Armin Kakas

Revenue Growth Analytics advisor to executives in charge of growing Net Revenue, Gross Profit, and Sales Productivity | Posts and articles about Revenue Management + Machine Learning insights, methods, and processes.

As the founder of Revology Analytics, I have the privilege of offering mid-market businesses unparalleled expertise in advanced Revenue Growth Analytics & Management, focusing on three core advisory areas: Pricing & Revenue Management, Sales & Marketing Enablement and Commercial Analytics Transformation efforts. 

Beyond my advisory work, I am the creator of the Revology Analytics Insider, a bi-weekly newsletter offering readers unique insights into the world of Revenue Growth Analytics. You can catch the latest issues on LinkedIn and our website, revologyanalytics.com/articles-insights.

To learn more, please visit revologyanalytics.com

Founder | Managing Partner

Revology Analytics is a Revenue Growth Analytics consultancy for middle-market companies in the Manufacturing, Retail, and Distribution industries.

We are an impact- and execution-oriented advisory firm founded on Outcome-Based Analytics™, with focused specializations in AI/ML-enabled Margin Analytics & Optimization, Sales & Customer Growth Analytics and Promotion Effectiveness & Optimization.

Know where your data and AI can take your business ✈️| Follow for data models delivering results | Data-Driven Growth Navigator 100+ Data Products Launched | $1B+ revenue, $100M+ savings driven by data and AI

Product and data executive who builds AI-powered platforms that generate billions in revenue. Led data strategy for Olo's $3.6B IPO. Launched the industry's first GenAI/LLM product for logistics in 2022—18 months before ChatGPT.

I bridge the gap between technical innovation and P&L impact. My teams have generated $2.5B+ in new revenue and delivered $150M+ in cost savings for Fortune 500 companies and high-growth startups.

𝗪𝗵𝗮𝘁 𝗜 𝗗𝗼 𝗕𝗲𝘀𝘁:
→ Turn data into products (built a $1B personalization platform at Best Buy for 15% of vendor cost)
→ Scale organizations (2 people → 300+ cross-functional teams)
→ Ship "impossible" roadmaps (90-day platform builds, 10X faster development cycles)
→ Navigate high-stakes moments (IPO prep, board presentations, M&A due diligence)

𝗧𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗘𝘅𝗽𝗲𝗿𝘁𝗶𝘀𝗲:
GenAI & LLMs (GPT, Llama, Claude) • Multi-Agent Systems (CrewAI, LangChain) • RAG & Vector Databases • Snowflake • Databricks • AWS/Azure • MLOps • P&L Management ($30M+)

𝗖𝗮𝗿𝗲𝗲𝗿 𝗛𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁𝘀:
• GenAI Pioneer: Built the industry-first GenAI/LLM product for logistics (2022), establishing competitive moat 18 months ahead of market
• IPO Leadership: Led enterprise data strategy for Olo's successful $3.6B IPO (NYSE: OLO)
• Revenue Growth: Built enterprise personalization platform at Best Buy generating $1B+ revenue at 15% of vendor cost
• Scale: Proven ability to build and scale teams from startup to 300+ employees

𝗖𝘂𝗿𝗿𝗲𝗻𝘁𝗹𝘆: Fractional CPO/CDO for 4 companies, focusing on AI product strategy, data monetization, and organizational transformation.

𝗢𝗽𝗲𝗻 𝘁𝗼: Chief Data Officer, Chief Product Officer, VP of AI & Product Strategy roles. Remote, hybrid, or relocation for the right opportunity.

Let's connect: edward.chenard@echenard.com | echenard.com

Edward Chenard

$20M.

That's how much revenue was hiding inside Target's data when I got there. Nobody had built anything to capture it. The data just sat there.

I keep seeing the same thing at every company I walk into. Millions in revenue sitting in data that nobody has turned into a product. Not because the technology doesn't exist. Because nobody asked the right question.
The question isn't "what can we build with AI?"

The question is "where are we leaving money on the table?"

When I asked that question at C.H. Robinson, the answer was freight pricing. They were a $15B company pricing loads without predictive models. We built one. $150M in new business. Microsoft and John Deere signed on before the platform was even finished.

