text,start,duration
morning everybody David Shapiro here,1.199,5.58
with your weekly video So today we're,3.419,5.401
going to talk about the data flywheel,6.779,4.261
the number one concept you need to know,8.82,5.46
in the age of AI,11.04,5.64
right off the bat what is a data,14.28,5.16
flywheel in short the concept of the,16.68,6.2
data flywheel is a simple virtuous cycle,19.44,7.02
whereby you have users and you use those,22.88,6.219
users to get more data you use that data,26.46,5.219
to get better AI you use that better AI,29.099,3.96
to get better products and services,31.679,3.961
which in turn gets you more users and,33.059,6.481
more data and so on and so forth and on,35.64,6.18
into Infinity you get compounding,39.54,4.56
returns and exponential growth,41.82,3.54
so,44.1,3.18
um I had to make this on my own graphic,45.36,3.66
because I think most of the ones out,47.28,3.36
there are probably copyrighted and I,49.02,3.18
didn't want my video to get taken down,50.64,3.96
but if you just Google image search data,52.2,3.48
flywheel you'll see what I'm talking,54.6,2.34
about there's plenty of good articles,55.68,3.539
out there but let's Dive Right In,56.94,5.16
okay so history of the flywheel,59.219,4.561
turns out that flywheels are actually,62.1,4.92
like super old like 8 000 years old,63.78,4.8
um most of the time they're used for,67.02,3.54
pottery wheels sharpening sticks and,68.58,4.26
stones but basically it's this little,70.56,4.98
bit down here that heavy Stone right and,72.84,5.099
so what you do is you spin it and hold,75.54,3.72
some of that energy and it'll keep,77.939,3.841
spinning and you add more and more,79.26,4.92
energy and it speeds up over time so,81.78,5.339
that's kind of where it started uh but,84.18,5.82
the concept of the flywheel applies,87.119,5.64
today uh just as much as ever so,90.0,5.52
flywheels are part of engines uh,92.759,5.04
particularly large diesel engines but,95.52,4.2
also uh flywheels are really important,97.799,4.32
for things like race cars pretty much,99.72,5.039
any electromechanical device or chemical,102.119,5.761
mechanical device because it Smooths out,104.759,5.82
the performance and and helps conserve,107.88,4.559
some of that rotational energy some of,110.579,4.801
that power but in all cases they start,112.439,4.68
at zero and they have to speed up you,115.38,4.5
can even have really big flywheels act,117.119,5.28
as generators for when you lose power,119.88,4.019
so all right so now you know what a,122.399,3.781
flywheel is electromechanical device or,123.899,3.961
chemical mechanical device that stores,126.18,3.9
rotational energy sure you don't care,127.86,4.019
about the physics of it,130.08,4.86
where did the data flywheel begin,131.879,5.881
um so apparently the whole thing was,134.94,6.6
pitched to Jeff Bezos back in 2003 by,137.76,5.52
Jim Collins,141.54,3.0
um and then of course everyone's,143.28,4.02
familiar with the Silicon Valley or are,144.54,4.74
they in Silicon Valley anyways the the,147.3,4.86
the tech darling of Amazon right Jeff,149.28,5.7
Bezos has obviously been incredibly,152.16,6.18
successful and Amazon has pioneered and,154.98,5.22
championed all sorts of innovative,158.34,4.5
things such as agile microservices,160.2,4.44
continuous deployment continuous,162.84,4.86
Improvement the data flywheel idea is,164.64,5.34
one of the foundational principles of,167.7,4.259
Amazon's success,169.98,5.339
which is I don't understand why people,171.959,4.5
more people don't know that that's why,175.319,3.241
I'm making this video so now you know,176.459,5.64
um okay so generative AI is basically a,178.56,6.84
fire hose of data every inference that,182.099,8.581
you send to uh gpt3 gpt4 uh for image,185.4,7.86
generation every single thing that you,190.68,4.8
send and get back is more data it's,193.26,4.08
literally just a data producer it's a,195.48,4.14
fire hose of data and so you should,197.34,5.039
record every single inference or,199.62,5.6
generation for LM,202.379,5.461
image generators everything,205.22,5.439
save it all save the input the output,207.84,5.36
the context the metadata the parameters,210.659,6.0
everything so there is so much data,213.2,5.679
going on here and this is above and,216.659,4.14
beyond the existing data that you get,218.879,4.561
from like your insights and Telemetry,220.799,4.86
from uh from your applications in web,223.44,4.859
right because everyone by now is,225.659,5.281
familiar with the fact that like uh you,228.299,6.421
know linger time time on page uh all,230.94,5.82
those other data points are available,234.72,3.96
through things but now you have even,236.76,4.8
more data which is your product is more,238.68,4.38
directly related to your data your,241.56,4.02
products and services so there's lots,243.06,4.5
and lots and lots of data to consume,245.58,4.799
already but then there's even more data,247.56,5.819
to consume in the age of generative AI,250.379,6.721
so how did I use this so I use this to,253.379,6.301
go from 20 000 subscribers at the,257.1,4.379
beginning of this year to over 50 000,259.68,4.86
subscribers right now so I more than,261.479,5.341
doubled my subscriber base in just a few,264.54,4.68
months now obviously some of that is,266.82,5.04
algorithmic luck right some of that was,269.22,4.979
I noticed a trend and I made use of it,271.86,4.8
but that was consuming the data so let,274.199,5.22
me tell you how it wasn't entirely on,276.66,4.44
accident,279.419,3.961
for those of you who don't know YouTube,281.1,5.76
gives you a dashboard for creators and,283.38,4.92
some of the information that it gives,286.86,3.779
you it's it tells you one which of your,288.3,4.86
videos are doing best over time it tells,290.639,5.041
you which uh channels and other videos,293.16,5.34
are leading to yours so you can see okay,295.68,5.519
who are my peers and competitors I don't,298.5,3.9
really think of it as a competitive,301.199,3.361
landscape because usually users want,302.4,3.72
more of the content that you give them,304.56,3.48
so it's more like peers right we're all,306.12,4.26
offering similar similar content it's,308.04,4.439
like a buffet,310.38,3.9
um so it's like okay so they're they're,312.479,3.841
doing this they're doing that,314.28,3.479
um and there's a lot of information that,316.32,3.12
you can consume above and beyond what,317.759,3.541
the dashboard gives you so there is a,319.44,3.539
service out there called vid IQ which I,321.3,3.54
tried and I didn't like it,322.979,3.241
um I didn't I didn't go for the paid,324.84,3.72
tier but I was just like I saw like I,326.22,3.479
was looking at the dashboard and I was,328.56,2.699
like oh they're obviously using like an,329.699,3.84
llm to generate some recommendations I,331.259,4.201
was like I can do better than this so,333.539,4.261
what I did was I took the transcripts,335.46,4.86
from my own videos and I compared that,337.8,4.86
with the comments that I was getting so,340.32,5.4
I I took the transcript and comments and,342.66,4.62
I said what about my videos is,345.72,3.78
resonating with my users and so I use,347.28,4.919
chat GPT to distill down and said okay,349.5,6.06
this is what people like this is why and,352.199,5.94
then I use those insights to generate,355.56,5.46
um ideas for follow-up videos and to,358.139,4.681
really kind of shape my narrative and,361.02,4.98
delivery and it worked I got from 20 000,362.82,5.28
Subs to 50 000 Subs in just a couple,366.0,4.979
months that was uh how I really kind of,368.1,5.159
nailed the algorithm them by using this,370.979,5.041
data flywheel because more users more,373.259,4.921
subscribers meant that I got more data,376.02,3.959
right and the data in this case was in,378.18,3.92
the form of comments but also the the,379.979,4.801
numerical Telemetry right because you,382.1,4.659
put out more videos you see which videos,384.78,4.38
resonate which ones don't and I feed,386.759,4.621
that into the AI in this case chat gbt,389.16,4.379
which led to better videos better,391.38,4.56
transcripts or Scripts,393.539,4.801
which led to more users,395.94,4.8
um so that's just I did it by hand right,398.34,6.0
didn't just manual copying and pasting,400.74,7.32
of of raw data into chat GPT obviously,404.34,5.34
this is not the most sophisticated or,408.06,3.54
sustainable model but it's just an,409.68,3.9
example of how easy it is to get started,411.6,4.92
with the data flywheel,413.58,6.839
so with this concept of the data,416.52,5.82
flywheel the goal the business goal is,420.419,3.961
compounding Returns the positive,422.34,4.26
feedback loop The Virtuous cycle or most,424.38,3.96
conventionally The Snowball Effect,426.6,3.84
because the more subscribers you have,428.34,3.84
the better reach you have the more data,430.44,3.36
you've got and that's just for YouTube,432.18,3.9
this applies to every business out there,433.8,4.56
whether it's for your marketing team,436.08,5.519
your product team everything,438.36,7.2
so how can businesses do this obviously,441.599,5.22
not everyone out there is a YouTube,445.56,2.88
Creator so you're like Dave come on get,446.819,3.841
on with the show so there's a few rules,448.44,3.36
of thumb,450.66,4.259
um out there so one record everything if,451.8,4.5
you're not recording your data you need,454.919,3.241
to be recording it and not just the,456.3,3.119
output from the models you need to,458.16,3.539
record what goes in and out because,459.419,4.861
that's how you're going to get the,461.699,4.981
feedback later you also need to record,464.28,4.44
the metadata the metadata the context,466.68,4.26
and the parameters because that,468.72,3.72
information will be critical with for,470.94,3.84
making sense of the data later on number,472.44,4.44
two you need to actually use the data,474.78,3.96
there are so many companies out there,476.88,3.78
that have a lot of data that they just,478.74,3.899
never even look at so then they're,480.66,4.68
spending money on storage,482.639,5.221
and backups and they're not using it you,485.34,3.84
might as well just delete the data if,487.86,3.839
you're not going to use it number three,489.18,4.44
and this is the hardest part is make,491.699,3.181
changes,493.62,3.66
pretty dashboards and charts and graphs,494.88,5.099
don't mean a damn thing if you don't,497.28,5.22
actually change your behaviors so,499.979,4.62
sometimes that means you need to say hey,502.5,5.52
like this email campaign didn't work,504.599,6.421
let's just stop doing it or you know,508.02,4.8
we're we're totally missing out on this,511.02,3.86
sector we need to go after that sector,512.82,4.38
you can change your outward Behavior,514.88,4.0
which is how you engage with the market,517.2,3.839
and your customers you can also change,518.88,5.159
your internal Behavior so in the age of,521.039,6.36
language technology that can be changing,524.039,5.581
how you run and use meetings it can,527.399,5.461
change it can be how you uh use jira and,529.62,5.52
slack and teams and all that stuff,532.86,3.9
so you have to change your behaviors,535.14,4.139
your rules your structures change the,536.76,5.88
system right and if you don't do that if,539.279,4.981
you're if you if you just consume the,542.64,3.36
data and look at it but then don't make,544.26,3.36
any changes again you might as well just,546.0,3.6
delete it and I'll tell you why that's,547.62,3.659
the worst decision you can make in just,549.6,4.2
a minute and number four iterate you're,551.279,3.721
not going to get this right the first,553.8,3.18
time right you're going to experiment,555.0,4.32
you're going to have a lot of uh of uh,556.98,4.979
false starts a lot of damp squibs that,559.32,4.68
just it just doesn't hit but then when,561.959,3.901
you do nail the algorithm you'll go,564.0,3.42
you'll do what I did which is go from,565.86,3.18
twenty thousand Subs to fifty thousand,567.42,3.9
Subs in just a couple months and that's,569.04,4.62
that's honestly the dream of every tech,571.32,4.139
product out there right that's what chat,573.66,4.92
GPT did they kept iterating on GPT,575.459,4.921
technology until they found the version,578.58,4.68
chat GPT that went to 100 million users,580.38,5.82
in a couple of months,583.26,5.16
all right so I'm talking about iteration,586.2,3.9
and all this other stuff in data,588.42,2.94
flywheels and you've probably never,590.1,3.12
heard about either of these things but,591.36,3.9
the good news is for all you business,593.22,3.54
types out there there are a few,595.26,3.3
Frameworks out there that you can just,596.76,3.9
go read about and hire a consultant and,598.56,4.98
uh and say Hey I want to improve my,600.66,4.5
business processes both internally,603.54,5.0
externally uh customer product whatever,605.16,6.239
and you say I want to do this from a,608.54,4.54
data Centric perspective because this is,611.399,4.141
the age of AI so first one is agile,613.08,5.4
agile is a software uh development thing,615.54,6.18
which focuses on tight feedback loops so,618.48,5.7
that is agile is how you can do the data,621.72,5.84
flywheel for your your software teams,624.18,6.9
Kaizen is about iterative Improvement,627.56,5.399
more broadly across your whole business,631.08,5.16
Six Sigma is a very similar thing where,632.959,6.461
it focuses on reducing defects,636.24,5.7
um and and flaws in the process and then,639.42,5.039
finally I till the it infrastructure,641.94,4.38
Library,644.459,4.5
um is about service process Improvement,646.32,6.06
internally so all of these read a book,648.959,5.041
or two on it you'll see the common thing,652.38,3.899
the what these all have in common is,654.0,4.74
rapid feedback loops and those feedback,656.279,5.341
loops are often human-based processes,658.74,4.68
but there's no reason that you can't,661.62,4.74
start adding more data and consuming,663.42,5.159
that data with AI to get that feedback,666.36,5.099
and make those Behavior changes,668.579,5.281
all right so I know some of you guys are,671.459,4.021
hot on the biscuit to get going so for,673.86,3.719
all you business Chads out there like,675.48,4.32
cool your britches for a second,677.579,4.44
so while,679.8,4.68
a data flywheel is really critical like,682.019,4.141
if you want to survive in the age of AI,684.48,4.62
you absolutely 100 need a data flywheel,686.16,7.26
I guarantee you that uh 80 or 90 of the,689.1,6.66
businesses that fail in the age of AI,693.42,4.56
won't have a data flywheel that being,695.76,3.9
said this isn't this alone isn't going,697.98,3.66
to save you right,699.66,3.6
um doing the if you say like hey we need,701.64,3.06
to do a data flywheel before you even,703.26,2.639
have a product,704.7,3.54
you're missing the point you need a,705.899,4.341
decent enough product or service first,708.24,5.099
to prime the pump right if you if you,710.24,4.36
look at that graphic where you start,713.339,3.12
with more users right I've already got,714.6,3.6
users on my YouTube channel I earned,716.459,4.201
that the hard way right if your product,718.2,4.68
doesn't have any users,720.66,3.9
data flywheel isn't going to matter,722.88,3.42
because remember the data flywheel is,724.56,3.36
based on inertia it's based on momentum,726.3,3.719
if it's sitting there inert it's,727.92,4.8
actually going to slow you down more,730.019,5.461
um it's not an it's not a shiny new toy,732.72,5.64
it is nice but as a business person you,735.48,4.32
shouldn't even be thinking about it once,738.36,3.539
it gets set up let your teams handle it,739.8,3.599
because they're going to be the ones to,741.899,4.141
to know how to do it,743.399,5.161
um most importantly that concept all,746.04,4.5
these philosophies of iterative,748.56,3.54
improvements start with executive,750.54,5.46
leadership if you don't buy in on a on a,752.1,6.359
visceral level and understand process,756.0,4.32
Improvement and iterative Improvement at,758.459,3.481
an executive level,760.32,3.18
doesn't matter you might as well not,761.94,4.68
even do it all right so,763.5,5.7
in order to really power up obviously,766.62,4.14
eventually you're going to want really,769.2,5.4
fat fat data pipelines to feed your uh,770.76,5.94
flywheel right,774.6,4.799
there's a few mantras that you can adopt,776.7,4.319
and this is for executives all the way,779.399,4.081
down to engineers and product owners and,781.019,5.641
Architects data is the new oil it,783.48,5.159
baffles me how few people have heard,786.66,4.38
this term in the age of generative AI,788.639,5.161
data is the new oil,791.04,7.14
um live it learn it embody it embrace it,793.8,7.38
think about it talk about it another one,798.18,6.06
is slow is smooth smooth is fast this is,801.18,5.76
the idea of kick starting your data,804.24,4.98
flywheel slowly at first remember it's a,806.94,4.62
big heavy thing right all your products,809.22,4.559
all your services all your data it's,811.56,4.019
really cumbersome you're not going to,813.779,3.661
light this fire real fast you got to,815.579,4.2
start slow that'll be smooth but the,817.44,4.079
idea is that it cranks up over time,819.779,4.321
right it starts accelerating,821.519,4.921
garbage and garbage out for those of you,824.1,3.66
who are not familiar with machine,826.44,3.36
learning and data science if you feed it,827.76,3.54
a bunch of garbage data you're going to,829.8,3.18
get a lot of garbage results so don't,831.3,3.36
put pressure to say like we need to,832.98,3.18
consume all of our data I know that I,834.66,2.58
said,836.16,3.299
record all your data that doesn't mean,837.24,4.8
consume all your data right you have a,839.459,4.62
big pile of data you need to figure out,842.04,4.799
which signals to pay attention to,844.079,4.5
and that goes to the last thing which is,846.839,4.321
what gets measured gets managed this,848.579,5.76
actually comes from ITIL so ITIL talks,851.16,6.06
about things like manage your uh your,854.339,5.041
time to resolution your mean your mttx,857.22,4.5
mean time to resolution and a few other,859.38,6.36
things in your it service Portal same,861.72,5.58
thing can happen for customers right,865.74,3.48
whether it's an internal customer or,867.3,3.42
external customer,869.22,3.6
how long does it take to satisfy,870.72,4.14
customer demand how much does it cost,872.82,4.86
you those kinds of things if you're not,874.86,4.44
measuring the right things you're,877.68,4.019
completely blind and this these are all,879.3,3.899
some of the mantras this is not an,881.699,3.301
exhaustive list but these are some of,883.199,4.14
the mantras that can go into building,885.0,5.639
that mentality of the data flywheel,887.339,7.56
so in the age of AI uh being where oil,890.639,6.541
is the new data stop hemorrhaging your,894.899,4.38
data or your oil because then you're,897.18,3.54
just losing money it's money out the,899.279,3.36
door and that's all there is to it I,900.72,3.179
believe yes that's the end of the video,902.639,4.7
all right thanks for watching,903.899,3.44