it’s our honor and privilege for me to
join this company you nearly were about
seven months ago or nine months ago now
one hundred and sixty one thousand
employees 190 countries and if you look
at our product suit we are okay that one
ought to in every markets we operate so
talk about the amount of data that this
company generates internally you can see
some stats here I think we generate
about also if you look at the way we run
our company so far we start with data
everything starts with the data so
that’s where you know as we move towards
looking at AI mission learning deep
learning it’s really imperative how we
build some foundation capabilities
leveraging data so if you look at look
at our approach to technology in
Unilever we have many technologies like
any large corporations thirty technology
platforms we literally run about 100,000
I to use as we support on a daily basis
these are white collars we use our
technology platform run about 28
petabytes of data today growing
exponentially on an average of few
petabytes every quarter and as you said
we have about 190 countries but
definitely a kind of about 900 connected
size with our factories with our users
except etc so the technology is at the
heart as a DNA of the company to run
this and when you have the technology
the data follows through it so what’s
our approach to technology I think every
digital transformation starts with a
message or figure you can’t just do
technology for the sake of technology
right you got to understand what is the
value of the technology that brains are
drives the company whether it is the
top-line growth or the bottom line
savings exit trap
so in unity well it is always toss with
purpose right we are everything that we
do is trust with purpose and what it
means to our end users and consumers and
customers and build on that to build
define what products and innovations you
want to bring to the market be it in our
personal care products a home care
products or or food and refreshments
capabilities and build that on top of
that without culture and skills you
nearly where has a very unique culture
being in operating 190 countries
sometimes it feels like 200 cultures but
there is an underlying theme that drives
that this is where again going back to
our purpose and value our technology
operations is really larger
we talked about and the platform is
pretty big but bottom line is as we said
digital platform is all about what is
the value the underlying technology
means to the table not about technology
flagging the DA in a tail light in the
dark right we don’t want to be in this
space we want to look at technology as a
foundation as capability to drive the
value for the company so many of you
might have seen this many of you know
that the word artificial intelligence
were coined in Dartmouth 1956 I think
the vision from that Dartmouth
conference is all about how do you
replicate human beings right we saw the
first incarnation about humanoids or the
2001 space so it is the Hal computer
that you have seen or the even the
Terminator movies where the Terminator
characters etcetera but I think you know
they are also science fiction right I
think they’re still far away from
getting those things into reality even
today’s theme about 2030 where my guys
talked about how we’ll be there in about
twenty thirty to get that parity but I
think we have old us food along quite
strongly from where we started in 1956
about artificial intelligence to more
grounded realistic about instead of
replicating human brain how do you bring
that thinking process to the world into
the machines leveraging through machines
with evolution or machine learning and
deep learning I think you know we had in
a very good spot to tackle that 2030
problem that the stage if you are
talking today I’ll come back to this at
the end of the speech to reflect back on
why now right because we’ve been talking
about this 50 years actually 70 years
and I’m sure we’ll talk about this for
the next 50 years
but why this has become extremely
critical and even now become pervasive
the large-scale systems we built our you
know traditional BI capabilities but BA
capabilities only gets you as far as you
can get right it helps to run the
company it only tells you what happened
what might have happened it doesn’t tell
you how to optimize how to predict for
the future if you look at the data sets
like you’re a lot as you said 30
platforms that generates 22 petabytes of
data but they are all very silos right
trying to bring any sense out of the
data across the systems it’s been very
difficult for
and also given large corporation we have
a country silos you have functional
silos etc and there is a data ownership
within those silos bringing them all
together it is a journey for us
and equally data are sitting in
different systems disparate systems lack
of proper interfaces etc getting access
to data is really painful for us and
most importantly I like any BI systems
you might have seen people start
questioning the reports it generates
because if the data is if you’re not
sure of the results that you are looking
for if the first instinct is going to
question that data is oh I don’t trust
data then you start building reports to
test your data and the report that you
produce and it’s a never-ending cycle we
are at that tipping point of
fundamentally changing the mindset about
how do you move away from instinctive an
approach to decision-making to data
driven approach I think this is where
the companies making significant
investment in actually not only creating
the cultural change about moving to data
driven approach and to help it migrate
to that approach build the foundation
with the technology that’s why we call
it a significant opportunity for us
looking at all these areas as a holistic
