YT YouTube2Text

YouTube2Text → Transcripts → How to Become a $200K Data Analyst in 2026 | Ex-Google

How to Become a $200K Data Analyst in 2026 | Ex-Google — Transcript

by Sundas Khalid · 4,302 words · 623 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00You're thinking about becoming a data
0:01analyst and you are scared AI is going
0:04to take the job before you even land it.
0:06Here's the truth. AI can already write
0:08SQL, generate Python, build
0:10visualizations, and even analyze data
0:12set on its own. So why would anyone hire
0:15a data analyst? I've spent about 12
0:17years in big talk as a data scientist,
0:19data engineer, and a data analyst at
0:21Google and Amazon where I earned top 1%
0:23performance rating. I have interviewed
0:25candidates at both companies and helped
0:28people around the world learn data
0:30skills. So I have seen this journey from
0:32both sides. By the end of this video,
0:34you will know exactly why and what you
0:36need to learn differently to make sure
0:38you succeed as a data analyst. The
0:40fundamentals of data analysis skills
0:42matter, but how you learn them and what
0:44you pair them with has completely
0:47changed with the rise of AI. I'm going
0:49to break this down into three things.
0:51the core technical skills you need, how
0:54to turn data into real business
0:55decisions, and how to actually use AI so
0:58it makes you faster instead of replacing
1:01you. Then I'll show you exactly how to
1:03turn this into projects and job. And
1:05yes, I'm going to be sharing resources
1:07as well. Let's get into it. The first
1:09thing you need to do is build a strong
1:11foundation in data. And I would start
1:14with SQL. You need to be comfortable
1:15working with databases and answering
1:18questions using data. Simple, right?
1:21Start with things like, let's say,
1:22select statements, filtering,
1:24aggregation, group by, joins,
1:26subqueries, and eventually window
1:28functions, but don't just learn SQL
1:30syntax. Try to actually solve problem
1:32with it. For [snorts] example, imagine
1:34you're working for an e-commerce company
1:36like I did for Amazon, and someone asks
1:38you which projects generated the most
1:40revenue last quarter. You should be able
1:42to understand the question, figure out
1:45where the data lives, write the query,
1:47and explain what the results actually
1:49mean. That's much more useful than just
1:52simply writing a SQL join. Trust me, I
1:54spent about 12 years in data science
1:57domain working as a data analyst, data
1:58engineer, and data scientist. And SQL
2:01never went away. No matter which company
2:04and which role I was at, SQL was core
2:06part of my entire workflow. If you're
2:09completely beginner to SQL, here are two
2:11of my favorite resources to start
2:13learning today. After you have learned
2:15SQL, you're going to move on to Excel.
2:17Excel is still useful because you will
2:19often work with business teams that use
2:21spreadsheets even when the company has
2:23more advanced data systems. Yes, 12
2:25years later into data analysis, I was
2:27still using Google Sheets, especially if
2:29I needed to do something quick and
2:31dirty. Uh you would be surprised how
2:33often you will end up using tools like
2:35Google Sheets or Excel. So, what exactly
2:38should you learn in Excel? You should
2:39learn how to clean and organize data,
2:41use formulas, create pivot tables, and
2:44build basic to advanced charts. Here are
2:46two of my favorite resources if you're
2:47completely new to Excel. Next skill is
2:50statistics. No, I don't want you to
2:52become statistitian. I want you to learn
2:54statics statistics in a way that you're
2:56able to uh understand your data and
2:59you're able to explain it and you're
3:00able to use statistics as an advantage
3:03to basically figure out what data you're
3:05working with. You should be able to
3:06understand concepts like averages,
3:08distribution, variability, correlation,
3:11probability, confidence interval, and
3:13hypothesis testing. One of my favorite
3:15subjects. Honestly, I need to do a video
3:16on this. And you should be able to take
3:18this statistics concept and be able to
3:20apply on actual problem that you're
3:23trying to solve and on the data that
3:24you're trying to work with. Let me give
3:26you an example. Let's say if two groups
3:28of customers have different averages,
3:29purchase values, you shouldn't
3:31immediately assume one group is
3:33performing better than the other. You
3:35need to actually understand whether the
3:36difference is actually meaningful. And
3:39that's where statistics will actually
3:41help you figure out if the two groups
3:43are comparable or if you need to like
3:45make some adjustment or find additional
3:47data set to be able to do that
3:48comparison. Here are some of my favorite
3:50resources for learning statistics. Now
3:52let's move on to something called
3:54Python. I think we all know what Python
3:56is. So the question is should you learn
3:58Python as a data analyst in the current
4:00job market and in the current age of AI?
