1.4

1.2

0.8

0.6

0.4

0.2

0
0 50 100 150 200 250 300 350 400 450

Lift His

12
LET’S LOOK AT AN EXAMPLE
1.8 Lets fit a line
1.6
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

1.4
400 adstocked GRP means 1%
1.2 lift
1

0.8

0.6

0.4

0.2

0
0 50 100 150 200 250 300 350 400 450

Lift His

13
IMPORTANT
Remember we cant think of lift in terms of volume as the volume will
depend on the Core
Lets see effect of executing 400 adstocked GRPs in different weeks:
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

Week Core Lift Volume


Volumeassumed
1 1000 1% 100

2 20 1% 2

3 300 1% 30

4 5000 1% 500

Hence same GRPs can generate different volumes depending upon


time of execution
14
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

REVENUE
COME BACK TO
LET’S SEE HOW WE

15
MATHEMATICS BEHIND REVENUE
CALCULATION
CORE VOLUME PRICE IN THE
IN THE GIVEN
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

GIVEN WEEK
WEEK
𝒘 = 𝒍𝒂𝒔𝒕

∑ 𝐶𝑜𝑟 𝒆 𝒘 ∗ 𝐿𝑖𝑓𝑡 𝒘 ∗ 𝑈𝑛𝑖𝑡 𝑃𝑟𝑖𝑐𝑒 𝒘


𝒘 =𝟏

%AGE LIFT IN
THE GIVEN
WEEK

16
GIVEN RESPONSE CURVE
How the Revenue Curve would look like assuming Core is flat and price is
constant
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

𝒘 = 𝒍𝒂𝒔𝒕
CONSTANT
∑ 𝐶𝑜𝑟 𝒆𝒘 ∗ 𝐿𝑖𝑓𝑡 𝒘 ∗ 𝑈𝑛𝑖𝑡 𝑃𝑟𝑖𝑐𝑒 𝒘
𝒘 =𝟏

Nielsen Confidential 17
REVENUE CURVE ON DASHBOARD
What is the difference ?

SUMMATION
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

OF CHILD
CAMPAIGNS

PEN = 50
EF = 3

PEN = 91
EF = 4

18
PROS AND CONS OF DASHBOARD

• LIKE
• It shows the relationship between total Spends (X-axis) and total Revenue
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

(Y-Axis)
• The curve gives the overall picture across dimensions (at aggregated level)
• Easy forecasting with key KPIs

• WATCH-OUTS
• Flight pattern changes could cause curve to change
• No Campaign level insights
• Strong (and steep) events can cause cluttering

19
LET’S UNDERSTAND
OPTIMIZATION

This artwork was created using Nielsen data.


Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.
Nielsen Confidential
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

Max
Revenue
• Where the engine would aim?
REVENUE OPTIMIZATION

Nielsen Confidential
21
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

• Where the engine would aim?


PROFIT OPTIMIZATION

Max Profit

Nielsen Confidential
22
WHICH ONE TO CHOOSE?
 No Rule – Depends on Clients Business life cycle
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

 Is your business Growing?

 Do we have revenue targets?

 Is client more interested in Expanding the portfolio?

 Normally Higher profit may result into lower revenue increase

23
COMMERCIAL ASPECT
• TYPICAL QUESTIONS FOR A SIMULATION
• Optimal Split between Media and Non-Media
• Optimized Flight pattern
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

• What if analysis
• MAP TRAPS
• Why my TV ROI decreased after I optimized?
• Why my Revenue Curve changed after I optimized?
• Why results are different when I forecast and when I optimize?
• Optimized TRPs are not commercially viable for purchase
• MAP is allocating all GRPs across campaign in few weeks only. Why?
• Why Promotion simulations are looking bad on MAP?
24
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

RESPONSE CURVE
WHAT TO LOOK FOR IN A

25
HOW TO REVIEW THE RESPONSE
CURVES?
• Within Nielsen Mix, Response Curves are generated while loading planner.
• They can be viewed from the ‘Response Curves’ link on the job details page
of the planner job:
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

26
HOW TO REVIEW THE RESPONSE
CURVES?
The ‘Response Curves’ link will open a list of response curves produced
by the Nielsen Mix
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

