B1 中級 7763 分類 收藏
開始影片後,點擊或框選字幕可以立即查詢單字
字庫載入中…
回報字幕錯誤
>> Predictive analytics has an incredible amount of potential
power and could lead to fantastic success, or your competitors
could leverage the principles to gain an advantage.
Learn the secret to managing your data to gain the competitive edge
next on "Modern Workplace."
[Music]
>> Prediction is the way of the future. Today we have access
to the roughly 5 exabytes of data created globally every single day.
That's the same as 1 billion gigabytes every day. But how can
this lead to success? According to Gartner, by 2016, 70 percent
of the most profitable companies will be using this data and
realtime predictive analytics to rapidly grow their businesses.
Those who don't start today risk becoming obsolete. Today's topic
is actionable insights, how to turn data into success. We are
delighted to welcome the best selling author of "Predictive Analytics,"
Eric Siegel. He says that perfect prediction is not possible,
but it's important to make your bets on predictive analytics
now in order to get value from them tomorrow.
Also with us today, we have the CEO of Aryng and author of "Behind
Every Good Decision," Piyanka Jain. She says that deciphering
data is not rocket science. Thank you both for being here.
First question, Eric, I'm interested, is BI, big data, that's
been a trend that most people are well aware of. Is predictive
analytics kind of the same thing as that, or is it different?
>> When people say big data, I'd say more than half the time they're
referring to a specific use of predictive analytics, but big
data is a much broader area. Predictive analytics is a subset.
Big data doesn't refer to any one specific method or technology.
It just means there's a lot of excitement because there's a lot
of data, there's a lot of things to do with it, a lot of value
to get.
>> Big data is just a pile of data whereas predictive analytics
is what you can do with it?
>> Exactly, right. So the data is what's so exciting about data,
the most valuable thing about it is that it's predictive.
It is a recording of things that have happened. It's a bunch
of experience sort of the collective experience of an organization,
from which it's possible to apply predictive analytics to learn
from it how to make predictions per individual. So in most cases,
predictive analytics is per individual, like per individual customer
or applicant for credit, this kind of thing.
>> The same sort of stuff as like the nudge theory, where you
can sort of predict what people are going to do based on previous actions?
>> Yeah, exactly. And the kind of prediction, as you said in the
opening here, is about putting odds on their behavior, whether
they're going to click, buy, lie, or die, or do something you
don't want them to do, commit an act of fraud or a crime.
>> One of the things we also said in the open, there was the Gartner
statistic of a lot of companies who are already using predictive
analytics today. Where are some people getting value out of this,
like right now?
>> Well, most large organizations are using it in multiple ways.
Marketing is a really hot area, which is just in terms of targeting
your direct marketing to customers more likely to respond or
targeting your retention offers to customers that are more likely
to leave.
The ultimate sales force and the most visible example of marketing
campaigns in this country are the U.S. presidential campaigns
for becoming the most powerful person in the world. And the
sales force are the armies of volunteers who go and knock on
the doors. Those are the sales calls. It's a marketing campaign.
They're selling a presidential candidate. That's the product
being sold. So very much the same type of marketing analytics,
predictive analytics, specifically, is being used by multiple candidates.
It was really a game changer in 2012 where the Obama campaign
used predictive analytics for the first time.
>> Now, that strikes me as interesting because we're sitting here
in 2015, almost 2016. We're well in the midst of another presidential
cycle, yet we're still highlighting something a few years ago.
Is it still the case that that remains the poster child? Or have
others really superseded what we saw there in that 2012 race?
>> Did you see yesterday's ad sorry to jump in.
Hillary's ads? Hillary Clinton has changed her entire advertising
game, and I think it's coming from the same analytics kind of perspective.
She now has four ads, which came out yesterday, and it's highlighting
four economic issues for women of certain age.
And the polls which have followed since then have already showed
a difference in her so I think it's powerful. That story will
read in 2016 for selection.
>> It sounds like it's already being employed. Eric, do you think
