Valiotti Analytics

At Valiotti Analytics, I lead a team that builds powerful, modern analytics stacks for companies ranging from Fortune 500s to fast-growing startups across EdTech, FinTech, GameDev, eCom, and mobile.
We’ve delivered 40+ data-driven solutions that have:

* Cut infrastructure costs by up to 40%
* Accelerated reporting by 70%
* Boosted client revenue up to 25x

Our end-to-end services cover everything from data warehousing and pipeline engineering to BI dashboards and machine learning.

Tech Stack: AWS, GCP, Snowflake, dbt, Airflow, Kafka, ClickHouse, BigQuery, Redshift, Looker, Tableau, Power BI, and more.

What we do:
✓ Data Strategy
✓ Data Engineering
✓ Data Analytics
✓ Advanced Analytics 

Need to scale your data capabilities? Let’s build something powerful.

CEO & Founder

Founder & CEO, Valiotti Analytics | Better Data Analytics for Data-Driven Companies

🚀 Founded valiotti.com — the go-to data agency for fixing messy marketing analytics.
Top-rated on Upwork & Clutch. Trusted by growth-stage teams to build dashboards, clean up data chaos, and drive faster decisions. Partnered with AWS, Google Cloud, and Salesforce.

Ex-Head of Consumer Marketing at the largest eastern retail company.
I’ve been the one chasing targets with thousand-row tables, slow reports, and siloed data — I know the pain.

Since 2019, I’ve helped marketing and sales teams:
→ Scale revenue up to 25x
→ Save hundreds of hours on reporting
→ Get insights that actually drive action

Let’s talk: https://calendly.com/nick-valiotti

Nikolay Valiotti

The hill I’ll die on:
“Natural language → SQL” will never — I said NEVER — replace analysts.

Every few months someone announces the same miracle:
“We cracked it — business users can query data in plain English now!”

And every time I think:
Great. We solved the part of analytics that was never the problem.

SQL was never the bottleneck, and typing isn’t the job — thinking is.
Most companies don’t struggle because nobody knows how to write a JOIN.
They struggle because nobody knows what question actually matters, or how to connect it to the business reality they’re living in.

Give someone a “natural language → SQL” tool and here’s what happens:

→ They ask vague questions.
→ The model confidently answers the wrong thing.
→ And everyone walks away thinking the data is broken.

The truth is that if analytics fails, it fails long before the query — at the decision, the framing, the context, AKA the part AI can’t do for you.

This tech won’t replace analysts. It just exposes how much of analytics depends on real judgment, not syntax. If anything, AI is about to make analysts more valuable — because someone still has to think clearly.

→ Follow Nick Valiotti, PhD for the data takes that won’t waste your time.
♻️ Repost if you’re tired of seeing “natural language SQL” demos pretending to solve a thinking problem with a typing feature.

Nick Valiotti, PhD

Author of Your Fractional CDO — Order on Amazon | Founder @ Valiotti Analytics | Data Strategy & Consulting for Modern Teams | PhD & MA in Analytics

Hard pill to swallow:

Excel was never the problem.
Dependence on Excel was.

Excel is brilliant.
It’s flexible, familiar, and absurdly powerful for its size.
But here’s the trap: the better Excel works, the more you rely on it. And the more you rely on it, the harder it is to grow beyond it.

Every company reaches that moment:
→ one file becomes 12
→ 12 becomes 40
→ 40 becomes a folder called “final_v7_really_final.xlsx”
and suddenly nobody knows which version is the truth

This isn’t a tooling problem. This is an infrastructure debt problem. And it compounds silently.

In my book “Your Fractional CDO” (Amazon: https://lnkd.in/dtvr7Y9Y) I break down exactly why spreadsheets become dangerous at scale — and what to do before your dashboards, KPIs, and decisions start contradicting each other.

Because the solution isn’t to “ditch Excel.” The solution is to stop using it as your entire data strategy.

Here’s where companies should shift their focus:

1 — Centralize your sources. Stop exporting CSVs like you’re throwing confetti. Put your data in one place — Sheets, a small warehouse, whatever. One source of truth beats ten perfect reports.

2 — Assign ownership. If everyone is responsible for the data, no one is responsible for the data. Quality only scales when accountability does.

3 — Automate the pipes, not the panic. Connect tools directly. Even Excel works beautifully as the front end if the plumbing behind it isn’t a disaster.

4 — Measure less, decide more. 50 KPIs don’t make you data-driven. A small set of metrics that the business actually uses — does. Excel absolutely has a place in a modern stack. Just not as the backbone of an entire company.

Tools don’t create clarity. Frameworks do. Excel should absolutely stay in your stack — just not at the center of it.

So fix foundations before foundations fail you.
And get your copy of Your Fractional CDO — unless you truly enjoy living dangerously with formulas that break when a butterfly sneezes.

Hot take we need to normalize in 2026:

Data engineering is not a one-time project.

Because here’s the truth nobody wants to admit:
Pipelines will break.
Schemas will drift.
APIs will betray you.

That’s not incompetence, that’s the nature of living, breathing systems.
The job isn’t to build a perfect stack and vanish into a Jira ticket sunset.
The job is to own it, watch it, and fix it before the business even knows a fire existed.

Great data engineers understand:
→ Code will fail — your detection speed is the real skill.
→ Models will drift — your adaptability is the differentiator.
→ Requirements will change — your shipping velocity decides your career.

This is why “set and forget” always becomes “set and regret.”

If you want trustworthy data, you need:
1 — Monitoring that screams before Slack does
2 — Engineers who debug like surgeons
3 — A real process for data contracts, not a Notion wish

Because data is infrastructure.
You’d never build a bridge and say, “ok, good forever 👍”
Don’t do that with your pipelines either.

Repost if a schema change has personally victimized you at 2am.
Follow me Nick for unfiltered takes on data engineering, data strategy, and keeping your pipelines from spontaneously combusting.