4:04I think yes you need to learn Python. In
4:06my previous videos I have mentioned that
4:08you can kind of like skip on Python but
4:10as a data analyst in the current time
4:13frame Python is important especially for
4:17you to understand how Python works. Even
4:19if you're using help of AI you should be
4:20able to write basic Python code to
4:22understand your data set but also like
4:24work with your data set and do the data
4:26manipulation and analysis as well as
4:28well as like those pretty
4:29visualizations. Okay. But here's the
4:30difference. When you're trying to learn
4:31Python for data analysis, you're not
4:33going to learn everything. You're going
4:34to focus on libraries such as pandas,
4:37numpy and get comfortable with data
4:39cleaning, exploratory data analysis and
4:41basic visualization. You can also learn
4:43scikitlearn libraries later on. But in
4:46the beginning, I would say about 90% of
4:48the work that you need to do, you should
4:50be able to do with pandas and numpy. As
4:52a data analyst, you won't be writing
4:54Python code for production. you are
4:56writing Python code for probably your
4:58Google collab notebooks or your Jupyter
5:00notebooks to do the basic data analysis,
5:02create visualization and so on. So
5:05that's the depth and breadth I would
5:07like you to focus on. So at this point
5:09you have learned the core hard skills
5:12necessary for data analysis. I think by
5:14now you know that SQL, Excel, statistics
5:17and Python are important. The next
5:18things that I'm going to talk about are
5:20even more important now in the age of AI
5:23because as I said earlier in my video,
5:25basic data analysis can be done with AI.
5:28So where exactly do you bring value as a
5:30data analyst? The next part is going to
5:32cover exactly that. You can do all the
5:34data analysis, but if you're not able to
5:37turn whatever you find into a decision
5:39or into an action or a recommendation,
5:41your work is not going to be useful in
5:44the age of AI. So the next skill that
5:46you need to master is how to turn data
5:48into a decision. Now knowing how to work
5:50with data is only one part of being a
5:53good analyst. The next question always
5:55after the analysis, what do you actually
5:57do with that analysis? This is where
6:00data visualization, business
6:02understanding and communication become
6:05important. Now why did I say data
6:07visualization? Because when you go
6:09present your insights and your uh
6:11findings and your recommendation,
6:14visualizations are one of the best ways
6:16to kind of tell a story and basically
6:19explain why you are recommending a
6:21certain action. Okay, by now you have
6:23learned SQL, Python, Excel, statistics.