it's still are they still leading the game?
>> The thing is we haven't had another chance since 2012 because
we haven't had another presidential like there have been other
election campaigns done on the smaller scale doing this, but
the presidential elections where they really have the budget
and the oomph to do it. Hillary for America has actually advertised,
not just for predictive analytics staff, but more specifically
for what's called persuasion modeling, which is a specific advanced
form of predictive analytics >> You
mean micro targeting.
>> that the Obama campaign did. Now, all predictive analytics
you could consider micro targeting, but in this case, it's the
definition of what's being targeted, which isn't just who's going
to vote for us or who's not going to vote for us, it's who can
be persuaded. They call it persuasion modeling, also known as
uplift modeling.
>> So that's happening in the presidential campaign, and that's
all well and good, but how would you relate that to someone who's
sitting at their desk in a commercial organization today?
What are the things that they can learn and apply from that example?
>> Well, we see lots of case studies from large corporations using
this same type of technology. So predictive analytics, normally,
you're predicting who's going to respond in order to target an
offer or direct marketing, direct mail, what have you, and that's
the traditional use. That's the most common way to use it.
It's very valuable. You see time after time again, PREMIER Bankcard,
for example, reporting a savings of $12 million in their marketing
costs because they're defining that audience and they reach their
targeting that much more effectively. But then the next step
we're seeing in a smaller handful, but even more successful,
is where you don't just predict who's going to respond if I contact,
you predict, who am I going to influence to respond because of contact?
That is, they wouldn't buy otherwise. So you're literally predicting
influence or persuasion in order to optimize for persuasion.
>> Much more nuanced?
>> Yes.
>> Awesome. We're going to have to take a break now, but it's
a great time to let us know what you think. You can tweet me
directly @amcbg, using #modernworkplace. Next we're going to
take a deep dive into how you can start using predictive analytics
in your organization today.
>> Every day I'm flooded with new information e mail, calendar
events, social feeds. Keeping my head above water long enough
to find what I need used to be nearly impossible.
Until now. As I work, Office Graph used machine learning to map
the relationships between the content and people I interact with
all across Office 365, then uses what it learns to make my job
easier in all kinds of new ways. For instance, Outlook relies
on that intelligence to remove the clutter from my inbox so I
can focus on important e mails first. And whenever I need to
get ready for a meeting, Office Graph checks my calendar to learn
the topic, then pulls the latest content together onto prep cards
that I can review on the way.
And if I need a file from OneDrive while I'm there, I just type
a question to find it. No more fumbling around in folders.
Even the groups I work with benefit from Office Graph. It not
only tells us what's trending within the group, it also pulls
in relevant content and resources from outside the group so we
can be even more productive.
Office Graph intelligence also powers Delve, a brand new app
that presents and collects content from anywhere in Office videos,
links, documents, the works. It almost feels like the right information
finds me. So even though the daily flood isn't likely to slow
down any time soon, I'm not worried. Office Graph's got my back,
making sure I'm more proactive and better connected with my network
than ever.
>> In a moment, our experts will share some tips on how you can
start utilizing your insights. But first, if you're not already
registered, please check out modernworkplace.com. You can register
to receive reminders about future shows and get access to exclusive
white papers and other key information to help your organization thrive.
Please register at modernworkplace.com. Today we're focused
on how you can stay ahead of the competition using predictive analytics.
Piyanka, it's now time to get down to brass tacks. In your consulting
business, how do you advise customers to get started in this field?
>> There are two aspects to it. One is at the leadership level,
how do we start competing in analytics? How do we start leveraging
the troll of data that we are been collecting? Maybe we started
collecting day one. How do we start leveraging that to start
making decisions? That's >> How
do we do it?
>> How do we do it? So we start, we do that by what we call three