This is the best dashboard I’ve seen this week.
Not because it’s perfect — it’s not.

But because someone clearly gave a damn.
Most dashboards feel like spreadsheets that went to design school and dropped out.

This one by Alice McKnight, MPH one actually tries to communicate.

1 — The layout breathes
No clutter. No panic. Every KPI has space to exist — to be seen, not shouted.
That’s rare. Most dashboards feel like PowerPoints after a bender.

2 — The hierarchy makes sense
Big picture up top. Details below. You can tell Alice actually thought about how someone reads, not just what they read.

3 — The numbers talk, not mumble
They’re big enough, clean enough, and balanced enough to be noticed without screaming “LOOK AT ME.” That’s confidence.

4 — Context lives where it should.
There’s a data source, update date, and even the developer’s name.
That’s how trust is built — through quiet transparency, not endless disclaimers.

5 — It’s not flawless — and that’s fine.
Some visuals take a second to decode.
A table assumes you know what you’re looking at.
But that’s the good kind of friction — the kind that says “almost there,” not “run away.”

6 — It tells a story, not a status.
You don’t stare at numbers — you read them.
Every element works toward one message: clarity.

Great BI design isn’t about decoration. It’s about earning trust.
Every chart, number, and title whispering the same story, in the same tone, to the same person.

If your dashboards can’t explain themselves to a stranger, they’re not finished yet.
Save this for your next design review

Fractional CDO | Founder @ Valiotti Analytics | Data Strategy & Consulting for Modern Teams | PhD & MA in Analytics

Best dashboard I’ve seen this week.

For one simple reason — it doesn’t scream, it breathes.

This one by Anastasiya Kuznetsova is one of those rare Tableau projects that feels considered instead of stitched together at 2AM.
Here’s what stood out to me:

1 — The layout is clean. Ruthlessly so.
KPI block on top, two neatly separated tables at the bottom — nothing fights for attention.
There’s air. Space. Respect for the reader’s eyes.
(You’d be surprised how rare that is.)

2 — The first KPI section is genuinely clever.
The distribution visual is tidy, intuitive, and lands the message fast.
Only gripe: red numbers on a red background = Where’s Waldo: Dashboard Edition.
Easy fix for an otherwise strong idea.

3 — The remaining KPIs? Thoughtful as hell.
Side-by-side visuals make the comparisons stupidly easy.

4 — The left table is the star of the show.
Not your typical range chart — more informative, more intentional.
Switchable dimension, clean values, no useless axes cluttering the view.
Honestly: love it.

5 — The right table is simple, pretty, and functional.
Bars inside rows always look good when done right.
Switchers add flexibility without blowing up the layout.

And here’s why dashboards like this actually matter:

A while ago, inside one company, we ran a deep-dive attrition study. Months of surveys, HRIS exports, interviews, features, models… everything. And the biggest predictor of someone quitting?

A sudden, unusually long vacation.

Not salary. Not performance. Not even manager relationships. Just a long break — the classic “I need to clear my head” PTO that HR loves to ignore.

Dashboards like this help you see those signals before they turn into resignations on your desk.

Overall? It’s polished, structured, and built with care.
A dashboard that respects logic, not just decoration.
Every element feels like it belongs — and that’s the kind of design that scales, teaches, and actually gets used.

→ Follow me for more breakdwons like this.

A very important message — and some of you need to hear it twice:
Stop downgrading data analysts into IT interns.

If your analyst is fixing printers, resetting passwords, or babysitting dashboards nobody opens… congratulations: you’ve just burned the most strategic hire in your business. Because analysts aren’t “report people”, neither are they back-office technicians, nor support staff for whatever broke today.

They are business operators — and if they’re not sitting with marketing, product, and ops, you’re wasting their entire potential.

When analysts are isolated under IT, here’s what you get:
→ Numbers with no story
→ Dashboards with no decisions
→ “Insights” nobody uses

But put an analyst in the room where decisions are made, and the whole machine shifts:

Marketing: “We’re launching a new campaign.”
Analyst: “Perfect. Here’s what worked, what failed, and where your budget will disappear if you’re not careful.”

That’s what an analyst is for: not maintenance, not admin — but IMPACT.
If your analysts don’t understand the business, they don’t produce value. And that’s on leadership, not them.

—

♻️ Repost if you’ve ever seen analytics trapped where they don’t belong.
→ Follow Nick Valiotti, PhD for more hard truths on building data teams that actually move revenue.

Author of Your Fractional CDO — Order on Amazon | Fractional CDO | Founder @ Valiotti Analytics | Data Strategy & Consulting for Modern Teams | PhD & MA in Analytics

“Data analytics is so difficult”

No it’s not, you just need to know these basics:

1 — ETL (aka “just move the damn data”)
Fivetran — Free up to 500K rows. Unless you’re Google, you’re fine.
Python script — For people who are either clever… or broke.
Goal: one place for your data. Not duct tape. Not Excel exports.

2 — Data Storage
BigQuery. That’s it.
Cheap. Scales forever. Doesn’t cry when you throw SQL at it.
Storage = pocket change.
Queries = fine unless you go full goblin mode.

3 — Dashboards That Don’t Suck
Make it visual. Make it clear. Make it clickable.
Tool doesn’t matter.
That’s how you get decisions, not headaches.

That’s the starter pack.
No enterprise fluff. No shiny $10K toys.
Just a stack that works.

Now, this has been a mega short rundown. If you want details, foundations, and basics, my book “Your Fractional CDO” is for you.

Order on Amazon: https://lnkd.in/dDzQC3MJ

It shows you how to:
→ choose metrics that won’t betray you 
→ build a stack you won’t rebuild in 6 months 
→ think about data the way operators think about business

If this post was the appetizer, the book is the full meal — without the enterprise-priced bill.

Most “dashboard problems” I get hired to fix,

have nothing to do with dashboards.

