The Measurement and Growth in the ROI of Marketing
Jun 20, 2015
The Measurement and Growth in
the ROI of
Marketing
Agenda for today
1. Brief introduction of media effectiveness measurement.
2. Quick case studies on effectiveness measurement.
3. Questions and Issues
3
The long & winding road of marketing
Some questions prompting need for effectiveness measurement What are the most effective marketing channels – TV, Radio, Outdoor, Print or Digital?
What has been the ROI of paid search and display ads?
How important are holiday promotions?
What is the impact on sales if we increased product distribution?
What is the impact on sales if we increased prices by x%?
What is the best way to allocate my marketing budget?
Which marketing messages are most effective in driving sales per investment?
Are there any synergistic effects from executing integrated marketing campaigns?
4
The ideal situation
Spend X on Activity Y
Sales increase by Z
The problem: Too many channels, fragmentation and confusion
Sales
…simply eye-balling historic media and sales data does not provide answers
7
…econometrics can help A statistical technique used to identify and quantify the incremental contribution made by marketing
investments on sales*
Sales
Macro-Economic
Factors
Distribution
Price
Seasonality (Temp & Periodic effects)
Promos/ Direct
Marketjng
Design & Packaging
Paid Digital Media (PPC,
Display)
Mass Media: TV,
Radio, Print, OOH
8
…simple econometric sales equation
9
Once we have determined the B weightings for each media variable, we can fit historic sales data to a very high
degree.
Sales = base + B1TV + B2Radio +
B3Digital + B4Outdoor
Estimate for sales generated from non-marketing activity Weighting for TV Weighting for Radio
Weighting for Digital Weighting for Outdoor
Key Deliverables from Econometrics Projects
10
-200
0
200
400
600
800
1000
1200
1400
1600
1800
Oct
-20
10
No
v-2
01
0
De
c-2
01
0
Jan
-20
11
Feb
-20
11
Mar
-20
11
Ap
r-2
01
1
May
-20
11
Jun
-20
11
Jul-
20
11
Au
g-2
01
1
Sep
-20
11
Oct
-20
11
No
v-2
01
1
De
c-2
01
1
Jan
-20
12
Feb
-20
12
Mar
-20
12
Ap
r-2
01
2
May
-20
12
Jun
-20
12
Jul-
20
12
Au
g-2
01
2
Sep
-20
12
Oct
-20
12
No
v-2
01
2
De
c-2
01
2
Jan
-20
13
Feb
-20
13
Mar
-20
13
Ap
r-2
01
3
May
-20
13
Jun
-20
13
Jul-
20
13
Au
g-2
01
3
Sep
-20
13
Un
its
Sold
Variance
Modelled Sales
Actual SalesForecast
R2 = 97.6% - This indicates a very high level of accuracy to the actual sales data. Holdout R2 = 89.9% - This indicates fit for the forecast (holdout) test. MAPE = 4.2% The average % variance between actual sales and fitted/predicted.
Step 1: Develop a highly predictive sales model Develop a predictive model by mapping historic media activity against sales. We deliberately holdout 10-15% of the dataset to test for predictive accuracy. The premise and validity of our models depends on validating predictive capabilities
Step 2: Overall media contribution to sales
Decomposition of sales provides a snap shot of the overall importance of media and marketing in driving total sales. In this case, 33% of total sales revenues are “due to” marketing expenditures.
12
The first step in improving marketing productivity is to determine precise financial returns to marketing spending by campaign/activity.
Step 3: Returns by channel (ROI per $1 spent)
Least efficient channels
13
Step 4: Response Curves & Diminishing Returns
Current Spend
Press Saturation Press Paid Search
Optimal Spend/Allocation
Increase Spend
Increase Revenue
Reduce Spend
Decrease Revenue
We can provide guidance on which channels may be getting saturated and where the opportunity may exist to increase sales.
14
Step 5: Optimal spend solution across all channels
Mathematical optimization of your marketing spend shows you how to generate between 2-8%
higher revenues without increasing total marketing investment.
$15.5m $15.5m
15
Play out marketing What-if scenarios
An interactive dashboard allows you to simulate different marketing mix/spend
scenarios and assess the resultant impact on sales and profitability.
1. Set marketing budgets.
2. Set your spend levels across media channels
3. Assess the resultant impact on sales & profit
16
Providing Answers to Business Questions
17
Issue 1: What if I increase the number of SKUs in stores?
Distribution Index % Impact on sales
55 1.8%
57 2.1%
60 2.5%
We can model the effects of retail distribution on sales of the brand In this case the response of sales distribution is highly nonlinear and continuing to increase distribution would increase sales by less and less.
0200400600800
100012001400
% In
cre
ase
in S
ale
s
Weighted SKUs
Distribution Increases - % Impact on Sales*
1010
18
Issue 2: Measuring the Halo Effect of Media
Our approach allows us to assess and quantify the Halo or Spill-over effect from one brands
media to another. For example what is the impact of Coca-Cola brand media on Diet Coke>
Parent brand
Sub brand
Sub brand
Parent brand
Issue 3: How does Competitor advertising affect my sales?
% Effect
Competitor A -3.7%
Competitor B -1.2%
Competitor C -2.0%
Here we show that not only do the activities of competitors have a negative impact on sales of the brand measured. But also that some competitors have a greater impact than others
0
50000
100000
150000
200000
250000
300000
350000
400000
450000
-
50
100
150
200
250
300
350
400
Dec 08 Dec 09
Co
mp
etit
or
Act
ivit
y (€
)
Sale
s
Aleve
Nurofen
Finmal
Actual
20
Issue 4: How responsive are my customers to changes in price?