High and positive AICc denotes bad fitting


Response curves
For convenience this page is sorted by
AICc

27
HOW TO REVIEW THE RESPONSE
CURVES?
• It is important to review the response curves and make sure they have a decent
fit to the data points available.
• Another thing to check is if the Response Curve has the correct trend – either
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

Increasing OR decreasing (Own media shouldn’t have decreasing trend)


• For example, if a TV campaign shows a decreasing trend, then either
something is wrong with the data OR the configuration of the response curve.
• Here is an example of a good fitting response curve:

28
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

BAD

UGLY
GOOD
LETS SEE SOME MORE EXAMPLE

29
LET’S LOOK AT THE TOOL
LIVE?

This artwork was created using Nielsen data.


Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.
ITS VERY IMPORTANT TO UNDERSTAND
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

THAT TWEAKING RESPONSE CURVE IS


NOT AN IDEAL PRACTICE – IF MODEL
PARAMETERS HAVE BEEN CHOSEN
CORRECTLY – THIS STEP CAN BE
AVOIDED

31
RESPONSE CURVES CAN ONLY BE INFLUENCED
FROM THE OMAR CONFIGURATION FILE
• While the underlying data cannot be changed for response curves, CG does provide with
a bunch of configuration settings that can be changed to tweak the response curves.
• One of them is RC Scaling. Refer to the ‘Response Curve Scaling’ document for mode
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

details on this.
• Along with that, the following configuration settings can be found under the
DueToDimension Table and EventDimTable of the Omar config file:
1. ResponseCurveOriginFlag: 1 if the curve has to pass the origin; 0 otherwise

2. ResponseCurveLiftSaturation: Controls when the curve should saturate

3. SaturationLB & SaturationUB *: Controls the saturation level

4. ResponseCurveShape: Controls the shape of the curve. The option is converted to the
min and max of the parameters

and written into ConfigTable.csv in OMAR process.
* This setting does not exist in the OMAR Config file by default. It can be added if required.
32
UNDERSTANDING LIFT SATURATION
Saturation

Contraint : the curve should reach 95% of the asymptote at LiftSaturation times
the max historical causal value
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

By default this value is 5.

95% of asymptote

Max historical Causal


value

3 30 X 5 15
0 0

33
UNDERSTANDING SATURATIONLB AND
SATURATIONUB
Constraint: the asymptote of the curve should be within SaturationLB times the
max historical lift and SaturationUB times the max historical lift.
Default value is -1, which means no constraints.
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

SaturationUB is 2
125 X 2

Max historical Lift

125

125 X 0.5
SaturationLB is 0.5

34
DIFFERENT VALUES OF RESPONSE
CURVE SHAPES SUPPORTED BY CG
• IncrC: Returns an increasing C curve.
• DecrC: Returns a decreasing C curve.
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

• IncrS: Returns an increasing S curve.


• DecrS: Returns an decreasing S curve.
• AutoC: Returns either an increasing C or an decreasing C curve.
• AutoS: Returns either an increasing S or an decreasing S curve.
• AutoIncr: Returns either a increasing C or an increasing S curve.
• AutoDecr: Returns either a decreasing C or an decreasing S curve.
• AutoAuto: Returns a curve that can be decreasing or increasing, C-shaped or S-
shaped.

35
WATCH-OUTS
Understanding the difference

• The Response curve can differ from Reach curve due to variations in Core
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

• Tweaking of Response curves can only be done at event level – We can tweak
Total TV which will apply to all campaign within TV. We cannot tweak a
particular campaign

• Avoid talking about optimal point on response curve as definition of optimal


can vary (Revenue v/s Profit)

• Understand the trend – C shaped curve tells that campaign is good – however
it also implies that it will saturate quickly 36
“ READING FURNISHES THE
MIND ONLY WITH MATERIALS
OF KNOWLEDGE; IT IS
THINKING THAT MAKES
WHAT WE READ OURS
Attributed to John Locke
This artwork was created using Nielsen data.
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.
THANKS – QUESTION?
YOU CAN ALSO REACH OUT TO VAIBHAV.GARG@NIELSEN.COM

This artwork was created using Nielsen data.


Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.
38
This artwork was created using Nielsen data.
Copyright © 2017 The Nielsen Company (US), LLC. Confidential and proprietary. Do not distribute.

You might also like