1. HMG Corporate Development Team
gloria.andreu@havasmg.com
ines.campanella@havasmg.com
oscar.munoz@havasmg.com
santiago.murillo@havasmg.com
2. 5 Continents
Billings –
$18billion
Over 100
Countries
15,000
Professionals
6th global media
agency group
3. Strong European presence
+70 countries 80 Offices 3,900 Media Experts
HavasMedia has the leading position in the key European markets:
• UK, Germany, France, Italy, Spain
4. Our clients
5,000+ customers all
over the world
100+ international
customers
5. Our job
• To establish relevant touch points between brands and
consumers
• Intermediating between brands and media for optimising
advertising budget allocation
• 80% of global advertising budget is managed by media
agencies
• Ensuring that brands’messages reaches consumers efficiently
(maximising ROI)
“Our aim is to be the world’s best company
at connecting brands
with people using creativity, media
and technology”
6. From a traditional communication model…
Brand
Brand
Brand
Brand
Brand
Brand
Brand
Brand
Brand
Brand
Brand
Brand
Brand
Brand
• Investment in mass media (TV, Radio, Press, Outdoor, ...)
• Communication performancemeasurement made through KPIs based on
audience (Gross Rating Points)
• Consumer insights are obtained through opinion surveys
7. …to a digital world
• Consumers are the most influent media (producing content in social media)
• Communication performancemeasurement is performed through many different
KPIs (visibility, engagement, conversion, recommendation)
• Market insights are extended with the analysis of content published in social
media by consumers
8. Havas Media Communication Process
Understand consumers
through data analysis
• Decision-making drivers
• Brand recommendation
factors
Design communication
strategy
Cross-media
environment (paid,
owned and earned)
Activate communication
strategy
• Budget Allocation
• Automated real-time
communication with
consumers
Optimise communication
processes
• KPIs evaluation
• Performance
measurement through
data analysis
DATA
DRIVEN
11. Intelligent Consumer Profiling
Distribution by
gender & location
Marketing Mix
for the fashion
domain
Conversation
topics when
evaluating a
product
12. Event detection and explanation
Monitors 23 KPI
− Based on volume, sentiment
and relevance
− Aggregations by brand, media
(ej. Twitter, Facebook) and
topics
Net Promoter Score
NPS = advocates - detractors
Time series analysis
13. Time series analysis
Analysis of relations between social buzz and advertising pressure
Advertising Pressure
− Audience (GRP) & Investment (€)
− Offline (TV, radio, press) y online
(display ad.) media
Social buzz
− 23 KPIs based on volume,
sentiment & relevance
− Brands, Media & topics
Discovery of relations between series
− Correlation analysis
− Events shared
14. Community Analysis
Detection of influencers, brand
ambassadors, detractors,
viralisers, …
Social Graph Analysis
Fashion Domain
15. Social graph analysis
Content propagation
− Posts most viralised (owned & earned, positive y negative)
− Cross-channel analysis of the propagation tree beyond Twitter (blogs, forums, …)
18. Volume of data sources
4,895 brands
1.3 B records
230 TB
8,940 brands
116.182 M records
178 TB
1,913 brands
6,108 M records
61 TB
301 brands
1,411 M records
13 TB
—
—
19. Consumer Connected Platform
https://vimeo.com/88775576
− Tracking of consumers across online and offline touch points generating consumer
profiles
− Real time automation of communication processes
✴Online ads, push messages sent to the smartphone, email recommendations, …
21. Conclusions & Challenges for media agencies
• Understand consumer through audience
measurement and opinion surveys
• Segmentation of consumer groups
• Small tabular data
• Batch budget allocation
• Understand consumers through Content Analysis,
Web Analytics & social media interactions
- Insights must go beyond polarity detection
- Multilingualismbeyond main European languages
- Accuracy of the algorithms is a key issue
• Segmentation of individual consumers
• Big Linked Data
- Integration of multiple heterogeneous & unstructured
data sources at scale
• Communication with consumer activated in real
time
- Meaningful recommendations beyond display
advertising
- Online-offline integration through mobile devices
22. THANK YOU
Óscar Muñoz-García
@omunozgarcia
oscar.munoz@havasmg.com
Notes de l'éditeur
Good morning everyone.
My name is Óscar Muñoz
I work for Havas Media Group and I will present our practises and needs regarding content analytics
First, let me start with a brief introduction to our company:
Havas Media Group is a media agency group focussed in the marketing and advertisement industry.
It is the 6th global media agency group and is present in the 5 continents in more than 100 countries.
With more than $18 billion billings more than 15,000 professionals work at Havas Media .
We have a strong European presence.
With 80 offices in more than 70 countries and near 4 thousand media experts we have the leading position in the key European markets, namely, UK, Germany, France, Italy and Spain.
Regarding our clients, we have more than 5,000 customers all over the world, from which more than 100 are very relevant international companies.
As and advertising media agency, our job is to establish relevant touch points between brands and consumers.