21
Household Cleaning Products Price Elasticity
We can measure the responsiveness of changes in product pricing to changes in sales. This allows
us to guide on pricing that will be profitable.
2013 Elasticity
22
Issue 5: We can explore the effect of price promotions In this situation when clients are price discounting to the same level of the competition there is a slight marginal benefit.
Impact 10% off promo 10% promo and
Feature and Display in 10% of stores
10% competitor promo
10% off Brand and competitor promo
Volume 22.6% 26.1% -7.8% 13.0%
Value 10.3% 13.5% -7.8% 1.7%
22.6%
26.1%
-7.8%
13.0%
10.3%
13.5%
-7.8%
1.7%
-10%
-5%
0%
5%
10%
15%
20%
25%
30%
10% promotion + Featured and display in10% of stores
10% competitor promotions 10% with competition
Ch
ange
in s
ales
Volume Value
Providing Answers to Business Case Studies
23
Case 1: Effectiveness for household cleaning product
Business Questions 1) To what extent has brand marketing driven sales and ROI?
2) Can we quantify the brand spill-over effect from parent brand (Y) spend on sub brand sales?
3) Does brand image attribute (Recommended by Doctors) impact upon brand sales?
Insight & Outcomes Overall marketing contributed to 32% of total sales
units
TV and digital media (network display and paid
search) were the largest contributors?
Doctors Recommendation drove 2% of sales.
Distribution was supported by additional in-store
features and display.
Spend optimization led to increased allocation for
digital display media and TV.
Modelling the impact of media and key brand image factor on sales
24
Case 2: Media effectiveness for Gamma toys
Business Questions
1) Gamma Playsets supported by substantial ATL activity leading up to Christmas, can we assess the ROMI
for this?
2) In addition, new store partners will help to increase distribution by 17%. Insight & Outcomes
Demand was highly seasonal and concentrated
to October, November and December.
Q4 sales were largely driven by expanded retail
store distribution and TV execution.
Retail distribution and increased brand
awareness led to a sales uplift in Q1 the
following year.
Measuring ROI for Gamma playsets
25
Case 3: Effectiveness and spend saturation for used car dealership
Business Questions
Increased spend in digital and offline channels without
media accountability?
Can we spend more to increase sales units?
Insight & Outcomes
Demand was seasonal and led by registration plate
changes.
Competitive advertising has greater cut through.
Upside potential to increase sales by spending more was
limited.
Significant synergies between Network Digital and Radio
pointed to growth opportunities.
Media contribution to sales units
Spend Saturation Analysis
Econometrics is especially important when
A new marketing campaign and sizeable investment New marketing director – new strategy and growth targets. Marketing spend up - flat sales A change in pricing or if retail distribution partners are calling for changes in pricing structure? A new product launch or change in package design Marketing allocated across a broader set of channels compared with past
Common Concerns
We only have 1 years worth of data? Econ requires 3 years worth of data, but if we can pool several markets together with 1 years worth of data, then we have a solution.
All our marketing and media is digital, so we can just measure it directly ourselves? Online marketing (display and paid search) does not build your brand equity. Your sales are most likely being driven by a multitude of external non-marketing factors such as distribution, pricing, competitor media, economic situation etc – if you measure ROI without taking these into account, you are very likely to misattribute and inflate the contribution from digital alone.
Econometrics (media effectiveness modelling) is too expensive…I mean what is the hard commercial value to our business? Expensive relative to what? As John Wannamaker said, “Half the money I spend on advertising is wasted; the trouble is I don't know which half”. Many of today’s surveys of top marketing companies find that 40-50% of marketing budgets are wasted. We will identify the 30-50% of your marketing spend that is wasted/underperforming and we’ll direct you to those activities that are much more productive in driving sales. Following this, we will optimise your marketing budget across media channels to ensure it is allocated in such a way that is geared for sales growth.
Our annual total marketing/media budget is small This works for budgets as small as £500K.
It all seems too complicated for our non-technical teams Start by explaining and visualizing to them the concept of correlation, which is the basis of MMM. Also explain to them that our job is to explain to them the complex in their own language and they will understand, and we have been doing that for years.
Our market is extremely volatile -there is no recurring patterns in our sales data. We have NEVER found a situation that could not be managed. If we are permitted to evaluate data in advance and find it cannot be modelled, we will not charge anything.
Why does Econometrics take so long to do? Developing an econometric model that accurately maps your historic sales and predicts sales going forward involves a number of iterations. This can take weeks.
What if we don’t meet the sales targets projected by the Econometrics? Econometrics is a statistical analysis, not a crystal ball. Whilst models purport to be very accurate, they are still only a yardstick. However, if you follow the optimised course of action in terms of media weight/allocation you should see an uplift in sales…this is assuming no external adverse events in the economy/marketplace.
Common Concerns, ctd.
Masood Akhtar Managing Partner, Analytics (EMEA) Bottom Line Analytics Global E: [email protected] M: +44 7970 789 663 www.bottomlineanalytics.com
30
Michael Wolfe CEO Bottom Line Analytics Global E: [email protected] M: +01 678.3143,8446 www.bottomlineanalytics.com