key questions to ask your data. We do that by not looking at
data and expecting the answer to pop up.
>> Not like The Matrix?
>> No, the data doesn't speak. It only responds. It responds
to intelligent questions.
>> That's an interesting way of putting it. It doesn't speak,
it only responds, but you have to know the questions to ask it.
>> Exactly. And there are many, many questions you can ask the
data, but where should you begin? Where you begin is by asking
the three key questions. The first one is how am I doing?
Which is another way of saying how do you measure your own success?
And for different organizations, it will be different. For some
companies it might be profit. Sometimes it might be shareholder value.
Sometimes it might be >> So
this is more about knowing your business and knowing what's important.
>> What's important for you. So that's the first part, which is
how am I doing? That's the first question. Second is what drives
how am I doing? What drives the success metrics? What drives
those metrics you define to be your KPI?
>> This is where you need to think about what are leading indicators
and lagging indicators.
>> All those, right, exactly. Leading and lagging are basically drivers.
Think of what do I really if I really want to have better margins
or better growth, what levers can I pull within my organization
to move that, to drive growth forward? So that's the second part.
What drives the success?
>> What do I actually care about, and what influences that?
>> Exactly. And the third one is customers. Who are my customers?
And how do I engage with them? So we all have hundreds and millions,
sometimes tens of thousands, different customer sets. How do
you do you understand your customers? What are their needs?
What do they want from you? How do you engage with them? So these
are the three key questions that you start with at your highest
level of the organization and start peeling the onion of your
it looks like three questions, but as soon as you have answered
them, there will be more questions coming up. Now, once you've
identified like, let's say you're going for profits or margin,
and you realize really it's the churn, the customer churn that
is killing us. Then the question is how do you reduce that churn?
And the question you may start with is our customer churn is
39 percent. How do we get it down to 35 percent? Once you have
a question defined at that level, you use a framework we call
BADIR, which is a five step method of really starting from defined
business question, then laying out an analysis plan, which is
the second step of B A in BADIR, collecting only relevant data
or pulling only relevant data, deriving insights, and making recommendations.
>> So just recap that, and then I'd love Eric's thoughts on how
to get started.
>> BADIR. Business question, what's the real business question?
Lay out an analysis plan, which has all the hypotheses and so on.
>> Business question, analysis.
>> Data collection. And then do your insights.
>> That's actually getting it, getting the data.
>> And then making the recommendations. So BADIR is the short term.
>> Eric, what are your thoughts?
>> So to build on what Piyanka just said, everything she just
said has to do with figuring out exactly how to position and
make the best use of predictive analytics, which ultimately means
what operational actions, decisions, treatments of customers,
whether it's regards to marketing or sales calls or customer
service, what operation is going to be improved with predictions?
So if it turns out, for example, you want to try to retain customers,
you're trying to figure out which ones are at risk of leaving
in order to target those retention offers. Which customer's
at risk of committing fraud, in terms of triaging the use of
auditors' time. So one way or another, it's improving mass scale operations.
This is where the rubber meets the road or what all organizations
are doing to treat and serve us. It's all the mass scale operations
in terms of deciding upon credit applications for a credit card,
marketing, fraud detection all these things can stand to be improved
because they're mass scale numbers games, and each individual
per customer decision can be driven by a per customer prediction.
>> Why do you think it is you mentioned marketing as kind of leading
the way.
>> Ahead of the game.
>> Exactly. Ahead of the game. Why do you think that is? Why is
marketing ahead of the game and, say, product development, just
picking it out of the hat, why not other parts of the business?
HR, for example?
>> So when you look at work force or you look at B to B instead
of B to C, the numbers are smaller. So it can be less of an
opportunity depending on the scale of the operation and the number
