dddddddddddddddddddddd

Digital marketing analytics
Second Edition

Chapter 13
Digital marketing
analytics

Copyright © 2022, 2018, 2016 Pearson Education, Inc. All Rights Reserved

Identifying Data Sources
Social listening tools—It offers insights into who your audience is, what content they
engage with and share, whom they interact with on a daily basis, and where you can
reach them by identifying which channels and platforms they are spending time on.
Search tools—One of the most popular search analytics tools is Google Trends, a.k.a.
Google Insights for Search, and it should be in your research plan toolbox not only
because of its usefulness but also because it’s free! Google AdWord Keyword Planner is
another option, especially if you are trying to assess keyword possibilities, affinities, and
overall volume.
Web Analytics—it is a critical data source to understand what user traffic has come to
your site, how they’ve engaged, how long they’ve spent on the site, and if they have
taken any of the desired actions It’s the most useful way to determine how your audience
behaves on owned media properties you may have.

Paid Media Data Sources— everything from television, to display advertising, to online
video, to native advertising, and pretty much every media format in between. Paid media
performance data comes from a variety of sources, including the publishers where
advertising was placed, data service providers (DSPs), and data management platforms
(DMPs).

Identifying Data Sources
Social platform insights—are great sources of data if you are conducting an analysis of
social media activation and how each of those networks has performed for the brand. – a
good complement to data and insights derived from reading social conversations
analyzed using social listening tools.

Traditional media monitoring tools— it can be an important tool in your toolset. Social
listening tools capture a significant number of online news sites, but tools such as Factiva
and Cision have even greater news media capture capabilities.
Third-party research tools—call on third-party data to help fill in whatever story you are
trying to tell.

Picking the Channels for Analysis
News—News outlets are the usual suspects that you should be familiar with by now. The
New York Times, Wall Street Journal, and USA Today are examples.

Blogs—Discerning a blog from a news outlet is becoming increasingly difficult these
days, but blogs are the most common channel identified for developing research plans.

Comments—Blog and news comments can also be a source of conversation data to
analyze, as well as fodder for the research plan.

Groups or forums—These are often closed networks of people who are talking about
single, or related, subjects online. A good example of a group or forum is
Babycenter.com, which is dedicated to all things “mom.”

Owned properties—Again, by owned properties we mean your corporate website or
blog. It is often one of the most important sources for understanding how your audience
behaves.

Picking the Channels for Analysis
Paid media sources—We mentioned earlier that paid media could be display
advertising, television, online video, or native advertising. If your campaign or company is
running any sort of paid media, and we’re imagining that they are, especially in this day
and age, it’ll be an important channel to analyze.

Facebook—Facebook is the most dominant social platform in the world with 2 billion
monthly users and is most commonly included in research plans. It’s important to
understand, however, that listening tools do not capture all available Facebook content,
only publicly posted content on their platforms.
Twitter—Twitter currently boasts more than 328 million monthly users and is one of the
most common channels analyzed, as brands of all sizes attempt to determine how it can
be best leveraged. It is also typically a large source of mentions for brands. One thing to
note, though, is that not every tool pulls in the entirety of Twitter mentions. Over the last
several years, Twitter has restricted access to its data to only a handful of providers.
Instagram—Instagram currently has more than 800 million monthly users due to the
rise in popularity of visual communications and sharing. Visual analytics tools or features
within social analytics tools like Sysomos or Brandwatch can be used to identify insights
from social data containing image engagement.

Picking the Channels for Analysis
Snapchat—Snapchat currently reports more than 255 million monthly users and is
heavily dominated by the highly coveted cohorts of millennials and Gen Z consumers.
Snapchat data that is available for analysis and reporting isn’t as robust as some of the
other platforms here, but can still provide a window into audience engagement, content
consumption behaviors, and affinities.
YouTube—YouTube currently reports 1.5 billion monthly users and is nearly the
largest search engine in the world, second only to Google. YouTube insights give video
content engagement data and social listening tools can help with the commentary and
discussion around video content.

Pinterest—Despite having the lowest monthly audience at 175 million active users,
Pinterest can be a valuable source of data for insights around image-based content and
sharing.
Reddit—Long ignored by marketers, Reddit has risen in potential importance due to its
thriving and highly engaged community of 234 million monthly active users. Analysis of
Reddit discussion can yield important indicators about consumer intent, perception, or
expectations.