way I think you know most of you are in
the room must be very familiar
especially fear from IT or if you are
running an IT department right
traditionally it always started with in
the way back in the 80s and 90s are what
we call traditional bi or a technical bi
but if we need a report if need a piece
of information you call your friends an
IT and they will produce a report for
you and you know that continued for a
while and then people got frustrated as
the time lag to produce the report
started going longer or longer and even
became a monthly project to quarterly
project through annual projects and
sometimes they’re in our active history
is it takes years to get a report right
and we all seen that then technology has
moved on to produce more of a
self-service bi okay here’s a piece of
data I’ll give you access to data you
generate your own insights I think you
know where we are in the time of raisin
with the maturity of the technology and
the availability of the data we can now
we are a perfect tipping point to
democratize data and insights for
everybody in the organization
as I said 110,000 white-collar users in
our company imagine giving the power of
data in their hands that they can
generate their own insights by
themselves to do the job better I think
that’s where we are at and what are you
going to talk about a little bit today
is the idea of a data Lake how that is
going to help change this paradigm for
Unilever only if you heard the term data
Lake okay so I think that’s a
significant hand rice if other ways we
just start thinking if we’re talking
about what is a data late I think you’re
in simpler terms data Lake is an idea
where you bring in any type of data
shape size color whatever you want to
call it bring it into a one place where
you can start making sense out of it
whether machine learning deplaning
whatever you want to call it but having
in one place of all types of data I
think it’s a powerful paradigm shift
from where we had come from
traditionally we were used our data
warehouse technology where the data is
extremely normalized I think there was a
term that often goes with data
warehouses like Hotel California right
you can only put it in you can never
take it out you can never leave the data
warehouse so I think there is a idea for
data Lake unshackles us from the
constraints that we had about data
warehouses especially as the volume of
data explodes I mean looking at our own
data right 22 petabytes of data there is
no technology in data warehouse that can
hold the kind of a data and generate the
analytics that we are looking for so far
who approached you at beat Diagnostics
and descriptive analytics what happened
why it happened
etc but not about the future what can I
do with that decision etc so that’s
where the idea who data leak comes
really powerful we spent a lot of time
trying to figure out what is a data leak
in the you delivered construct is it one
leg is it multiple legs because every
country started asking ok I need a leg
for myself and our function started
asking our wound you know we did our own
leg then it kind of a defeats the whole
purpose of a data Lake right data link
has to be universal
if you want to democratize data and our
insights for everybody so that’s why we
took a very hard stance
you really were scale we’re going to
have one physical data Lake that’s going
to have all the data that’s evenly
whatever produces produce I’ll continue
to produce any other external data like
Nielsen’s Kantas all the external
sources social chatters that we want to
worry about our custom consumer behavior
you want to bring all those data into
this platform that’s why we named it
Universal data Lake there are two
important aspects of this data data like
first and foremost it is stored in this
native form because we want to make sure
the data that we receive from a
particular source hey there it is a
supply chain data or of twitter feed if
we change the shape of the data before
we processing it then the lineage is
broken and people would not trust the
data certamen sure that starting the
trust we want to start with that to
ensure that there is a linear that we
can insure so that we got the source
data we can start trusting the data
either your source is consistently right
or consistently wrong but at least we
can trust the data then the most
important point is we want to bring it
in here is the idea of time series how
deep you want if you look at the
traditional transaction systems they
store the data for two years some
countries it’ll be seven years depending
on the various country regulation but if
you want to run deep learning and
machine learning algorithms you need lot
more data the more the depth is better
in both cases in some cases will be
decremental but as much as we can
depending on the use cases we our time
series thing as deep as possible try
some cases we’re going up to ten years
of data so that’s the idea of universal
data laid then we said okay you got the
source data
I’m not generating any insights out of
it or I’m not even aggregated it
consolidated I’ve not done it yet
so that’s where the first processing
starts so introduce the idea of a
business data Lake again is an idea it’s
not a physical thing because it still
sits in the universal data Lake but it’s
where we start processing to a
particular function for example
marketing marketing may be interested in
a set of kaypea supply chain may be
interested in another set of KPIs or nd
may be looking in certain KPIs
so that’s why I’ve introduced idea of a
business data Lake again a logical comes
that sits on the physical literally now
that we have that done we have the
source data we have the kpi’s process
etcetera now we can serve up at once
analytics through our products be