of corporate clients. When you talk about, as you said, product
development, what are the number of operational choices?
How many different products are you considering making? How many
different projects are you considering green lighting? If the
numbers are high enough, as they are in some cases, you can apply
the same thing because the defining characteristic of predictive
analytics is it's applying on that sort of micro level one prediction
per individual over many cases of people or of projects or of
products, this kind of thing. So there has to be a scale.
But where there's scale and the potential to optimize over mass
scale operations on that level of detail across many cases, that's
where you get the benefit of prediction. Prediction is the way
to optimize those things.
>> And as big data sort of has its effect on different parts of
the organization, this will become more and more prevalent, I
would imagine.
>> Exactly. This is where big data becomes most actionable.
So to clarify, there's two stages. The data comes in. You learn
from it to make the predictions, and then that predictive model
is applied. That's input for that one customer, one customer
at a time, into the model for the prediction for that one customer.
>> I think that's a particularly important point because sometimes
people think of it as, oh, this is just business intelligence,
and it's not just applying a sort of standard model on something.
With this volume of data, 5 exabytes created globally every single
day, there's no way the traditional models are going to help
us in that regard. And what would your tips be in terms of how
to help people understand how to get value out of this?
>> How to get value out of predictive analytics internally?
>> Yes. Is it about sort of investing in data? Is it about your leadership?
What is it in that sense?
>> So there are four aspects to analytics maturity.
Number one is data maturity. If your data is not mature, you
have unclean data. You have >> What
does that mean, unclean data?
>> Accuracy. Accuracy of data. And then there's data access.
How about you might have the perfect data sitting somewhere in
this server somewhere, which nobody can access. Or maybe only
two people, who hopefully they don't get hit by the bus, only
they know how to get the log file and all that.
>> So that's data cleanliness. So that one is data maturity.
>> First one is data maturity. The other one is leadership, leadership
being truly data driven. They're truly wanting to learn from
the data.
>> And making decisions based on it.
>> Open minded, not like data, prove what I want to be proven,
like I really believe in this, that position.
>> My gut says this, therefore, find this.
>> Exactly. Like I want to spread the business and charge the
customer twice and prove it. That's not what we're talking about.
So data driven leadership. The third aspect of having a mature
analytics organization is decision making process, having some
process of making decisions and then looking back. How did we do?
Did it work? Did it not work?
>> We sometimes term that here at Microsoft e corporate memory.
We're sometimes very bad at sort of taking a moment and looking
back because you can often sort of infer sort of these patterns.
But we're all sort of moving.
>> Like move, move, move, move. Q1, yes, Q2, everybody is looking forward.
How about looking backwards? And the fourth aspect is analytics
talent, which you need that for an organization to be successful.
>> So can you just sum those up again for us.
>> The four aspects of analytics maturity is data maturity, data
driven leadership, decision making process, and analytics talent.
>> This all sounds great, but it runs the risk of potentially
being a little bit academic.
If I'm sitting at work right now, what can I do to get started?
>> So specifically, you can start asking questions of your data.
So let's say you're a marketing manager, and you say last quarter
I spent 100K in my marketing effort, and I made $500,000 on this
product or this segment for my company.
Can I do better than that? That was a five times, 500 percent ROI.
Can I do better than that? So the question is that you might
start asking yourself is how do I improve the ROI for that?
>> So these are questions so you're not so much asking the data,
but asking of yourself with what you can do about the data?
>> So you start with a business question like we were talking about.
You start with a business question, how do I improve the ROI?
What's the best way for me to spend my 100K? Let's say that's
the question. And then you come up with hypotheses,
and you're not supposed to be coming up with hypotheses alone.
You come up with them together with your team members and say,