6:26The next skill that I want you to focus
6:28on is how to build dashboards. And for
6:30dashboarding, you can use tools like
6:32Tableau or PowerBI and learn how to
6:35build dashboard. Believe me, even in the
6:37age of AI when I was at Google,
6:39literally four months ago, we were all
6:42still building dashboards by hand
6:44because there is so much nuances
6:46involved in ter when it comes to
6:47building dashboard and especially if
6:48you're building a very very custom
6:50dashboard that doesn't have very clean
6:51data set that AI can easily pull and
6:53build a dashboard. You still have to
6:56know how to build those dashboards and
6:58how to basically present that data that
7:00actually helps businesses and leaders
7:02make decision. So for a dashboarding
7:04skill I would say the best resource is
7:06to do handson learning. You can download
7:09Tableau which has a free software and
7:12like start practicing. Take a simple
7:14orders data from Kaggle and like start
7:16building a orders data dashboard or
7:17wherever whatever your interest is. That
7:19reason I keep going back to orders data
7:21because that's what I enjoy working
7:22with. uh but for you you can pick up any
7:25data set and figure out what story and
7:27what recommendation you want to give
7:29with that dashboard or you want to
7:31monitor and then build that dashboard
7:33hands-on dashboard building is the only
7:36way I would recommend to learn dashboard
7:38learn dashboarding skills yes you can
7:40watch videos to learn but like you have
7:41to apply hands-on I mean any other skill
7:44that you learn you have to apply but for
7:45dashboarding I would say like hands-on
7:46is very very important okay now the next
7:49skill that I want to talk about is very
7:51very important especially in the wage
7:53API because as I said earlier in the
7:55video you can actually do basic data
7:57analysis with AI. So why is a data
8:00analyst needed? The reason data analysts
8:03are still needed because even when AI
8:06can do basic data analysis, a human who
8:10has the context, business domain
8:12knowledge, who has the judgment and who
8:14has the full product context and nuances
8:17of the business is going to be able to
8:20apply their domain knowledge and
8:22recommend uh basically a business
8:24decision. So when you're sharpening your
8:27analysis skills, when you're practicing
8:28your projects, always look at the data
8:31and the insights that you generate from
8:33a lens. Okay, what exactly am I going to
8:35do with this? If I'm presenting this to
8:37my vice president or director tomorrow,
8:39what recommendation I'm going to give to
8:42my leader that they take action at the
8:45end of the day as a data analyst, your
8:47job is to drive decision. And in the age
8:49of AI, this is going to become
8:51critically more and more important to
8:54for you to drive decision-m and for you
8:56to tell stories. And one example of that
8:59could be let's say analyze you're
9:00analyzing sales data and you found that
9:02sales in one region drop significantly.
9:05You're definitely going to root cause
9:06it, but you also have to give a
9:08recommendation. So you're going to look
9:10at why did sales decline? Where was
9:12there like any drop in customers? Was
9:15there a particular product that was
9:17affected? Um, and what should the
9:19business do about it? Like what is the
9:21next step? Now, the chances are AI is
9:24probably maybe able to root cause if
9:25your company has AI integrated in all
9:28over the places where it's possible, but
9:31the chances are it's going to be missing
9:33all the nuances and business context
9:35which you as a domain expert and um uh
9:39basically an expert data analyst will
9:41bring. So yes, AI can do the basic
9:44analysis, but decision making,
9:46recommendation, and the trust that
9:48you're going to have with your
9:49recommendation is not going to be
9:51replaced in the real world. Yes, a lot
9:54of the times the business domain and
9:56context comes from like your time at the
9:58company and your experience, but you can
10:01actually start building that muscle now
10:03while you're learning to become a data
10:04analyst. Let's say if you're working in
10:06marketing, learn the basic marketing
10:08metrics. If you're working in product
10:10analytics, understand things like user
10:13retention and conversion. You don't need
10:15to become an expert in every industry,
10:17but you should understand the business
10:20problem you are trying to solve.
10:25Now, the next skill which is also
10:27related to decision-m and recommendation
10:29is communication and storytelling. The
10:31only way you're going to be able to
10:32communicate your action, your
10:34recommendation is with good storytelling
10:36and good communication. Let me actually
10:38give you an example. Imagine you spent
10:40three days analyzing a data set and then
10:42present 20 different charts to your
10:44manager. You might have done excellent
10:46analysis, but if the important insight
10:48isn't clear, the analysis is not going
10:51to do anything. You should be able to
10:53explain here is what we found, here is
10:55why it matters, and here is what I would
10:57recommend to do next. That ability
11:00becomes even more important as AI makes
11:03some of the technical parts of the
11:04analysis easier. And I can't emphasize
11:06decision-m action recommendation enough
11:10because no matter in what world we live
11:12in, businesses are not going to trust AI
11:15blindly because it could cost a lot of
11:17money. That's where you as a human come
11:19into the play and play your role as a
11:22domain expert who also is an expert in
11:24the data analytics field. Now let's talk
11:27about the biggest addition to this road
11:29map which is AI. You shouldn't think of
11:31it as just another tool that you just
11:32add to your resume. You need to learn
11:34how to use it as part of your workflow.
11:37For example, you can use it to help you
11:40write SQL queries, explain errors in
11:41Python code, explore a data set, or
11:44identify patterns worth investigating.
11:46You can also use AI features built into
11:48analytics or BI tools to speed up parts
11:50of your workflow. But there is one very
11:53important skill you need here, which is
11:55being able to check what AI gives you is
11:58correct or not. If a AI tool can
12:00generate SQL query, don't just run it
12:01and assume that it's correct. Look at
12:03the joins, check the filters, make sure
12:05the C calculations are actually done
12:08correctly. The same thing applies to
12:10Python statistics and business
12:12conclusions. AI can give you a starting
12:15point, but you are still responsible for
12:17understanding and validating the
12:18results. And I think this is going to be
12:20one of the most important differences
12:22between analyst in the future. It's not
12:25simply can you just use AI. It's
12:27actually can you use AI effectively and
12:30know whether the output is wrong. You
12:32can also start looking at the repetitive
12:35parts of your work and ask whether AI
12:37can actually automate some of that. For
12:39example, if you are repeatedly
12:40downloading a report, clean the same
12:42data and update dashboard and send an
12:44email. Ask yourself, can any of this be
12:47done with AI? When I was working at
12:49Google about 4 months ago before I left
12:51my job, I was actually using AI in part
12:54of my every workflow possible. Uh I'll
12:56be honest, like the AI that I was using
12:58wasn't that great. uh because like
13:00Google is a huge company. It's an
13:02impossible for an AI model to like have
13:05all the context and like have like a
13:07full-blown rag model that understands
13:09everything. But it was good at certain
13:11aspects. It was really good at writing
13:13SQL, but I still had to validate it. It
13:16was able to quickly figure out where the
13:17data lives, but I still had to validate
13:19it. And the only way I could validate
13:21any of this is if I myself knew where
13:24the data lives and if I myself knew
13:26where the SQL lives. Now, I'll be
13:28honest. I didn't build any AI automation
13:31workflows when I was at Google because
13:33in my opinion, the AI integrations for
13:35data analysis weren't that good. Don't
13:38quote me on it, but I I'm sure like
13:40they're getting better now. It's getting
13:42better actually every day, but couple of
13:44months ago, I don't think they were
13:45there. I wasn't able to like automate
13:47anything um like that I would use on a
13:50regular cadence. A lot of the work that
13:52I would do is custom analysis because as
13:54a principle I would get pulled into only
13:56things that have not been done before.
13:58So a lot of it was like new work that I
14:00would have to do. Uh for those it's
14:02really hard to like use AI to automate.
14:04But I still use AI to like write SQL
14:06code and um sometimes I would do the
14:09analysis in Google sheet myself and then
14:11I will take a screenshot and then put it
14:13into Gemini and say can you write a
14:15quick one paragraph summary of what you
14:18see in the table. I think I just got
14:19lazier because I wanted AI to give me a
14:22starting point and then I'll like change
14:23it to write like me. So there's so many
14:25ways that you can use AI as a data
14:28analyst in your workflow. I'm definitely
14:30need to cover that in one of my upcoming
14:32videos. Let me know if you would be
14:34interested because I did a very detailed
14:36video for data scientist role of how I
14:38use uh Claude um as a data scientist in
14:41my day-to-day workflow and that is one
14:43of the most popular videos that you all
14:46actually love. I'm going to link it
14:47here, but I definitely need to do
14:49another one that focuses specifically on
14:52data analyst roles. Let me know in
14:54comments if you would be interested in
14:55that. And if you want practical advice
14:57about building career in data and AI
14:59from someone who has actually done the
15:02work, subscribe. That's exactly what I
15:04cover on this channel. Another skill you
15:06need to have is measuring whether AI is
15:08actually streamlining your work. You may
15:10be using more tokens, compute, and money
15:13without making the work faster. Here,
15:15for example, I have an AI agent data
15:17analyst connected to sales data. I can
15:19ask it questions like, why did revenue
15:20drop last month? And it analyzes the
15:22underlying metrics and give me the
15:24answers. But one correct answer is
15:26actually not enough. I need to know
15:27whether the agent stays accurate, fast,
15:30and affordable when other people start
15:32using it. So, how are we going to
15:33evaluate it? We actually going to
15:35evaluate it using Forge within Corv
15:37Arena, who is also sponsoring this
15:39portion of the video. So, here I'm
15:41comparing two configurations of the same
15:43agent. Agent X passes most evaluations
15:46while agent Y passes only 55%. So with
15:50Forge I can actually compare the
15:52response time, throughput, failure rate
15:55and cost. I can also use agent lens
15:57where I can inspect conversations and
15:59incorrect results. For example, on
16:01harder questions like which region and
16:03product contributed most to the revenue
16:05decline, agent X can answer while agent
16:08Y doesn't have enough context. And here
16:10Arya basically analyzes that evidence
16:12and shows me what to investigate next.
16:15Here the difference is how much relevant
16:17contacts each agent receives. So that's
16:19the problem I need to solve on my own
16:21workload. I can also benchmark the
16:22newest infrastructure and validate the
16:25relevant forge capabilities. Using Arya
16:28and working with coreweave AI expert, I
16:30can actually identify improvements,
16:32determine how far the workload can scale
16:35and estimate the cost before production.
16:37That's how you build smart AI workflows
16:40whether you work as a data analyst, data
16:41scientist, product analyst, data
16:43engineer and whatnot. Now looking at
16:45this, this basically is telling me
16:46whether the AI is actually saving time
16:48and whether the infrastructure can
16:50support more users. That is exactly when
16:53you become a real AI power user. You can
16:56check out core and apply for an arena
16:58pass using the link in description
17:00below. Now let's move on to the next
17:02topic. All right, we're not done yet.
17:04After you have learned the skills, now
17:05we're going to get to work because
17:07that's where you're going to start
17:08building your hands on projects. And you
17:11don't need 10 projects. I would much
17:13rather see three strong ones where you
17:16can clearly explain the problem you were
17:17trying to solve, the data you actually
17:19worked with, your findings, and what you
17:22would recommend based on that like your
17:24recommendation. So focus on three
17:26projects. Try to choose projects that
17:28reflect the kind of work that you
17:29actually want to do. One of my best one
17:31of the best advice that I was given is
17:33like look at the job descriptions and
17:35look at the people who are currently
17:36working as a data analyst at a company
17:38that you really want to work at and look
17:40at their portfolio like your you know
17:43like when people add on LinkedIn what
17:45they do sometimes they add like what
17:46exactly the type of work that they're
17:48doing there that's one of the best ways
17:50to basically work backward and kind of
17:53like build your project portfolio if
17:54there is a specific specific company
17:56that you want to target let's say Amazon
17:58or Meta you can actually specifically
18:00basically find data set on Kaggle and
18:02like do projects on that and kind of
18:04like build your portfolio for that job
18:06that you can apply to later. So and you
18:09also like in a way you're also preparing
18:11for the interview that you will
18:12eventually do. So be very strategic with
18:14your projects and prioritize them. Don't
18:17just think that you have learned the
18:19skill and now you can just go apply to
18:21jobs. No, you actually need to build
18:22hands-on projects before you actually
18:25start applying to jobs because that's
18:27how you're going to solidify your
18:29knowledge. Okay, one last thing I would
18:30say here is that don't just upload a
18:32GitHub to notebook or like your code to
18:34notebook. Actually try to explain the
18:36problem you were trying to solve. Show
18:38how you approached it. Highlight the
18:39insights that actually matter and your
18:41thought process and importantly explain
18:43why someone should do what some someone
18:46should do with these findings that would
18:47give recruiter much better sense of how
18:49you think and work as an analyst. Okay.
18:52The whole point of learning all these
18:54skills is to get the job. So once you
18:56have learned the skills and start
18:58building the projects, I would say in
18:59parallel start applying to jobs. I would
19:02also say that the data analyst job
19:03market is very competitive right now. So
19:05your goal should be continuously
19:07improved while you're applying. Prepare
19:09for SQL question, Excel question,
19:10statistics question, case studies and
19:12behavioral interviews. You can also use
19:14platforms like strata scratch or leak
19:16code to practice technical questions and
19:19problem solving. My favorite one is
19:21definitely lead code. I've recommended
19:22it to so many people. I don't have any
19:24partnership with them, but I just love
19:26love love Stratus Scratch. If I were to
19:29do an interview tomorrow, I would open
19:30Stratus Scratch and I'll go through all
19:32the free problems. I don't have a paid
19:34subscription. They there are some
19:35problems that are available on the free
19:38version. So, I would just practice SQL
19:40and Python there and have a friend to
19:42mock interview you and um do some case
19:45study questions where uh when I
19:47interviewed at Meta, like I think Meta
19:49is like a really good example. When I
19:51interviewed there, they would actually
19:52start with a it would be a coding round,
19:54but the coding round would also be
19:56inclusive of problem solving, logical
19:58thinking, metric definition, like
20:00everything in one question. Um, so like
20:02they build on top of each other. So like
20:04look at some of those interview
20:05questions and like have a friend mock
20:06interview you cuz sometimes when you're
20:08practicing on your own, you're not able
20:10to like tell what exactly you need to
20:12improve. But a friend or somebody else
20:14who is can actually give you more
20:16constructive feedback to when you where
20:18you need to improve. The important thing
20:19here is to not rush through a checklist.
20:22You want to reach to a point where you
20:24can take a real business question, work
20:26with the data and use the right tool and
20:28explain what the results mean. I know
20:30some companies such as Meta even allow
20:32you to use AI. So, this is where if you
20:35already know how to use AI to do your
20:37data analyst work, you will actually
20:39shine in that interview. And if you're
20:41worried that AI is going to replace your
20:43job, actually I covered this in a very
20:45detailed video. So I would recommend
20:47that you watch this video next where I
20:49talk about what exactly AI has and can
20:52automate and where human a data analyst
20:55is still needed. So if I had to
20:56summarize the data analyst road map I
20:59would focus on three things. First
21:02you're going to build your foundation
21:03with SQL, Excel, statistics and Python.
21:06Second, you're going to learn to turn
21:08analysis into business decisions through
21:11visualization, communication, and
21:13storytelling with a strong domain
21:15knowledge and business understanding.
21:17And lastly, learn how to work with AI
21:19effectively, not just to generate
21:21answers, but understanding, checking,
21:24and improving its output. Then use all
21:26of these skills to build a few strong
21:28projects and start applying to jobs.
21:30Yes, AI is changing what data analysts
21:32do, but that doesn't mean there is no
21:35place for data analysts anymore. It
21:37means the role is evolving and if you
21:39can combine technical skills, business
21:41thinking, communication and AI fluency,
21:44you will be much better prepared for the
21:46data analyst role in next coming years.
21:49I hope this road map was helpful. Let me
21:50know in comments what else you would add
21:52to this road map and if you have any
21:54favorite resources that you would like
21:56to share with other people. I hope
21:58you're having a beautiful day and I'll
21:59see you in the next video. Have a good
22:00one. Bye.

About this transcript

This page contains the full transcript of How to Become a $200K Data Analyst in 2026 | Ex-Google by Sundas Khalid, generated from the public captions YouTube serves with the video. The transcript has 4,302 words across 623 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.