through API is dashboards natural
language processing we have actually NLP
interface to some other analytics
recently we did some pure C’s to
demonstrate this capability they talk
about slow moving goods can we predict
slow moving goods before they become
slow moving goods very interesting idea
isn’t it right if you know that this
particular product is a slow moving
product before it becomes slow moving I
rather have I won’t make major
investment in the product so I think
what do you believe in this in this idea
is this is going to completely unlock
and create the democracy we talked about
for both data and analytics first then
the question is why now and why what are
some of the impedance ease for this
technology I think you know today we are
the crossroads with perpetual
scalability with a cloud providing the
foundation capabilities for the data
laid and the algorithms that in it can
run on the infinite scale that we now
have available through the cloud
computing I think you know the ml and
deep learning is now here looking at the
chart is not 30 years from now 20 years
from now Israel right now it’s available
at a price point that is economical for
all of us to democratize data and
insights but at the same time there are
a few things that’s going to be really
cumbersome for us
first and foremost porosity thrust right
as we being in more and more sources of
data if you don’t have a trust in the
data nobody’s going to trust the
analytics this comes out of it so one of
the things that’s keeping me awake is
how do I make sure that our data Lake
doesn’t become velum do Lake like a
swamp if you know you weren’t from
Bangalore bias right human beings as
they create the models the bias
naturally comes it breaks it people
missed it because there was a hidden
bias right I think there’s a joke about
who’s going to beans
well cup Germany everybody predicted is
Germany even though dot what Conference
created that algorithm there was
inherent bias that they won the World
Cup last time and they’re the favorite
they predicted those inherent bias that
keeps me up night and then the recent
one is data residency loss that’s coming
very various countries China Russia as
recently as Vietnam two weeks ago and
now India they’re all trying to create
boundaries in terms of where the data
gets stored purely for jurisdiction
purposes so that they can have access to
the data and that’s going to have an
impact on this strategy but we are
watching it very closely because most of
the data that we put in right now is not
personally identifiable data so you’re
okay but somebody becomes PII data we
will have an impact so just wrap it up I
think the idea of a data Lake is is a
huge unlock when you come from the
traditional be able to look at
predictive and prescriptive analytics
and eventually automate decision-making
in the future I think there’s a great potential with the relics thank you
join this company you nearly were about
seven months ago or nine months ago now
one hundred and sixty one thousand
employees 190 countries and if you look
at our product suit we are okay that one
ought to in every markets we operate so
talk about the amount of data that this
company generates internally you can see
some stats here I think we generate
about also if you look at the way we run
our company so far we start with data
everything starts with the data so
that’s where you know as we move towards
looking at AI mission learning deep
learning it’s really imperative how we
build some foundation capabilities
leveraging data so if you look at look
at our approach to technology in
Unilever we have many technologies like
any large corporations thirty technology
platforms we literally run about 100,000
I to use as we support on a daily basis
these are white collars we use our
technology platform run about 28
petabytes of data today growing
exponentially on an average of few
petabytes every quarter and as you said
we have about 190 countries but
definitely a kind of about 900 connected
size with our factories with our users
except etc so the technology is at the
heart as a DNA of the company to run
this and when you have the technology
the data follows through it so what’s
our approach to technology I think every
digital transformation starts with a
message or figure you can’t just do
technology for the sake of technology
right you got to understand what is the
value of the technology that brains are
drives the company whether it is the
top-line growth or the bottom line
savings exit trap
so in unity well it is always toss with
purpose right we are everything that we
do is trust with purpose and what it
means to our end users and consumers and
customers and build on that to build
define what products and innovations you
want to bring to the market be it in our
personal care products a home care
products or or food and refreshments
capabilities and build that on top of
that without culture and skills you
nearly where has a very unique culture
being in operating 190 countries
sometimes it feels like 200 cultures but
there is an underlying theme that drives
that this is where again going back to
our purpose and value our technology
operations is really larger
we talked about and the platform is
pretty big but bottom line is as we said
digital platform is all about what is
the value the underlying technology
means to the table not about technology
flagging the DA in a tail light in the
dark right we don’t want to be in this
space we want to look at technology as a
foundation as capability to drive the
value for the company so many of you
might have seen this many of you know
that the word artificial intelligence
were coined in Dartmouth 1956 I think
the vision from that Dartmouth
conference is all about how do you
replicate human beings right we saw the