They start with a CEO wondering why Stripe says one thing, their BI tool says another, and the cohort table sits in the corner pretending nothing is wrong.
Then someone quietly admits they’ve been “double-checking everything in Excel… just to be safe.”

That was exactly the case with one of our clients at Valiotti Analytics. They were not fighting the dashboard. They were fighting the system underneath it.

Broken foundations.
KPIs built on guesswork.
Two databases that didn’t speak the same language.
Refund logic that erased churn.
ETL scripts nobody remembered writing.

By the time dashboards started disagreeing, the real damage already happened months earlier — inside the pipelines, definitions, and identity logic that nobody had time to untangle.

This week’s Datapreneur is about how we fixed it and why it mattered.
If you’ve ever found yourself reconciling metrics manually… or trusting your gut more than your dashboards… you’ll want to read this one.

→ Issue 24 — “What Really Happens When You Skip Data Foundations”: https://lnkd.in/dSm8Ydud

I’ve just released a book that will save you a lot of money.
It’s called Your Fractional CDO.

It’s a sort of 101 to data analytics for founders, CXOs, and anyone who participates in decision-making.

And while it explains technical concepts in plain language, the book also includes a separate section on data roles — for a reason.

Your first data hire can either accelerate your business or quietly waste months and money. Not because they lack skill, but because every role solves a different layer of the system — and there’s a natural order to building your team.

So here’s a quick rundown:

Data Engineer — pipelines, integrations, storage, identity stitching. They create the foundation everything else relies on.

Data Analyst — insight generation, investigations, decision support. They tell you what changed — and why it matters.

Analytics Engineer — models, transformations, metric definitions. This is where CAC, ROAS, LTV, retention, and other numbers become trustworthy.

Data Scientist — forecasts, experiments, advanced modeling. Only useful when the basics above are stable.

CDO / Head of Data — strategy, architecture, prioritization. The role that ensures data work drives ROI, not noise.

In the book, I break down each role in detail — what they actually do, when they’re needed, what problems they solve, and the risks of hiring them too early or too late.

And I also share a simple formula for assembling your first data team, based on the stage of your business and the kind of decisions you’re trying to make.

So if you’re hiring your first (or next) data person, read this book first.
Your Fractional CDO is out now on Amazon: https://lnkd.in/dh7xP34x

You hired a CDO and nothing changed,

because one person can’t fix an 8-person problem.

Most data disasters don’t come from bad talent.
They come from asking one person to do eight different jobs.

Here’s what it looks like when roles are wrong:
Your “data scientist” is building pipelines. Your analyst suddenly owns AI because “they’re good with numbers.” Your BI dev is trying to architect a warehouse.

Each department has its own dashboards, none of which reconcile.
Everything is busy… and nothing is reliable.

Here’s what it should look like:

Data Engineers move data from your sources into your database/warehouse — clean, accurate, automated. If they do their job well, you never think about pipelines again.

Data Architects decide how all systems connect, where data lives, and how everything scales.

BI Developers build the interfaces: dashboards, reports, visuals — they’re the ones who turn tables into something teams can actually use.

Data Analysts provide meaning. They look at the numbers, explain what’s driving them, and tell you what decisions those numbers support.

ML Engineers turn patterns into predictions. Churn risk, lead scoring, recommendations — anything that forecasts the future lives here.

This is the basic setup. AI adds a couple more roles in the mix like AI Engineers for LLM integrations, Prompt/LLM Engineers for optimizing the use if AI, and AI Product Minds to decide where to add AI at all, but that’s a whole new level.

Think of it like a relay:

One role prepares the data, another structures it, another presents it, another explains it, another automates it, another embeds it into product.

When each team carries the right baton at the right time, everything works.

And this is exactly where a Chief Data Officer comes in. A CDO creates the system: who owns what, what gets built first, and how every part ladders up to revenue, product, and growth.

Without that leadership, you get activity. With it, you get outcomes.
If your team is scaling and you want this system built properly, let’s talk → https://lnkd.in/dzGjJgWM

This is how you do dark mode —

without turning your dashboard into a rave.

Most “dark mode dashboards” just go black = done and call it innovation.
This one? Actually built for low-light decision-making.

Here’s why it works:

→ Strong visual hierarchy — clean cards, deliberate contrast
→ Red as the accent — great for night use (yes, your eyes love red light)
→ KPI cards feel alive — that legend toggle is chef’s kiss
→ Filters + toggles speak the same UI language — no drift, no noise
→ Waterfall on tab two — intuitive, readable, fast
→ Tooltips — finally doing something useful instead of restating labels

In short: dark theme done right.
Subtle. Functional. Designed to be used — not just admired.

Amazing work, Pradeep Kumar G!

Nobody’s using your dashboards.

And it’s your fault.

Here’s why it happens — and how to fix it fast (read on for a free cheatsheet):

1 — Trust isn’t there (your numbers ≠ their numbers)

They’ve been making calls off a hand-rolled sheet for years. You ship a shiny dashboard, skip real QA/reconciliation, and it doesn’t match their file. Trust drops to zero.

Fix: reverse-engineer their sheet. Document every calc, filter, and caveat. Run the dashboard side-by-side for a few cycles, show a reconciliation tab, and agree on a single source of truth. If their sheet is wrong, prove it with traceable logic — politely.

2 — It’s usable… for you, not for them

Excel/Sheets let them scribble notes, add helper columns, and drop quick charts. Your dashboard stripped those workflows away.

Fix: interview them. What questions do they answer daily? What shortcuts do they use? Recreate the flow: default filters that mirror how they think, prebuilt “answers,” and fast paths for the 3 decisions they make most.
And if they want a table... just do the damn table. 

3 — You ignored the stakeholder journey

That report doesn’t live in a vacuum. It’s forwarded to investors, pasted into decks, compared with finance, exported for ops.

Fix: map the journey (inputs → decisions → outputs). Who touches it, when, and why? Build for the full loop: export formats that survive copy/paste, drill paths for board slides, and snapshots that audit clean.

Do these three and adoption stops being a prayer and starts being process. Otherwise, you didn’t build a tool, you shipped art.

And if you want the exact template I use to run dashboard discovery sessions with clients — it’s yours, free → https://gum.co/u/nrnjlnqa
🤠 Save now, thank me later!

Data isn’t magic.
But if LinkedIn is Hogwarts… a lot of wizards walk among us.

Every day on LinkedIn, someone breaks down a concept, simplifies a workflow, or explains a tool in a way that suddenly makes everything click.

It’s collaborative, generous, and strangely close to what a modern Hogwarts would look like — minus the robes and floating candles.

Here’s how I see today’s Hogwarts Faculty of Data:

Transfiguration — Data Engineering
The ancient art of turning raw chaos into clean, structured data.
Professor: José Siles
He’s been posting DE spells for so long that half of our pipeline discipline is basically his homework.

Flying — Pipeline Optimization
Airflow, but make it aerodynamic.
Professor: Benjamin Rogojan
His content is basically: “Here’s how to keep your broom from crashing mid-air.”

Charms — Data Analytics
SQL, dashboards, insights — the everyday spells.
Professor: Donabel Santos
Her posts are the closest thing we have to Wingardium Levio-SQL.

Defense Against the Dark Arts — Data Governance
Keeping your company safe from metric inconsistencies, rogue definitions, and shadow spreadsheets.
Professors: Charlotte Ledoux & John Wernfeldt
Blocking curses like:
→ KPIs that contradict each other
→ Definitions that shapeshift every quarter
→ “We thought Finance owned this metric?”

Care of Magical Creatures — Data Leadership
Where analysts learn how to survive messy data, stranger decisions, unclear ownership… and the occasional fire-breathing timeline.
Professor: John Cook
His content is basically a field guide to keeping analysts alive:
how to ask the hard questions, push back gently, spot bad logic early, and protect your team from the kinds of “creatures” that wander into meetings pretending to be business requirements.

Muggle Studies — Explaining Data to Non-Data People
The noble work of translating dashboards into human language.
Professors: Morgan Depenbusch, PhD & Andy Kriebel
Their content is the bridge between:
“I don’t get this chart” → “Oh… THAT’S what it means.”
The Muggles absolutely love them for it.

If you were hiring at Hogwarts…
Who else belongs on the list? Tag them in the comments!

I wrote a book.

Unlike AI, it might actually replace me.

A few years ago I realized something obvious but uncomfortable:
There wasn’t a single book that explains analytics to business people in a way that actually makes sense to business people.

There were tons of technical books — great ones, even. But for a founder, a CMO, a COO, someone running a real business, most of that material was unusable.

So I wrote “Your Fractional CDO”.

It’s not a technical manual. It’s a decision-making manual.
Because most people still think a CDO is the person who “knows how to count things correctly.” While CDO is actually the one who:
→ Knows when a metric is lying
→ Understands the cost of fixing vs. the cost of ignoring
→ Picks the system that gets you to ROI fastest — not someday
→ Connects marketing, product, ops, finance, and sales into one coherent story
→ Prevents decisions that look smart in dashboards but kill you in reality

And this is exactly what this book helps you see: context, judgement, trade-offs; how business reality collides with data, and how to actually make decisions with the thing you worked so hard to measure.

It won’t promise you a “data culture.” It will help you solve the one problem slowing your business down right now. And honestly, that’s what most leaders actually want.

If you want to support me:

→ Grab the Kindle edition (it’s free if you’re on Kindle Unlimited): https://lnkd.in/dKbMg8fp
→ Or just repost this so more founders, CXOs and whoever might need it see it

And a quick note: print-on-demand edition goes live soon — hardcover + paperback, with reviews on the back from some of your favorite data creators: José Siles and John Cook. 

Designed to be highlighted, dog-eared, and carried around like a field manual.

Most dashboards fail for a simple reason:

They answer questions nobody asked.

We jump straight into charts, colors, and layouts… but skip the one step that actually makes a dashboard useful:
Understanding how people will use it.

Here’s the part so many teams miss:
A useful dashboard isn’t built around metrics. It’s built around decisions.
Before opening Figma, Tableau, Power BI, or anything else — ask your users:

→ What decision do you need to make faster?
→ What is the pain with your current process?
→ What do you consider a “good” or “bad” week?
→ Which slices, filters, or comparisons do you need every time?
→ Which metrics do you actually trust?

If you can get clear answers to these five things, the dashboard basically designs itself.
No more 12-widget monsters. No more “Can we add one more chart?” No more reports that look great… and get ignored.

Because the truth is simple:
Dashboards fail when they’re built for the tool, not the user. Dashboards succeed when they’re built for the decision.

If you want a shortcut:
I put together a Dashboard Interview Cheatsheet — the exact framework I use with clients to turn vague requests into dashboards that actually get used.

Happy to share it: https://lnkd.in/d_tq_kGb
Let’s make dashboards people actually use.

There’s just one difference between an AI product that fails and one that grows.
And it’s not the AI.

Right now we’re living through the noisiest tech bubble since crypto and dot-com.

Investment dollars are flying into AI companies not because they’ve built something groundbreaking, but because everyone hopes they might someday.

And in this environment, every founder eventually thinks: “Maybe I should build an AI product too.” So the market fills up with thousands of tools that all look identical. 

Some raise millions before they have a single real user. But here’s the part no one wants to admit: if you add AI without adding product discipline, your AI becomes a liability, not an advantage.

The products that survive this hype wave won’t be the ones shouting “AI” the loudest. They’ll be the ones that:

→ measure what actually drives acquisition & retention
→ understand real user behavior
→ iterate based on data, not vibes
→ know where AI helps — and where it breaks
→ build a feedback loop around the model, not treat it as a miracle

So AI doesn’t fix a weak product. AI exposes it.

This week’s Datapreneur issue dives into a real case where analytics didn’t just analyze an AI platform — it reshaped how the AI sells, behaves, and improves over time.

If you’re building AI right now (or funding it), this one is worth your time.
Full story → https://lnkd.in/dcNF3TWi

The hill I’ll die on:
Your company doesn’t need to be “data-driven”.

You need the one fix that removes today’s bottleneck. And make sure it doesn’t occur again.
Might be a report you rebuild every week, a KPI no one can calculate twice the same way, a funnel that breaks every time you need it, or a dashboard that looks great but tells you nothing. You name it.

You just want one thing: an exact solution to the exact problem setting everything on fire as you read this.

That’s why I wrote Your Fractional CDO. Not as a manifesto, but as a practical first step for founders who need clarity fast — and don’t have six months to waste wondering what they actually need.

Because once you peel back the noise, most “data problems” come down to a few simple truths: → Metrics break when no one agrees on their definitions. → Dashboards lie when systems don’t tell the same story. → Your first data hire can quietly save you a fortune — or quietly burn it.

The book exists to help you spot these traps before they cost you time, money, and momentum.

→ Get the book: https://lnkd.in/drHQsQgK
Because you don’t need a data revolution.
You just need your next decision to be right.

They say being a data analyst is about SQL and dashboards.
In reality, it looks more like this:

1. Drink a giant mug of coffee
2. Management asks to “fix something” in the dashboard
3. Spend 40 minutes decoding what they actually mean
4. Drink another coffee
5. Fix it
6. Hear: “It looked better before”
7. Inhale deeply, consider moving to a monastery
8. Three back-to-back standups that could’ve been one email
9. Breakfast at 4 PM — the most spiritual meal of the day

Data skills matter.
But caffeine and patience are the real tech stack.

Smash Repost if you’ve ever whispered “SELECT * FROM sanity” into your mug.

Some dashboards fight you.

This one actually wants you to understand it.

I spent a good half-hour playing with this insurance portfolio dashboard — and honestly, it’s one of the rare cases where you can feel the designer thinking.

Here’s what stood out immediately:

1 — The layout breathes.
Clear spacing, clean separators, nothing screaming for attention.
Just calm, intentional design — the kind that makes a CFO exhale.

2 — A tiny detail that changes everything:
The settings-style legend on the left. It’s subtle, elegant, and genuinely helpful.

And the wild part? Every chart using that arrow icon supports cross-filtering — and it takes exactly two seconds to figure out.
That’s good UX.

3 — KPIs with personality.
Each metric comes with its own deviation indicator — maybe slightly duplicated across the board, but it works.

Feels neat, approachable, almost “mobile-app clean.”

4 — Tooltips, source labels, developer credits.
This is what “designing for trust” looks like.
People forget that credibility is a UI element too.

5 — Benchmarking page? Thoughtful, but not perfect.
Some spacing inconsistency.
Sparklines don’t need Y-axes — they’re meant to whisper, not lecture.

But the idea of darkening the selected quarter? Love it. Not 100% intuitive, but intuitive enough.

6 — And then the BROKERS view hits you.
Super-light bars with subtle borders — aesthetically beautiful.
A different stylistic choice from the first sheet… but honestly, it works. Clean, airy, almost Scandinavian.

Overall? This dashboard has a soul. Simply because someone cared enough to make every interaction just a bit smoother than you expect.

If more dashboards were built this way, we'd have fewer “what am I looking at?” meetings and more actual decisions.

Great work, Ellen Blackburn!

Let’s be honest: running a marketing agency is like always being on trial. 
Every action and every dollar has to prove it wasn’t in vain. 

Good reporting can do that.
Bad reporting makes everybody look confused.

At Valiotti Analytics we worked with a marketing agency that was managing reports for 70 clients — all updated by hand.
Download. Paste. Fix. Repeat. By the end of each month, nobody trusted the numbers anymore.

They weren’t maintaining 70 reports, they were repeating one broken process 70 times.

So we tore it down and rebuilt it from the ground up:
→ automated data collection
→ standardized logic and definitions
→ one dashboard with secure, trackable client views

That’s when the magic happened: reports refreshed automatically, formulas never broke, and the team got 80% of their time back to actually think — not just maintain.

They didn’t just improve reporting — they improved how the whole agency learns.

If you’ve ever looked at a dashboard and thought,
“Cool… but what am I supposed to do with this?”
you’ll want to read this one.

Dive into the full story → https://lnkd.in/dyPVMC2M

🏄‍♂️ Dmitry Anoshin

Surfalytics

Surfalytics = Surfing + Analytics. Inspired by West Coast surfing spots and Pacific ocean vibes. Created with goal to help you start a new career in data analytics space, develop data engineering and analytics skills through coaching. It will teach you not just dry skills, but will keep your focus to deliver significant value to businesses in the analytics realm as well as help to get fair compensation for the work you're passionate about.

There are several key producs:

1. Membership in a community with the goal of landing the job, preparing for the interview and getting the best offer. Every month, we do a Data Engineering and Analytics project and discuss our progress on the journey.

2. Community support to
2.1. Land 1st job in data
2.2. Move from analyst to engineer
2.3. Promote from middle to senior and beyond

3. Free course based on my 14 years of experience in Data Engineering and Data Analytics

Data Engineering and Analytics Mentor

University of Victoria

Instructor | Cloud Computing, Analytics and Data Engineering

Rock Your Data

- Data Team building and solving bottlenecks in team delivery
- Building Data Engineering Solutions with Modern Data Stack
- Data Infrastructure and Data Ops (DevOps)
- Data Security and Compliance

Advisory Board Member

Microsoft

- Building Delta Lake on top Databricks and Azure for Microsoft Gaming
- Structured Streaming
- Self-Service BI

Data Engineering & ML

Amazon

- Designed and Implemented Cloud Data Platform for Alexa Quality Metrics
- Designed and Implemented ML platform into production for Customer Perception, Onsite Experience and Churn
- Analytics Modernization with AWS
- Data privacy and Security
- Tracked and Optimized solution cost

Locations: Vancouver, Cambridge, Seattle

Communities:
- Lead Amazon Tableau User Group with 2k+ users.
- Lead Amazon BI Tech Talk community with 100+ BI teams globally.

Packt

•	Author of "Azure Data Factory Cookbook" - Build and manage ETL and ELT pipelines with Microsoft Azure's serverless data integration service. Solve real-world data problems and create data-driven workflows for easy data movement and processing at scale with Azure Data Factory.
•	Author of "Tableau Desktop Certified Associate: Exam Guide" - This book helps to develop your Tableau skills and prepare for Tableau certification with tips from industry experts.

•	Author of "Tableau Cookbook 2019.x" - This book has over 600 pages and includes over 115 recipes that help you to master Tableau Desktop, Prep, and Server.
•	Author of "Mastering Business Intelligence with Microstrategy" - This book is intended to master the craft of Business Intelligence using the best BI platform and discover the full potential of Microstrategy 10 in order to increase performance and improve the user experience.

•	Author of "SAP Lumira Essentials" - Learn data discovery techniques, build amazing visualizations, create fantastic stories and share these visualizations through an electronic medium with one of the most powerful tools – SAP Lumira. 

•	Author of "Learning Hunk"​ - Explore, analyze and visualize Big Data in Hadoop or NoSQL data stores with powerful full-featured big data analytics platform Hunk. 

•	Reviewer of "Learning Tableau 2019"​
•	Reviewer of "Mastering Tableau 2019.1"​
•	Reviewer of "SAP BusinessObjects Reporting Cookbook"​
•	Reviewer of "Creating Universes with SAP Business Objects"​
•	Reviewer of "Learning SAP Business Objects Dashboards"
•	Reviewer of "Splunk 7 Essentials 3rd Edition"

Author & Technical Reviewer

Apress

Jumpstart Snowflake A Step-by-Step Guide to Modern Cloud Analytics. This book is about modern analytics in cloud data warehouse age. It covers cloud analytics fundamentals and highlights modern analytics requirements. The book helps you to start with Snowflake and explore unique features and architecture. and learn more about the modern analytics ecosystem. You will review the analytics solution architectures and migration/modernization scenarios.

Author

Wawanesa Insurance

Toolbox: Tableau, Microstrategy, Oracle, Erwin Data Modeller, Informatica PowerCenter

Data Engineering & BI

Forex Club

DI toolbox: Tableau, Amazon Redshift, Postgresql, Pentaho DI, Tealium, DMP, Klipfolio, Adobe SiteCatalyst, Google Analytics, Adjust

Data Engineer

Teradata

• Delivered technical concepts, designed & developed of BI/ETL/DWH, Splunk, Web Analytics, Campaign Management, Data Science, BigData Analytics
• Designed and Implemented BI architecture and developed the full life cycle of the Business Intelligence part of the corporate BI solution for 1,000+ end users

Data Engineering Leader with 14 years of experience working in Business Intelligence, Data Warehouse & Data Integration, BigData, Cloud, and ML space across North America and Europe.

Leading Data Engineering initiatives, working on a petabyte size data platform built using cloud and big data technologies for supporting machine learning experiments, data science models, business intelligence reporting, and data exchange with internal and external partners. Handling privacy compliance and security-critical data sets.

Apart from work, I am teaching a Cloud Computing course at the University of Victoria, mentoring high school students at CS faculty, and volunteering my time for coaching people with analytics engineering skills in the CIS region. Moreover, I am the author of analytics books and speaker at data related conferences and user groups.

Data Engineering and Analytics Chief Instructor

Surfalytics = Surfing + Analytics. Inspired by West Coast surfing spots and Pacific ocean vibes. Created with goal to help you start a new career in data analytics space, develop data engineering and analytics skills through coaching. It will teach you not just dry skills, but will keep your focus to deliver significant value to businesses in the analytics realm as well as help to get fair compensation for the work you're passionate about.

There are several key producs:

1. Membership in community with the goal to land the job, prepare for the interview and get best offer. Every Saturday we do a Data Engineering and Analytics projects and discuss our progress on the journey.

2. Career consultations from Dmitry with focus on
2.1. Land 1st job in data
2.2. Move from analyst to engineer
2.3. Promote from middle to senior and beyond

3. Free course based on my 14 years experience in Data Engineering and Data Analytics

"𝐉𝐮𝐬𝐭 𝐮𝐬𝐞 𝐏𝐨𝐬𝐭𝐠𝐫𝐞𝐬 𝐚𝐬 𝐲𝐨𝐮𝐫 𝐝𝐚𝐭𝐚 𝐰𝐚𝐫𝐞𝐡𝐨𝐮𝐬𝐞"

Until you need:
- 5TB+ of data
- Complex joins over billions of rows
- Sub-second dashboard queries for 300+ users
- Time travel, cloning, and zero-copy environments
- Elastic compute separation for ETL vs BI

Then you’re not using Postgres.

You’re building an MPP cloud data platform.

Most engineers think “Postgres as a warehouse” =
- Create a cluster
- Add indexes
- Partition big tables
- Throw more CPU/RAM
- Done

Reality hits around 1–2TB and 200M+ rows.

You’re no longer running analytics on Postgres.

You’re fighting Postgres.

What scaling actually means:
- Query queueing & concurrency management
- Columnar storage for analytical workloads
- Elastic scaling for transformation pipelines
- Automatic clustering & micro-partition pruning
- Separation of compute & storage

Your <1s dashboard breaks down like:
- Network: 30–50ms
- Data fetch: 100–200ms
- Join across 800M rows: 300–600ms
- Aggregations: 200–400ms

You have 0–200ms buffer.
"Just add indexes" doesn't work.

Storage is the silent killer.
- 3TB in Postgres x 3 replicas x WAL overhead = 10+ TB real usage
- Vacuuming & reindexing = hours of block time
- Backups restore time = multi-hour downtime

Contrast with Snowflake / BigQuery / Databricks:
- Storage = pennies
- Compression = 4–10x
- Zero-copy clones = instant
- Auto-scaling warehouses

“We need to run transformations while dashboards are live”

Now your platform needs:
- Separated compute clusters
- Task orchestration & workload isolation
- Query optimization with statistics
- Automatic cluster suspend/resume
- Materialized views that actually work

This is where 80% of Postgres-as-DW projects die.

When Postgres is fine:
- Store up to 300–500GB
- Run BI on 10–20 users
- Rarely need heavy joins
- Can tolerate 3–5s latency
- ETL runs once per day

When you move to MPP:
- Switch to Snowflake / BigQuery / Databricks / Redshift / ClickHouse when:
- 2TB+ data
- 50–500 dashboard users
- Real-time ELT
- Sub-second latency requirements
- Multiple compute workloads

Cost math:
- Postgres cluster 64 vCPU / 256GB RAM on cloud: ~$2,500/mo + ops & tuning
- Snowflake / BigQuery on demand compute: ~$800–1500/mo for same workload

But add $100K/year engineering & admin cost.

8 months later — MPP pays for itself.

Production data platforms have 4 layers:
- Query planner & optimizer
- Distributed SQL execution engine
- Storage engine with micro-partitions & pruning
- Observability, caching, workload isolation


Postgres is brilliant.
It’s not a data warehouse.

At scale, it becomes technical debt.

Staff AI Data Engineer | Data Infrastructure for AI | Data Engineering & Analytics System Design Community | 8x tech book author

I recently read “5 𝐓𝐡𝐢𝐧𝐠𝐬 𝐈𝐧 𝐃𝐚𝐭𝐚 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐓𝐡𝐚𝐭 𝐇𝐚𝐯𝐞 𝐂𝐡𝐚𝐧𝐠𝐞𝐝 𝐈𝐧 𝐓𝐡𝐞 𝐋𝐚𝐬𝐭 10 𝐘𝐞𝐚𝐫𝐬” by Seattle Data Guy — and it strongly resonated with my experience working across North America and globally. 

Here are the points that stood out to me the most, with my personal perspective👇

1) 𝘌𝘷𝘦𝘳𝘺𝘰𝘯𝘦 𝘞𝘢𝘯𝘵𝘴 𝘚𝘦𝘯𝘪𝘰𝘳𝘴
𝘛𝘦𝘢𝘮𝘴 𝘢𝘳𝘦 𝘮𝘶𝘤𝘩 𝘴𝘮𝘢𝘭𝘭𝘦𝘳 𝘯𝘰𝘸, 𝘦𝘹𝘱𝘦𝘤𝘵𝘢𝘵𝘪𝘰𝘯𝘴 𝘢𝘳𝘦 𝘩𝘪𝘨𝘩𝘦𝘳, 𝘢𝘯𝘥 𝘣𝘶𝘴𝘪𝘯𝘦𝘴𝘴𝘦𝘴 𝘸𝘢𝘯𝘵 𝘪𝘮𝘱𝘢𝘤𝘵 𝘧𝘢𝘴𝘵 - 𝘴𝘰 𝘵𝘩𝘦𝘺 𝘩𝘪𝘳𝘦 𝘰𝘯𝘭𝘺 𝘴𝘦𝘯𝘪𝘰𝘳 𝘦𝘯𝘨𝘪𝘯𝘦𝘦𝘳𝘴 𝘸𝘩𝘰 𝘤𝘢𝘯 𝘥𝘦𝘭𝘪𝘷𝘦𝘳 𝘧𝘳𝘰𝘮 𝘥𝘢𝘺 𝘰𝘯𝘦. 

Entering the industry as a junior has become extremely difficult. Companies rarely want to invest in training. If you’re not from a top-tier university, the path is even tougher. And the course-to-job pipeline is mostly broken.

That’s why the strategy “enter as a senior without traditional experience” is becoming extremely attractive — build strong portfolio projects, real results, and business value instead of certificates.

2) 𝘊𝘭𝘰𝘶𝘥 𝘉𝘦𝘤𝘢𝘮𝘦 𝘵𝘩𝘦 𝘋𝘦𝘧𝘢𝘶𝘭𝘵
10 𝘺𝘦𝘢𝘳𝘴 𝘢𝘨𝘰 𝘤𝘭𝘰𝘶𝘥 𝘸𝘢𝘴 𝘢𝘯 𝘰𝘱𝘵𝘪𝘰𝘯.
𝘛𝘰𝘥𝘢𝘺 𝘪𝘵’𝘴 𝘢 𝘳𝘦𝘲𝘶𝘪𝘳𝘦𝘮𝘦𝘯𝘵.

AWS / Snowflake / BigQuery / Databricks / GCP — you’ll find them in 95% of job descriptions globally. On-prem still exists only for regulation or very specific cost cases.

3) 𝘕𝘰 𝘖𝘯𝘦 𝘉𝘶𝘪𝘭𝘥𝘴 𝘊𝘶𝘴𝘵𝘰𝘮 𝘗𝘪𝘱𝘦𝘭𝘪𝘯𝘦𝘴 𝘈𝘯𝘺𝘮𝘰𝘳𝘦
𝘈 𝘥𝘦𝘤𝘢𝘥𝘦 𝘢𝘨𝘰 𝘌𝘛𝘓 𝘸𝘢𝘴 𝘢𝘭𝘭 𝘤𝘳𝘰𝘯 𝘫𝘰𝘣𝘴, 𝘚𝘚𝘐𝘚, 𝘗𝘺𝘵𝘩𝘰𝘯 𝘴𝘤𝘳𝘪𝘱𝘵𝘴 𝘢𝘯𝘥 𝘪𝘯𝘵𝘦𝘳𝘯𝘢𝘭 𝘧𝘳𝘢𝘮𝘦𝘸𝘰𝘳𝘬𝘴.

Low-code / no-code tools and AI automation allow engineers to focus on business impact instead of debugging legacy pipelines. Technical fundamentals still matter - but reinventing the wheel doesn’t.

4) 𝘚𝘘𝘓 𝘞𝘰𝘯

If your data product doesn’t support SQL, it’s a bad product. Period.
Yes, some analysts love living inside Pandas notebooks, but eventually, engineers need to productionize and clean it all up. 😅

5) 𝘈𝘐 𝘊𝘩𝘢𝘯𝘨𝘦𝘥 𝘌𝘷𝘦𝘳𝘺𝘵𝘩𝘪𝘯𝘨
𝘈𝘐 𝘢𝘤𝘤𝘦𝘭𝘦𝘳𝘢𝘵𝘦𝘴 𝘦𝘯𝘨𝘪𝘯𝘦𝘦𝘳𝘪𝘯𝘨, 𝘳𝘦𝘥𝘶𝘤𝘦𝘴 𝘳𝘦𝘱𝘦𝘵𝘪𝘵𝘪𝘷𝘦 𝘸𝘰𝘳𝘬 𝘢𝘯𝘥 𝘮𝘢𝘬𝘦𝘴 𝘴𝘪𝘯𝘨𝘭𝘦 𝘥𝘦𝘷𝘦𝘭𝘰𝘱𝘦𝘳𝘴 𝘴𝘪𝘨𝘯𝘪𝘧𝘪𝘤𝘢𝘯𝘵𝘭𝘺 𝘮𝘰𝘳𝘦 𝘱𝘳𝘰𝘥𝘶𝘤𝘵𝘪𝘷𝘦.

But it also creates a risk: people rely on LLMs without understanding what they are doing. They copy code instead of thinking. Movement ≠ progress.
The winners are those who combine AI with deep engineering fundamentals and create real business value.

For me, the biggest shift in 10 years isn’t even tech — it’s economics:
->lower purchasing power
->higher taxes and cost of living
->disappearing job security
->no stability

Everyone wants a high salary and remote work, but there won’t be enough seats for everyone.

The next decade will belong to:
🔥 engineers who think like owners
🔥 experts who move fast and deliver value
🔥 those who can adapt, learn and use AI as leverage

Shubham Srivastava has a great list of data engineering things that are great to have. But if you want just to start and get your first job, you can focus on building one end-to-end project with low/no code:

1. Pick a public API (for example, GitHub)
2. Use Firvetran to extract data (free up to 500k rows per month)
3. Use Snowflake to ingest data (free trial up to 6 months with help of Coursera)
4. Use dbt core (free forever) to transform data with SQL to answer business questions (make up these questions)
5. Use GitHub to store the code and schedule dbt
6. Connect open source BI like Metabase or a free trial of Looker/Tableau
7. Add Airflow if you want a fully end-to-end

Advance topics
8. Replace Fivtran with Airflow and Python
9. Add AWS resources
10. Use Terraform to create resources in Snowflake and AWS
11. Add tests with dbt and soda

Master these topics and mock the system design interview, and you will be fine.

Principal Data Engineer @ Amazon | Data Engineering

Deciding to get into Data Engineering is one long chain of decisions:

- decide you want to be a data engineer
- realize you need to learn SQL first
- spend three weeks on window functions
- pick Python or Scala
- pick Python
- learn pandas
- rewrite everything in polars
- choose a cloud platform
- second-guess that platform
- set up AWS (then get lost in IAM hell)
- learn Docker
- debug Docker networking
- add docker-compose because fuck it
- realize you need Kubernetes
- decide Kubernetes is overkill
- pick an orchestrator: Airflow vs Prefect vs Dagster
- install Airflow locally
- spend two days fixing dependencies
- learn about DAGs
- write your first pipeline
- it fails silently
- add logging everywhere
- learn about idempotency (probably too late)
- set up a data warehouse
- choose between Snowflake, BigQuery, Redshift
- immediately regret your choice
- design your first star schema
- refactor it into a snowflake schema
- go back to star schema
- learn dbt
- rewrite all your transforms in dbt
- add tests (after breaking prod)
- handle incremental loads
- forget about deduplication
- add partitioning
- add clustering
- learn about slowly changing dimensions (ugh…)
- set up CI/CD
- break it
- fix it, add data quality checks
- learn Spark because the data got big
- debug Spark memory issues
- optimize Spark jobs you don't understand
- add monitoring and alerting
- get paged at 2am for a failed pipeline
- learn about data contracts
- write documentation nobody reads
- build a data catalog
- update your resume
- apply to jobs
- get rejected for not knowing tool X
- learn tool X
- that tool is now deprecated
- start a portfolio project
- abandon it halfway
- start another one
- ship it
- immediately find a data quality issue

it's the learning path for probably 90% of data engineers who entered the field in the last 5-10 years, and you're doing all this while trying to stay relevant in a field that changes every 6 months.

📖 Just finished "Empire of AI" by Karen Hao - and it made me realize: AI isn't just a story of efficiency and innovation. It's also a story of hidden trade-offs - in labour, resources, and power.

The uncomfortable truth? Building AI requires:
-> Data "cleaned" by workers earning <$1/day in Kenya
-> Data centers consume as much energy as entire cities
-> The creative work of millions, scraped without consent

Hao calls it what it is: a new age of empire.
Whether you're bullish or skeptical on AI, this is essential reading.

taplio

The Battle

Nikolay Valiotti

🏄‍♂️ Dmitry Anoshin

Followers

Posts

Engagements

Likes

Performance ScoreNikolay Valiotti's LinkedIn Performance

Performance Score🏄‍♂️ Dmitry Anoshin's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Nikolay Valiotti

🏄‍♂️ Dmitry Anoshin

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Nikolay Valiotti's LinkedIn Performance

Performance Score🏄‍♂️ Dmitry Anoshin's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreNikolay Valiotti's LinkedIn Performance