At Best Buy, the answer was personalization. Online conversion was 1%. One percent. We built a personalization engine in 90 days that took it to 17%. Over $1B in revenue. Vendors had quoted $20M to $30M to do the same thing. We spent $3.2M.

At Shipwell, the answer was reporting. Enterprise clients were spending half a day building a single report. We built a natural language interface where you just ask the data a question. Took reporting from 12 hours to 6 minutes. That became a $2.2M product line.

Every one of these started the same way. Not with a data strategy deck. Not with a vendor evaluation. Not with a hiring plan. With one question about money.

I talk to AI leaders every week who are 8 months into building infrastructure and have zero revenue to show for it. They have beautiful architecture diagrams. They have governance frameworks. They have a data catalog.
They don't have a product anyone is paying for.

Meanwhile the company across the street shipped something ugly in 30 days and is already generating revenue.

If you're an AI or data leader and your board is starting to ask hard questions about ROI, stop what you're building and answer this first: what is the single biggest revenue problem your data can solve in the next 30 days?

Start there. Everything else comes after.

What revenue is hiding in your company's data right now that nobody has touched?

I help companies turn data into revenue streams, connect with me if this is something you want to do.

#AIStrategy #DataLeadership #GenAI

AI & Data Executive | Built the First GenAI Product for Logistics Before ChatGPT | $2.5B Revenue Impact

I've been selected by 4 companies in 18 months to lead AI and data strategy as a fractional CDO/CPO.

Here's what I've learned that most full-time executives are too busy to see:

Full-time leaders are optimizing the machine they're inside of.
Fractional leaders are pattern-matching across machines.


In the last year alone I've seen:

A manufacturer bleeding warehouse churn that nobody had named yet. We cut it 13% in one quarter with a multi-agent GenAI platform.

A hospital drowning its staff in paperwork. Not a staffing problem. An AI deployment problem. Non-patient workload dropped 31%.

A bike shop losing money to inventory guesswork. Predictive models cut that waste by 50%.

Three completely different industries. Same root cause: data sitting unused while decisions got made on gut feel.

The fractional model works because I'm not protecting a budget. I'm not managing up. I'm not waiting for the next performance cycle.

I show up, find the highest-leverage problem, and build.

If your company is in a growth or transformation phase and you're not sure whether you need a full-time data hire yet, you probably don't. You need someone who has already solved the problem you're about to have.

What's the biggest AI or data decision your team is staring down right now?

Looking for help with your data and AI, I can help.  See my profile for more information

One page.

That's all I give teams to define a data product before we start building. If you can't explain the what, why, who, cost, and the expected revenue on a single page, you don't understand the project well enough to start it.

I built this template after watching too many data products fail for the same reason. Bad clarity.

A team would spend 6 months building something, then someone in leadership would ask "wait, what problem does this solve?" and nobody could give a clean answer. Or worse, everyone gave a different answer.

So I created a one-page brief that every data product has to go through before it gets a single hour of time. I've used it at Fortune 500 companies and 10-person startups.

Here's what's on the page.

BUSINESS STREAM: Which part of the business does this support? If the answer is "everyone" then you haven't thought hard enough. A data product that serves everyone serves no one.

PROJECT CONTEXT: What's happening in the business that makes this necessary right now? Not why it would be nice to have. Why it matters today.

ROLES: Owner, Sponsor, Benefits Owner, Maintainer. If you can't fill in all 4, the project will stall. The one most people skip is Maintainer. Who keeps this thing running after it ships? If nobody owns that, you're building something that dies in 6 months.

DEFINITION OF SUCCESS: Not "better insights" or "improved decision making." A number, revenue generated, cost reduced, time saved.  If you can't put a number here, go back and figure out what problem you're actually solving.

SCOPE: Two columns. In-scope and out-of-scope. The out-of-scope column is more important than the in-scope column. This is where you kill feature creep before it starts. I've seen data products go from "build a pricing model" to "build a pricing model with a dashboard and an alert system and an integration with Salesforce and a mobile app" because nobody wrote down what was out of scope.

THE FINANCIALS: Cost vs. benefits over 3 years. Capital, expense, operating costs on one side. Sales, gross margin, etc... This is the section that separates real data products from science projects. If you can't project the financial impact, you're guessing.