first incarnation about humanoids or the
2001 space so it is the Hal computer
that you have seen or the even the
Terminator movies where the Terminator
characters etcetera but I think you know
they are also science fiction right I
think they’re still far away from
getting those things into reality even
today’s theme about 2030 where my guys
talked about how we’ll be there in about
twenty thirty to get that parity but I
think we have old us food along quite
strongly from where we started in 1956
about artificial intelligence to more
grounded realistic about instead of
replicating human brain how do you bring
that thinking process to the world into
the machines leveraging through machines
with evolution or machine learning and
deep learning I think you know we had in
a very good spot to tackle that 2030
problem that the stage if you are
talking today I’ll come back to this at
the end of the speech to reflect back on
why now right because we’ve been talking
about this 50 years actually 70 years
and I’m sure we’ll talk about this for
the next 50 years
but why this has become extremely
critical and even now become pervasive
the large-scale systems we built our you
know traditional BI capabilities but BA
capabilities only gets you as far as you
can get right it helps to run the
company it only tells you what happened
what might have happened it doesn’t tell
you how to optimize how to predict for
the future if you look at the data sets
like you’re a lot as you said 30
platforms that generates 22 petabytes of
data but they are all very silos right
trying to bring any sense out of the
data across the systems it’s been very
difficult for
and also given large corporation we have
a country silos you have functional
silos etc and there is a data ownership
within those silos bringing them all
together it is a journey for us
and equally data are sitting in
different systems disparate systems lack
of proper interfaces etc getting access
to data is really painful for us and
most importantly I like any BI systems
you might have seen people start
questioning the reports it generates
because if the data is if you’re not
sure of the results that you are looking
for if the first instinct is going to
question that data is oh I don’t trust
data then you start building reports to
test your data and the report that you
produce and it’s a never-ending cycle we
are at that tipping point of
fundamentally changing the mindset about
how do you move away from instinctive an
approach to decision-making to data
driven approach I think this is where
the companies making significant
investment in actually not only creating
the cultural change about moving to data
driven approach and to help it migrate
to that approach build the foundation
with the technology that’s why we call
it a significant opportunity for us
looking at all these areas as a holistic
way I think you know most of you are in
the room must be very familiar
especially fear from IT or if you are
running an IT department right
traditionally it always started with in
the way back in the 80s and 90s are what
we call traditional bi or a technical bi
but if we need a report if need a piece
of information you call your friends an
IT and they will produce a report for
you and you know that continued for a
while and then people got frustrated as
the time lag to produce the report
started going longer or longer and even
became a monthly project to quarterly
project through annual projects and
sometimes they’re in our active history
is it takes years to get a report right
and we all seen that then technology has
moved on to produce more of a
self-service bi okay here’s a piece of
data I’ll give you access to data you
generate your own insights I think you
know where we are in the time of raisin
with the maturity of the technology and
the availability of the data we can now
we are a perfect tipping point to
democratize data and insights for
everybody in the organization
as I said 110,000 white-collar users in
our company imagine giving the power of
data in their hands that they can
generate their own insights by
themselves to do the job better I think
that’s where we are at and what are you
going to talk about a little bit today
is the idea of a data Lake how that is
going to help change this paradigm for
Unilever only if you heard the term data
Lake okay so I think that’s a
significant hand rice if other ways we
just start thinking if we’re talking
about what is a data late I think you’re
in simpler terms data Lake is an idea
where you bring in any type of data
shape size color whatever you want to
call it bring it into a one place where
you can start making sense out of it
whether machine learning deplaning
whatever you want to call it but having
in one place of all types of data I
think it’s a powerful paradigm shift
from where we had come from
traditionally we were used our data
warehouse technology where the data is
extremely normalized I think there was a
term that often goes with data
warehouses like Hotel California right
you can only put it in you can never
take it out you can never leave the data
warehouse so I think there is a idea for
data Lake unshackles us from the
constraints that we had about data
warehouses especially as the volume of
data explodes I mean looking at our own
data right 22 petabytes of data there is
no technology in data warehouse that can
hold the kind of a data and generate the
analytics that we are looking for so far
who approached you at beat Diagnostics
and descriptive analytics what happened
why it happened
etc but not about the future what can I
do with that decision etc so that’s
where the idea who data leak comes
really powerful we spent a lot of time
trying to figure out what is a data leak
in the you delivered construct is it one