Identifying Search and Source Languages
One of the reasons there is so much variability is that a lack of clarity
exists around what tools to utilize in what markets.
Resources in this case can refer to money and
human resources.
Capturing all these languages is not
imperative, certainly, but this list is a good
starting place if you are launching a global
analytics project.

In almost every case,
companies start with
global English and
expand from there.
Organizations with heavy
operations overseas need
to think about global
analytics.

Nailing Down the Research Methods
A critical part of a research plan is the research methods you will utilize
after the data has been identified and gathered.
Hypothesis—Do not worry about whether your science teacher is looking over your
shoulder and grading your hypothesis, but it is important to have some general statements
about what you expect to see in the analysis.

Time frame for analysis—Gathering all the available information on the Internet is not
practical, nor is it practical to gather information for all time. You need to select a specific
time frame to properly scope your analysis.

Project team—Not everyone in the organization is going to be involved with your
research project, so clearly defining roles is important.

Depth of analysis— According to an IDC report, the world is expected to generate 180
trillion gigabytes annually by the year 2025; this is up from 10 trillion gigabytes in 2015
(Forbes, 2016).2 Because of the volume of information, your analysis can be very granular.
Before you start analyzing the data, determine how deep you want to go with the project.

Nailing Down the Research Methods
It is a critical part of the process, obviously, but without a consistent set
of methods, the analysis or project can fall apart.
Coding framework—You should have a standard approach to coding mentions or pieces
of data as you are conducting the analysis. By coding, we simply mean a method for
categorizing or notating the topic mentioned in the post, the sentiment of the post, or the
location of the post (blogs, Twitter, forums, and so on).

Sentiment approach—How people are talking about your brand online (positive,
negative, or neutral) is important. Is it the most important metric? No, it is not that important
in every case. If you have a consumer brand, though, it might be very important to
understand. As with the coding framework, you should have a standard approach to
measuring sentiment; otherwise, the results could be inaccurate.

Spam/bot filtering—Because of the way the tools we have talked about in previous
chapters work (that is, using keywords to gather information on the Internet), you are likely
to capture mentions that you do not care about or that are spam. Before conducting the
analysis, your team should have a clear understanding of how to treat those mentions.

Developing a Hypothesis
A hypothesis is a proposed explanation for some kind of phenomenon.
unless you have outlined the behavior you are trying to analyze, your
hypothesis is incomplete. Try the following:
●
●
●
●
●

Preliminary research—You probably have a set of media-monitoring terms
laying around that you can pop into Google for some initial searching.
Gathering existing market research—See if you can obtain the volumes of
offline testing your market research team has already done. It can be a
valuable source for developing a hypothesis statement.
Interviewing marketing or communications colleagues—Some of your
compatriots in the communications or marketing functions might have some
knowledge based on work they have already completed.
Looking at existing paid media or website data—If your brand has a large
media or website presence, this might be a daunting exercise before
developing hypothesis statements.
Asking your online community—There is a very good chance that your
community will offer up an opinion if asked and that can serve as a good
behavior (or question) to test.

Time Frame for Analysis
Capturing all the available data is impractical.
Identifying the Project Team
Project leader/champion—This person does not necessarily need to be the one who does
the work, but she needs to be the person who assembles the research plan.
Research leader— The research leader is the one who ensures that the parameters
identified in the research plan are followed.

Analyst—the person who completes the research in conjunction with the research leader.
Research Quality Assurance (QA)—Someone on the team should be designated to
double-check the coding.
Content strategist/engagement leader—This person works hand-in-hand with the analyst,
research lead, and project lead to develop insights from the data. Without this person, all
that is completed is the collection of a massive amount of data.

Determining the Depth of Analysis
You can pick from four different methods of analysis for your project:
Automated—Many of the tools we have talked about in this book offer automated
dashboards that count mentions across a variety of potential metrics. This is not a very
desirable state, though, as the data has not been vetted for spam or checked for
relevancy.
Manual—Whether or not you decide to go with a manual process of reading and
analyzing every post depends on the size of the project and the resources available to
you. If you decide to analyze only content that mentions your brand, and there are not
many mentions, then doing the analysis manually can work.
Hybrid—Most companies take a hybrid approach, in which they rely on an automated
dashboard and supplement it with manual analysis.
Random sampling—This method utilizes all the data you can gather from the tool(s) you
are using and then randomly samples a selection of those mentions. How large the
sample is depends on the confidence interval you are comfortable using. The confidence
interval indicates how reliable your data will be.

Building the Coding Framework
Coding the random sample—assuming that you choose the random
sample method—involves applying a qualitative label to a much larger
post.
Media type—This basic tag assesses whether the mention came from a news, blog,
forum, Twitter, comment, video, or image site.
Sentiment—Sentiment coding is simply understanding whether a piece of content is
positive, negative, or neutral. Read more about sentiment in the next section of this
chapter.
Messaging pillar—Most companies have a set of messages they are trying to convey to
the marketplace. One of your tags should be which bucket that mention falls into.
Company spokesperson—This is an obvious yes or no tag to include in the analysis.
Type of post—Is the post a customer complaint, or is it a product mention? Could the
mention be categorized as an HR issue? Capturing the type of post helps you segment
and share the data with other parts of the organization, if appropriate.
Target journalist or media outlet—Again, this is an obvious tag, but you should be able to
sort based on the tag and see whether a majority of mentions came from target
publications.

Taking a Sentiment Approach
Online sentiment is still one of the most controversial subjects in the
digital analytics community today. The debate is centered around two
different core topics:

Automation versus manual—The social listening tools we have mentioned throughout the
book all have an automated sentiment-scoring tool. Unfortunately, those automated
sentiment-scoring tools are far from accurate—primarily because they have a hard time
discerning sarcasm from authenticity. However, manual sentiment analysis introduces
issues of human bias and scale that have yet to be overcome.

Value to the brand—It is important to note that online sentiment is not necessarily a
proxy for overall brand reputation. It is possible that during a period of crisis, the two
could be related, but it is not always the case. Some brands place too much emphasis on
online sentiment, whereas others do not look at it at all. The answer lies somewhere in
the middle of those two extremes to yield sentiment analysis that is actionable and useful
to brands rather than just interesting.

Taking a Sentiment Approach
The most common scale for online sentiment analysis is positive,
negative, and neutral.
Positive—The positive posts will likely be the most obvious. They are the posts that
advocate for the brand in some way or that complement an action the brand has taken.

Somewhat positive—A step below the positive mentions, somewhat positive can be
tempered endorsements of the brand. These mentions might mention the brand positively
but may do so only briefly in the course of a post.
Neutral—The neutral posts are probably the hardest to classify because determining the
intent from a casual read of the posts is often difficult. Furthermore, some posts say
positive and negative things in a period of a couple of paragraphs. Typically though,
these posts don’t advocate for the brand in anyway, and they likely just mention the
brand’s name.
Slightly negative—These posts often use negative terms in association with the brand,
but one of the key differentiators between slightly negative and negative is that slightly
negative posts do not focus on your brand exclusively.
Negative—Negative posts focus exclusively on the brand and are hypercritical of its
actions, behaviors, or messaging.

Filtering Spam and Bots
The final portion of your research plan should be an outline of how to
deal with spam and bots.

Spam and bot filtering is only one part of the process. The other part is factoring in news
or press release syndication.

Modelling marcom value
Chapter 16
This is an analytic model that gives to marketers and
advertisers a quantitative way to value a number of media
vehicles.

Single and Multiple Equation Approaches
● Past chapters have showed PDLs, and elasticity and time series
analysis applied to marcom valuation. Those were modelled using a
single equation.
● Media mix modelling is far broader and far more general. All of the
above techniques can be incorporated but using multiple equations.
That’s why we get the big bucks.
● The results of MMM will be the ability to tell marketers and advertisers
the impact of media on revenue, taking into account a simulation of the
whole marketing spectrum.

Adstock models
It is a simple transformation to include a decay in the variable of
exposures

Single equation and PDLs
Most analysts simply include PDLs as independent variables.

UNITS = a + B1(direct mail PDLs) + B2(email PDLs) +
B3(TV ads PDLs) – B4(price) + B5(seasonality) +
B6(promo events)

Simultaneous equations
A multi−equation approach to MMM is more powerful and more
realistic.

Review the Business Case

Purchase answer to see full
attachment