In this video, we're going to talk about how to create
a successful interface with your data science team.
Your inputs to the data science team
are not that different than the inputs you would give to development team.
They are vivid descriptions of the customer journey and the customer experience,
probably more so with the product as it is and of course data.
Hopefully, coming from your own product or your own infrastructure,
your touchpoints with the customer because you can't do data science without data.
And back to you, the data scientists are going to try things and
give you inferences and maybe even some ideas on how to act on those.
You'll try those out and that's how you create
a robust interface here with your data science team, is through experimentation.
Because even though it's called data science,
it is like any science experiment.
You're going to try things that work out well and
move the needle for you and you're going to try things out that don't.
You need data to do data science.
So, one failure what you want to avoid is that,
you're not collecting any data off your product or your
touchpoint because that will make it really hard.
It'll create certainly fewer opportunities for data science team to
do something valuable for you.
My advice to you is to pay special attention to acting on
the inferences that data science team gives
you and iterating on those and trying them out.
Because even if early on,
they're sort of rough and you're unable to use them that well,
this is such an important area and
such an important learning opportunity for you and
your career that I think this is an area,
I would ascribe special importance to even if it's not on your A list at the moment.
The data science process we have in here,
we have questions of interest and an idea of what's the user experience,
how does the user interact with both the product and kind of like let's say
our support system or all the different touch points that we have with them.
And from here, we get insights and hopefully those are actionable for us and
then we're going to see if we can move the needle with those and we're going
to iterate through these things and try different stuff.
And one of the interesting things about
the practice of data science is you have to be
able to kind of move forward and move backward.
By forward, I mean the best way to get a good experimental result is to go
into it with a nice strong hypothesis and that is where these inputs come into play,
questions of interest, vivid depictions of user experience.
The way that you move backward is a little less obvious.
This is something in fact I've been learning about myself.
This is a case that we use here at Darden to teach data science called movie lens.
This is a plot of two factors.
Factor one is here,
and factor two is here and these are movies.
Like this says Dungeons and Dragons.
This says Jackass the Movie.
And this is a plot created by the machine intelligence about the variables,
the attributes of the movies that are most
likely to explain people's different preferences for them.
So, for instance, this might be a scale of you know happy to
sad and this might be scary to not scary.
I don't think that's actually what these things are but this is the kind of output
from your data science team that you might need to kind of work on with
them a little bit to understanding human terms,
to make actionable for next step and testing.
You need to make sure that you're collecting data off of your product.
Your data science team will help you with that but let's walk through a simple example.
Let's say there is this company Enable Quiz and they
offer lightweight quizzing solutions for companies that hire a lot of engineers,
so that the HR manager and the hiring manager can
assess whether somebody is really skilled in something,
really experienced in something or not.
And then through that they can assess the fit with the organization.
So let's say Enable Quiz,
we have inside the company,
we have this customer or user of the HR manager,
whose job it is to get candidates and
the hiring manager who is ultimately going to be hiring these people.
These two worked together inside the company and then we have job candidates
and we provide a quiz that they use to assess the fit.
So, what data might the product manager of this product collect from the customer?
Well, one idea is you know for forgiving these quizzes,
we might want to make sure that we're instrumenting and we
have ready access to the pass rates on the individual questions.
Because a really, really unusually high pass rate might
mean the question is too easy or too obvious and
an extraordinarily low pass rate might mean that the question
is too hard or poorly worded or just plain wrong and you want to fix it.
That one is an easy one because the instructional design team here,
we get this data from the learning platforms that we use.
What else might be interesting?
While this company is in the general space of skills assessment.
So, they might want to know that for
a given type of job that they're screening candidates for,
what skills do they select to put into their quiz.
Because let's say the quiz can have multiple skills in it.
Ruby on Rails, Linux systems administration, whatever.
If they collect that and they have data on the success rate of different hires.
They might be able to proactively suggest to their customers what skills and what kind of
quizzes they might want to make for
a given job description they could analyze and that would be really great.
Anything that makes the customer's job easier is really great.
If we go back to the Google Analytics example,
if Google Analytics can accurately predict
exactly what I have to do with my website to make it perform better.
Well, that's a much more valuable product than a product that need to dig
through a lot of analysis and make my own inferences by hand with.
And that's what so exciting about the future of data science among other things.
So, we've talked about how to create a successful interface with your data science team,
we've talked about some of the activities you guys might undertaking your collaboration.
Alex CowanFaculty & Batten Fellow