where do you think we'll get the best ROI? And some team members
may come together and say the hypothesis number one is really
the top buyer segment is fairly >> Collaboratively.
>> Collaboratively, you come up with the hypothesis and you start
with asking the really important question. Today my success
is ROI for my campaign. I have 100K to spend this quarter.
Where can I get the best ROI for my data? And then you use the
value framework and say, if this is the question, what are my hypotheses?
And you may come up with hypothesis number one, top buyer is
actually exaggerated. We can't get an incremental ROI from that,
but here's this group which might be actually going away.
They might be at risk for churn and things like that. So you
start with using the same framework that we've been talking about,
the value framework. But you start with asking the question and
use your value framework to get to your actionable insights.
>> Do you really think it's as simple, if I'm sitting there, what
can I do today, it's as simple as start asking questions?
>> Start asking the highest level question. So not just any question.
The question that you may come up with is like, oh, last quarter
we did spend 20K in optimizing our subject lines.
Should we spend 10K this quarter? That's not a high level question.
The highest level question, which starts, again, from using the
three key questions. How do you define success? What drives success
for you? And who are your customers? How do you engage with them?
So keeping that at that level, you start asking questions.
>> The same problems, just stepping back?
>> Yes.
>> Any thoughts, Eric?
>> That sounds about right. I think that by asking the bigger
questions, you're going to hone down on that exact application area.
What should be optimized? What operations, as they are existing
today, could be tweaked with individual predictive scores that
improve all those individual treatment decisions or actions?
>> And that transcends this is not just a marketing thing.
That actually transcends where you're using this.
>> So Piyanka gave in terms of a marketing example, but what we've
been saying applies predictive analytics is the same no matter
how you use it, the same basic data requirements and the idea
that you're predicting and using those predictions to improve
an operational decision.
>> All right. Final question, important one. If you were to give
organizations or people just one tip about predictive analytics
and insights, what would it be? Piyanka, I'll start with you.
>> I'll go by I'll borrow Albert Einstein's, when he said something
like keep it as simple as possible and no simpler.
Borrow that to your business problems. Solve any business problem
that you have at hand with as simple a methodology as you can
and no simpler, which means go and use simpler methodologies
first, and commonplace tools like Excel.
Use aggregate analysis, correlation analysis, sizing, trend analysis.
Use that before and exhaust it. Get the ROI from that before
you move on to more advanced predictive analytics linear aggression,
statistical aggression, decision tree. So that would be my condition.
>> Keep it simple?
>> Keep it simple.
>> Awesome.
Eric?
>> I would say that, in determining how you're going to use predictive
analytics, defining a product and initiative,
it's about defining what am I going to predict in order to help
with what type of operation? So is it in marketing, in credit
scoring, in the behavior of a website? Whatever that operation
that needs improvement, stands for improvement, whatever the
lowest hanging fruit within the realm of marketing, churn modeling
may be the lowest hanging fruit in a lot of cases.
>> Would it be fair to sum that up as sort of keep it simple,
start simple. Like almost pick the >> I
would say that, if you were going to list these two,
mine's sort of a prerequisite for what Piyanka's saying.
She's saying do it right in a way that's simple, don't overcomplicate it.
But what is it you're doing in the first place is the question
I'm interested in. So define your project and then follow Piyanka's advice.
>> Which is line with the BADIR framework I talked about, the
business question aspect, which is discussed in the book in chapter
3 and 4.
>> What's the right question?
>> What's the right question? Sometimes when I first say, I have
this problem, reduce churn, if you don't dig deeper, you might
not realize actually it's a different question underlying it.
So coming up with the right question and we've done this enough
number of times that we've worked with our clients that they
first come in, the e mail which comes in, we have this problem.
And then you go in for an interview, and it's like a completely
different problem.