leg is it multiple legs because every
country started asking ok I need a leg
for myself and our function started
asking our wound you know we did our own
leg then it kind of a defeats the whole
purpose of a data Lake right data link
has to be universal
if you want to democratize data and our
insights for everybody so that’s why we
took a very hard stance
you really were scale we’re going to
have one physical data Lake that’s going
to have all the data that’s evenly
whatever produces produce I’ll continue
to produce any other external data like
Nielsen’s Kantas all the external
sources social chatters that we want to
worry about our custom consumer behavior
you want to bring all those data into
this platform that’s why we named it
Universal data Lake there are two
important aspects of this data data like
first and foremost it is stored in this
native form because we want to make sure
the data that we receive from a
particular source hey there it is a
supply chain data or of twitter feed if
we change the shape of the data before
we processing it then the lineage is
broken and people would not trust the
data certamen sure that starting the
trust we want to start with that to
ensure that there is a linear that we
can insure so that we got the source
data we can start trusting the data
either your source is consistently right
or consistently wrong but at least we
can trust the data then the most
important point is we want to bring it
in here is the idea of time series how
deep you want if you look at the
traditional transaction systems they
store the data for two years some
countries it’ll be seven years depending
on the various country regulation but if
you want to run deep learning and
machine learning algorithms you need lot
more data the more the depth is better
in both cases in some cases will be
decremental but as much as we can
depending on the use cases we our time
series thing as deep as possible try
some cases we’re going up to ten years
of data so that’s the idea of universal
data laid then we said okay you got the
source data
I’m not generating any insights out of
it or I’m not even aggregated it
consolidated I’ve not done it yet
so that’s where the first processing
starts so introduce the idea of a
business data Lake again is an idea it’s
not a physical thing because it still
sits in the universal data Lake but it’s
where we start processing to a
particular function for example
marketing marketing may be interested in
a set of kaypea supply chain may be
interested in another set of KPIs or nd
may be looking in certain KPIs
so that’s why I’ve introduced idea of a
business data Lake again a logical comes
that sits on the physical literally now
that we have that done we have the
source data we have the kpi’s process
etcetera now we can serve up at once
analytics through our products be
through API is dashboards natural
language processing we have actually NLP
interface to some other analytics
recently we did some pure C’s to
demonstrate this capability they talk
about slow moving goods can we predict
slow moving goods before they become
slow moving goods very interesting idea
isn’t it right if you know that this
particular product is a slow moving
product before it becomes slow moving I
rather have I won’t make major
investment in the product so I think
what do you believe in this in this idea
is this is going to completely unlock
and create the democracy we talked about
for both data and analytics first then
the question is why now and why what are
some of the impedance ease for this
technology I think you know today we are
the crossroads with perpetual
scalability with a cloud providing the
foundation capabilities for the data
laid and the algorithms that in it can
run on the infinite scale that we now
have available through the cloud
computing I think you know the ml and
deep learning is now here looking at the
chart is not 30 years from now 20 years
from now Israel right now it’s available
at a price point that is economical for
all of us to democratize data and
insights but at the same time there are
a few things that’s going to be really
cumbersome for us
first and foremost porosity thrust right
as we being in more and more sources of
data if you don’t have a trust in the
data nobody’s going to trust the
analytics this comes out of it so one of
the things that’s keeping me awake is
how do I make sure that our data Lake
doesn’t become velum do Lake like a
swamp if you know you weren’t from
Bangalore bias right human beings as
they create the models the bias
naturally comes it breaks it people
missed it because there was a hidden
bias right I think there’s a joke about
who’s going to beans
well cup Germany everybody predicted is
Germany even though dot what Conference
created that algorithm there was
inherent bias that they won the World
Cup last time and they’re the favorite
they predicted those inherent bias that
keeps me up night and then the recent
one is data residency loss that’s coming
very various countries China Russia as
recently as Vietnam two weeks ago and
now India they’re all trying to create
boundaries in terms of where the data
gets stored purely for jurisdiction
purposes so that they can have access to
the data and that’s going to have an
impact on this strategy but we are
watching it very closely because most of
the data that we put in right now is not
personally identifiable data so you’re
okay but somebody becomes PII data we
will have an impact so just wrap it up I
think the idea of a data Lake is is a
huge unlock when you come from the
traditional be able to look at
predictive and prescriptive analytics
and eventually automate decision-making
in the future I think there’s a great potential with the relics thank you
Please follow and like us: