Name: How marketing mix modelling can boost your sales and profitability
Uploaded: 2024-04-12T15:49:36.335Z
Duration: 403.87
Description: Step into the Streamly studio with Hilary Borndahl, CEO of Mixed Analytics, as she teaches us about the intricate cogs that makes Marketing Mix Modelling a marketing strategist's go-to tool, and a favourite for transforming data into actionable insight.

MIMS MMM V2.mp4

MIMS MMM V2.vtt

English

{{00:00:00}}
If you are looking to improve

{{00:00:05}}
MMM is a top solution to do that.

{{00:00:08}}
I am Hilary Borndahl, the founder and
CEO of Mixed Analytics, and I'm here

{{00:00:13}}
to talk about Marketing Mixed Modeling.

{{00:00:16}}
MMM, my favourite topic.

{{00:00:36}}
MMM stands for Marketing Mixed Modeling.

{{00:00:39}}
Marketing Mix as a term was coined by
Professor Neil Borden at Harvard in 1949.

{{00:00:46}}
As the industry has evolved, marketing
mix modeling is now the technique of

{{00:00:52}}
understanding all of the levers that
can influence conversion for a brand.

{{00:00:57}}
That conversion metric could be a sale,
it could be a click on a website, it

{{00:01:01}}
could be an add to a cart, and it could
be a variety of other topics as well.

{{00:01:06}}
But it's understanding how
each element of the marketing

{{00:01:11}}
communications plan influences that
overall KPI that you're modeling

{{00:01:21}}
Marketing mix is a blend of human
touch and AI machine learning.

{{00:01:28}}
We like to say that it's a mix of art and
science and the science that we rely on

{{00:01:33}}
would be the statistical analysis, the
modeling work along with using all of the

{{00:01:40}}
data and the human element, the art of it
is looking at how to transform that data.

{{00:01:46}}
If you consider 20 different data
sets might be added into a marketing

{{00:01:52}}
mix analysis, each of those is going
to be coming in a different format.

{{00:01:56}}
And it takes a data scientist to
harmonise those data into the same square

{{00:02:04}}
database that you can run a model on.

{{00:02:06}}
Then that data needs to be transformed.

{{00:02:09}}
And what do I mean by
transforming the data?

{{00:02:12}}
If you have a television campaign
or a digital campaign in market, It

{{00:02:16}}
doesn't influence the sale just in the
week when the marketing is present.

{{00:02:22}}
It actually influences the
sale for the weeks that follow.

{{00:02:26}}
That's called the decay rate of
advertising or the media half life.

{{00:02:30}}
Those transformations have to go into
the model and it takes the artist or

{{00:02:35}}
the human to know which transformation
should be applied to which data sets.

{{00:02:40}}
Once all of that data is in the
model and it's running together, you

{{00:02:44}}
have to look at what is correlated,
the intercorrelations of the model.

{{00:02:50}}
It also takes a human data scientist
to evaluate which of those variables

{{00:02:55}}
should be prioritised over others
if they end up happening at the

{{00:02:59}}
same time in the same market.

{{00:03:02}}
And so while we can use AI and machine
learning to make our jobs easier in

{{00:03:07}}
terms of the data processing, I believe
that Marketing Mixed Modeling will

{{00:03:12}}
always be a tool that requires a good
portion of human expertise and artistry.

{{00:03:24}}
Model accuracy comes down to how many
data points you leverage in your model.

{{00:03:30}}
There is a term you can Google
called degrees of freedom.

{{00:03:34}}
And that is the number of independent
data points that are in your model and

{{00:03:39}}
the ratio of those to the dependent
variable, which is your sales.

{{00:03:43}}
And to build an accurate model, you
should start with a lot of data.

{{00:03:48}}
Data is available to most clients
in across industries for both the

{{00:03:53}}
conversion and the media channels.

{{00:03:56}}
Many of the platforms
globally have been set up.

{{00:03:59}}
to ingest the data directly into
the MMMs, which is very helpful.

{{00:04:09}}
The timing between marketing mix models
depends on the sales cycle along with,

{{00:04:16}}
or that corresponds to the industry
and the products that you're analysing.

{{00:04:21}}
It also should be married up
to the window of media change.

{{00:04:27}}
When can I make a change or influence
the overall annual marketing

{{00:04:32}}
budget and the different time
periods around media seasonality?

{{00:04:36}}
For example, if you have a highly
seasonal skew that sees most of its

{{00:04:41}}
volume in the summer months, you probably
want to wait until that sales season

{{00:04:45}}
is over to then analyse the changes.

{{00:04:48}}
And provide back the insights
ahead of next year's seasonal peak

{{00:04:58}}
Across business and industries, most
people in marketing roles can benefit

{{00:05:04}}
from the use of marketing mix modeling.

{{00:05:06}}
However, MMM starts with the data and
you have to have data that defines both

{{00:05:13}}
the sales along with all of the elements
that are in the marketing mix plan.

{{00:05:19}}
And certain industries have an easier
time than others to get that data.

{{00:05:23}}
And so while everybody can benefit,
not all industries are typically using

{{00:05:28}}
marketing mix to help inform the business.

{{00:05:35}}
With many countries going to cookie
less environments, there will be

{{00:05:41}}
limitations on tracking consumer
journeys online along the entire path.

{{00:05:48}}
And while customer journey analysis
is very, very interesting, there's

{{00:05:53}}
industries that frankly just can't
do that in a number of countries.

{{00:05:57}}
And so I see marketing mix modeling
actually rising as a discipline across

{{00:06:03}}
industries and businesses over the
years to come, while solutions like

{{00:06:07}}
multi touch attribution are going
to become more and more difficult.

{{00:06:11}}
There's also going to be an increasing
need to prove the value of marketing

{{00:06:17}}
dollars during recessionary times and
any time where the budget can be highly

{{00:06:23}}
scrutinised and come under question.

{{00:06:25}}
And so if you are looking to improve
your sales or improve your profit,

{{00:06:31}}
MMM is a top solution to do that.

{{00:06:34}}
And I can only see it becoming
more and more popular.

I am Hilary Borndahl, the founder and
CEO of Mixed Analytics, and I'm here

Marketing Mix as a term was coined by
Professor Neil Borden at Harvard in 1949.

As the industry has evolved, marketing
mix modeling is now the technique of

understanding all of the levers that
can influence conversion for a brand.

That conversion metric could be a sale,
it could be a click on a website, it

could be an add to a cart, and it could
be a variety of other topics as well.

But it's understanding how
each element of the marketing

communications plan influences that
overall KPI that you're modeling

Marketing mix is a blend of human
touch and AI machine learning.

We like to say that it's a mix of art and
science and the science that we rely on

would be the statistical analysis, the
modeling work along with using all of the

data and the human element, the art of it
is looking at how to transform that data.

If you consider 20 different data
sets might be added into a marketing

mix analysis, each of those is going
to be coming in a different format.

And it takes a data scientist to
harmonise those data into the same square

And what do I mean by
transforming the data?

If you have a television campaign
or a digital campaign in market, It

doesn't influence the sale just in the
week when the marketing is present.

It actually influences the
sale for the weeks that follow.

That's called the decay rate of
advertising or the media half life.

Those transformations have to go into
the model and it takes the artist or

the human to know which transformation
should be applied to which data sets.

Once all of that data is in the
model and it's running together, you

have to look at what is correlated,
the intercorrelations of the model.

It also takes a human data scientist
to evaluate which of those variables

should be prioritised over others
if they end up happening at the

And so while we can use AI and machine
learning to make our jobs easier in

terms of the data processing, I believe
that Marketing Mixed Modeling will

always be a tool that requires a good
portion of human expertise and artistry.

Model accuracy comes down to how many
data points you leverage in your model.

There is a term you can Google
called degrees of freedom.

And that is the number of independent
data points that are in your model and

the ratio of those to the dependent
variable, which is your sales.

And to build an accurate model, you
should start with a lot of data.

Data is available to most clients
in across industries for both the

Many of the platforms
globally have been set up.

to ingest the data directly into
the MMMs, which is very helpful.

The timing between marketing mix models
depends on the sales cycle along with,

or that corresponds to the industry
and the products that you're analysing.

It also should be married up
to the window of media change.

When can I make a change or influence
the overall annual marketing

budget and the different time
periods around media seasonality?

For example, if you have a highly
seasonal skew that sees most of its

volume in the summer months, you probably
want to wait until that sales season

And provide back the insights
ahead of next year's seasonal peak

Across business and industries, most
people in marketing roles can benefit

However, MMM starts with the data and
you have to have data that defines both

the sales along with all of the elements
that are in the marketing mix plan.

And certain industries have an easier
time than others to get that data.

And so while everybody can benefit,
not all industries are typically using

marketing mix to help inform the business.

With many countries going to cookie
less environments, there will be

limitations on tracking consumer
journeys online along the entire path.

And while customer journey analysis
is very, very interesting, there's

industries that frankly just can't
do that in a number of countries.

And so I see marketing mix modeling
actually rising as a discipline across

industries and businesses over the
years to come, while solutions like

multi touch attribution are going
to become more and more difficult.

There's also going to be an increasing
need to prove the value of marketing

dollars during recessionary times and
any time where the budget can be highly

And so if you are looking to improve
your sales or improve your profit,

And I can only see it becoming
more and more popular.

Make it Make Sense

Explainer

Step into the Streamly studio with Hilary Borndahl, CEO of Mixed Analytics, as she teaches us about the intricate cogs that makes Marketing Mix Modelling a marketing strategist's go-to tool, and a favourite for transforming data into actionable insight. 

Marketing insights

Marketing strategy

Data Collection

Data analytics

marketing mix modeling

data-driven decisions

Data

thumbnail_Hilary-Borndahl_503276_1685630778.png

xsmall_Hilary-Borndahl_503276_1685630778.png

Hilary-Borndahl_503276_1685630778.png

Hilary Borndahl

Founder & CEO

Borndahl kicks things off with an introduction to MMM, explaining how it works and what metrics are needed for this model.

Understanding marketing conversion with MMM

You need an array of datasets added into the model to have a good marketing mix. Borndahl goes into more detail on what those datasets could look like. 


Wide Datasets

Once everything is in the mix, what happens next? Borndahl walks us through reading the intercorrelations of the model and how to manage data accuracy.


Reading the model

Borndahl reveals how seasonality, media change, and the type of product you're analysing can all affect your MMM output.  


Seasonality and SKUs

With many countries going cookie-less, Borndahl shares insight on how the future of MMM may be affected. 