>> That peeling the onion, sort of finding out exactly what the
issue is.
>> Exactly.
>> So two things I'm taking away, as not an expert in this sort
of area, but an enthusiastic individual.
First of all, I really enjoyed understanding the difference between
predictive analytics and its relationship with big data, how
the two are related, but this is different. And then for me,
this sort of light bulb went off when it wasn't just about taking
old models and applying it sort of in a linear form. The prediction
is that key part, and with that volume of data we can't hope
to ever get insight into that without that prediction form.
Thank you so much for being here. I really appreciate it.
>> Thank you for having us.
>> It's been great. Thanks, Alex.
>> Please, now it's your turn. Talk to us. Who or what should
be featured on future episodes? You can tweet me directly @amcgb
using #modernworkplace. Or take our brief survey at modernworkplace.com.
Up next, we will explore a new tool that makes collaboration
quick and easy.
[Music]
>> Windows 10 embodies what our consumer and
our enterprise customers are demanding.
[Music]
>> We'll be making available a free upgrade.
Cortana, we're excited to welcome you to the PC.
[Music]
>> The Windows Insider program gives us that connection,
an earlier phase than at any other time in the history of Windows.
[Music]
>> We want to move from people needing Windows to choosing Windows
to loving Windows.
>> Welcome back to "Modern Workplace." Today we are talking about
how the world of insights has enhanced how we can help and connect
better to our teams and our customers.
Please go to modernworkplace.com and get this month's special
white paper, capturing the $1.6 trillion data dividend.
This first of its kind study takes a deeper dive into data and analytics.
While previous studies have looked at the benefits of any point
investment in data and analytics over no investment at all, this
study compared the returns of a comprehensive approach with the
returns of making more limited investments in a range of data
and analytics capabilities. Check it out. On "Modern Workplace,"
we look at tools and innovations that drive business. Today we
are going to take a look at Delve, a new solution to keep up
to date with your connections, documents, and colleagues in an
easy way.
They're geared to tailor and display the most relevant content
for you at any given time from across your working environment.
What's considered relevant is based on you, who you work with,
and what topics your colleagues are working on. In addition,
the new experience provides an engaging and natural way to search
for content across any source from within a single view.
One way to catch up quickly on several high priority projects
is to use your personal home view in Delve.
There you can quickly see items that you contributed on your
personal page, boards to which specific content has been contributed,
or people with whom you've recently interacted. The second thing
I want to highlight is easily finding what you need.
Imagine you're about to embark on a new project and need some
extra help. In the past, you might have known who to talk to
about this, but that person has moved on. To find another expert
in this area, you can search for the topic, and you'll see content
on this subject from a number of authors. At least one of those
authors should be a good lead.
Lastly, it's critical to connect and engage. Delve helps you
learn more about your colleagues, their activities, and their content.
Of course, as your role, workload, and projects change, so does
the group of people and the topics that you interact with the most.
Delve automatically adapts to the reality of these changes.
By working with this productivity tool, you can increase your
team's connection and make collaboration easier and faster for everyone.
Please chat with me. You can tweet me directly by using my handle
@amcgb and please use #modernworkplace. Or go to modernworkplace.com
and take our brief survey. Next month we will explore how you
can harness the limitless potential in your organization.
I'm Alex Bradley. Thanks so much for joining us on "Modern Workplace."
Until next month, here's to making your business work better.
提示:點選文章或是影片下面的字幕單字,可以直接快速翻譯喔!

載入中…

預測分析 (Watch Actionable Insights to learn how to turn data into success on Modern Workplace (Ep 203))

7763 分類 收藏
Chris Lyu 發佈於 2015 年 12 月 6 日
看更多推薦影片
  1. 1. 單字查詢

    在字幕上選取單字即可即時查詢單字喔!

  2. 2. 單句重複播放

    可重複聽取一句單句,加強聽力!

  3. 3. 使用快速鍵

    使用影片快速鍵,讓學習更有效率!

  4. 4. 關閉語言字幕

    進階版練習可關閉字幕純聽英文哦!

  5. 5. 內嵌播放器

    可以將英文字幕學習播放器內嵌到部落格等地方喔

  6. 6. 展開播放器

    可隱藏右方全文及字典欄位,觀看影片更舒適!

  1. 英文聽力測驗

    挑戰字幕英文聽力測驗!

  1. 點擊展開筆記本讓你看的更舒服

  1. UrbanDictionary 俚語字典整合查詢。一般字典查詢不到你滿意的解譯,不妨使用「俚語字典」,或許會讓你有滿意的答案喔