That is, we intermediate between brands and media for optimising budget allocation, ensuring that brand’s messages reaches consumers efficiently.
In summary, we optimise brand’s Return Of Investment in advertising.
Up to a few years ago, our work was an easy task, the more investment in mass media, the more visibility for the brand.
In this scenario the performance of communication was measured through KPIs based on the audience of the spots (the so called rating points).
If we needed to perform any market study, we had to made opinion surveys, since we did not have another way of capturing data about consumers.
But since some years ago, the advertising panorama has changed drastically, with the advent of Web advertising, social media, and mobile devices.
Nowadays consumers are the most influent media because of the content produced in social media by them.
In this context, communication performance measurement is getting more and more complex since we have to take into account dozens of KPIs that measure more dimensions, namely visibility, engagement, conversion and recommendation.
Therefore the market insights that we had are extended with the analysis of content published in social media by consumers.
For dealing with this complexity, in Havas Media, we have designed a communication process, in which data is the most important part of every activity, either
for market research in the understanding phase,
for automated communication with consumers in the activation phase or
for campaign performance measurement in the optimisation phase.
Next, I will show you different the different research that we have carried out at Havas Media regarding content analytics.
The first use case is focussed in segmenting consumers along different socio-demographic and psychographic categories with the aim of performing market research.
The first variable that we analyse is the purchase funnel, which is related to the different stages that consumers transit when making purchase decisions.
Then we classify opinions about products and brands according to the MM framework, which shows if consumer opinions are related to the price of the product, the point of sale, specific promotions, or the product characteristics.
We also segment consumers according to socio-demographic variables, like gender and location.
In addition we perform sentiment analysis of content, but going beyond polarity detection. UPM has developed this classifier for us, which is able to detect a rich set of human emotions about brands, including satisfaction, dissatisfaction, trust, fear, love and hate.
Finally, we also identify conversation topics in an unsupervised manner helping us to summarise huge volume of conversations.
In this slide we see some example dashboards outputted by the tool.
The one on the top-left corner shows distributions by gender and location of consumers that talk about the brands Zara, Mango and H&M
The one on the top-right corner shows the different conversation topics corresponding to consumers that evaluate the brand H&M. We can see that H&M products are mostly compared with Zara (more than with other brands).
The one in the bottom shows the marketing mix elements and their volume corresponding to the fashion domain. In this domain the most important marketing attributes are the point of sale, the price and the design.
The second use case is related with the analysis of time series generated by aggregating content analytics results.
We perform atypical events’ detection, and summarise the conversation of consumers within the day of the atypical event comparing it with the previous and the next day.
As an example, in this slide you can see the evolution of a KPI called “Net Promoter Score” which shows the difference between brand’s advocates and brand detractors.
EXPLICAR EJEMPLO DE RENAULT
In summary, we have developed 23 based on volume of posts, sentiment and relevance,
We perform KPI aggregation by brand, media and topics.
In addition, we analyse the relations between social buzz and advertising pressure.
We correlate time series that measure the evolution on investment in different media with series of social buzz.
Discovering relations between series based in correlation analysis and the events shared by these series, checking whether a given ad campaign had influence over the social buzz.
The last use case is about social graph analysis.
We perform community analysis clustering users into communities,
Detecting who are the influencers for a given brand or domain
We also perform analysis of content propagation,
Not also attending to a particular media like Twitter, but performing a cross-chanel analysis, being able, for example to detect opinion propagation from blogs, news, etc. to social networks.
Finally, I will show you the different challenges that we are trying to face at Havas Media
The first challenge regards the Variety and Velocity of the different data sources that we have to manage, analyse and integrate.
Regarding structure we have to handle data sources that go from structured ones to highly unstructured ones.
Regarding format all of these data sources are heterogeneous, having many formats for each kind of data source.
Regarding velocity the data is produced in an speed that currently we are not able to process at real time
Another challenge is related to the volume of the data sources.
We have to process billions of records and hundreds of terabytes.
Our goal is to integrate all of this data sources in order to track consumers across online and offline touch points gathering information about them
with the aim of perform real time automation of communication process
Serving online ads, push messages sent to the smartphone, email recommendations, etc.
Let me illustrate this view with a short video
The idea here is to integrate the data gathered from web sites with the content extracted from social media and the events captured in the real world through mobile devices.
From each media we can extract different knowledge about consumers.
That, once integrated, help use to produce different outcomes like the ones presented in the slide.
As a conclusion, Media Agencies are moving from communicating in mass media to a communicating in a digital eco-system in which social media and mobile devices are more and more important.
In this context, our challenges are the following:
Understand consumers through Content Analysis, Web Analytics & social media interactions
Insights must go beyond polarity detection
Multilingualism beyond main European languages
Accuracy of the algorithms is a key issue
Segmentation of individual consumers
Big Linked Data
Integration of multiple heterogeneous & unstructured data sources at scale
Communication with consumer activated in real time
Meaningful recommendations beyond display advertising
Online-offline integration through mobile devices