KEY PHASES AND MILESTONES: Broken down by quarter. Not a 47-page project plan. Just the big milestones that tell you if the project is on track or not.

HELP NEEDED: What's blocking you right now? What do you need from leadership? This section forces honesty. Most project briefs are designed to make things look good. This section is designed to surface problems early.

Every data product I've shipped started with some version of this page. When teams push back and say it's too simple, I tell them that's exactly the point. Complexity is where bad projects hide.

If your team is working on a data or AI product right now, could you fill this out today? If not, that tells you something important about how well the project is actually defined.

#AIStrategy #DataLeadership #GenAI

Last week I shared my 90-Day Proof framework and Phase 2 got many comments.

"Ship something ugly."

People hated it and loved it at the same time. So let me tell you what ugly  looked like.

Best Buy, 2011. The company was in the middle of a turnaround. E-commerce conversion was at 1%. Vendors came in pitching personalization platforms. $20M to $30M. 18 to 24 months to deploy.

We didn't have 18 months. The board needed results now.

So my team built a personalization engine in 60 days. Was it elegant? No. Was the code something I'd show off at a conference? Absolutely not.

But it took conversion from 1% to 6%. And the ROI hit in 4 months.

The vendors were still doing discovery workshops.

Here's what I've learned about shipping ugly across 6 companies and 100+ products.


UGLY RULE 1: Solve one problem completely, not ten problems halfway.
At Shipwell in 2022, our customers were spending 12+ hours building enterprise reports. We could have built a beautiful analytics dashboard. 

Instead we built a natural language interface. You type a question, you get an answer. That's it. One problem. Solved completely.

It looked rough. But dispatchers were using it every day within weeks. 


UGLY RULE 2: Use the data you have, not the data you wish you had.
Every company I walk into says the same thing. "We can't do AI yet because our data is messy." I've heard this at Fortune 500s with thousands of engineers and at startups with 12 people.

Your data will never be clean enough. Ship with what you have and clean it up after you've proven the value. At C.H. Robinson I built an AI prediction platform on data infrastructure that didn't exist 6 months earlier. It was messy. It still won Microsoft and John Deere as customers.


UGLY RULE 3: If it takes more than 30 days to get to a user, you're overbuilding.

I give every first product a hard 30-day deadline from kickoff to a real user touching it. Not a demo for leadership. Not a sandbox. A real person using it to do real work.

This forces brutal prioritization. You can't build a data lake, a governance framework, and a product in 30 days. So you skip the first one and build the product with governance that covers the basics. You earn the right to build infrastructure with the revenue from v1.

A bike shop owner I worked with recently needed an inventory planning tool. We had AI predictive models running inside of 3 weeks. Reduced his inventory management time by 50%. It was ugly. He didn't care. It saved him hours every week.

The discomfort you feel about shipping something imperfect? That's the exact instinct that keeps AI projects stuck in pilot purgatory for years.
Your competitor isn't building something better. They're just building something first.

What's the ugliest v1 you ever shipped that ended up working? I want to hear the real stories.  I'm here to help build so reach out if you need help.

#AIStrategy #DataLeadership #GenAI

I shared my 90-Day Proof framework. Yesterday, I went deep on Phase 2, "Ship Something Ugly."

Today let's talk about Phase 1. The one that makes data leaders uncomfortable.

Find the money first.

Not build the data lake first. Not hire the team first. Not set up governance first. Find the revenue problem that data can solve right now.

I know this makes some of you twitch. But I've done it 6 times and it keeps working. Let me show you why.

When I joined C.H. Robinson, they had zero data infrastructure, data products, or data teams. Just me.

The safe move would have been to spend 6 months building a foundation. Instead, I asked one question: "Where is this company leaving money on the table?"

The answer was pricing. A $15B logistics company was pricing freight without predictive models. The opportunity was massive.

Within months we had an AI prediction platform in production. It generated $150M in new business. Microsoft and John Deere became customers. I got $20M investment from the board.

That investment funded the infrastructure, the team (0 to 45 people), and 15 more products. But none of that happens if I start with the data lake.

At Target, same playbook. First month I looked at their email program. Wide open opportunity. We built an ML-powered trigger algorithm that drove 400% higher open rates and $3M in incremental sales.

At Olo before the IPO, I found $20M in incremental revenue hiding in data products that nobody had thought to build yet.

The pattern is always the same.

Revenue first. Infrastructure second. The money funds the platform. Not the other way around.

HOW TO FIND THE MONEY IN YOUR FIRST 30 DAYS
Talk to sales. Not engineering. Not the data team. Sales. Ask them what questions they can't answer, what deals they're losing, what takes too long. That's where the money is.

Look for manual processes that touch revenue. If someone is spending 12 hours building a report that drives a pricing decision, that's your target. At Shipwell we turned that exact problem into $2.2M in new ARR.

Find the executive who owns a P&L and is frustrated. Not the executive who wants a dashboard. The one who wants more revenue or lower costs and doesn't care how you get there. That person becomes your sponsor, your champion, and your budget.

Ignore anyone who says you need a data strategy document before you can start. The strategy is: make money with data. The document can come later.

I realize this approach isn't for everyone. If you're at a company where compliance or regulation requires infrastructure first, that's a different situation.

But for the other 90% of you sitting in planning meetings about your AI roadmap while your competitors are shipping? Go find the money.

What's the biggest revenue opportunity hiding in your company's data right now? The one everyone knows about but nobody has built anything for yet?

Check out my profile, I help companies find ways to make money with data.
#AIStrategy #DataLeadership #GenAI

You are not the same person at 8am and 8pm.

But every personalization system treats you like you are.

This is the biggest mistake in AI-driven personalization and almost nobody talks about it. I've built personalization engines at Best Buy, Target, and Olo across 100M+ customers. The thing that made the biggest difference wasn't a better algorithm. It was a concept from psychology called the Fundamental Attribution Error.

Most personalization programs assume your behavior comes from who you are. Your traits. Your profile. So they build one model of you and serve the same recommendations whether it's Tuesday morning or Saturday night.

That's wrong. Your behavior is mostly driven by your situation, not your identity.

Think about food. Hungry at noon on a workday, you want something fast and close. At 7pm on a Friday, you're browsing, aspirational, open to trying something new. Same person. Completely different buying behavior.

At Olo, I built personalization strategy around this for 80,000 restaurant clients. Instead of one static profile per customer, we used day parting. Breakfast you, lunch you, and dinner you are three different customers. Research on this showed 30-40% sales increases versus traditional one-identity personalization.

This applies way beyond restaurants. At Best Buy, conversion went from 1% to 17%. A big part of that was understanding someone browsing laptops at 10am Monday is researching for work. Same person browsing TVs at 9pm Saturday is in a completely different headspace. Same customer ID. Different person.

At Target, we built cross-device personalization spanning 100M+ loyalty members. The biggest unlock wasn't the technology. It was mapping behavior to context, not just to a customer profile.

The psychology of personalization matters more than the technology of personalization.

Most teams jump straight to the algorithm. Collaborative filtering. 

Recommendation engines. ML models. Those are tools. If you're feeding them a single-identity model of your customer, you're optimizing a flawed assumption really efficiently.

Start with one question: who is my customer right now, in this moment? Not who are they in general.

Anyone else building personalization that accounts for time of day and context? 

Or is everyone still stuck on one profile?

#Personalization #AIStrategy #DataScience

I've presented to boards over 20 times. The first time, I almost got laughed out of the room.

I walked in with a beautiful deck, model accuracy charts, architecture diagrams, and a roadmap full of technical milestones. The board listened politely for about 4 minutes. Then the CFO interrupted and asked one question.

"How much money does this make us?"

I didn't have a clean answer. Meeting over.

That failure changed how I present to every board since. At C.H. Robinson, I presented 9 times in one year. At Olo, 6 times during IPO preparation. At Shipwell, quarterly. Every single time I walked in with the same structure.

Three numbers. That's it.

What it cost. What it made. What it will make next quarter.

Everything else is supporting material. The board doesn't need to understand your model. They don't care about your tech stack. They definitely don't care about your data pipeline. They care about one thing: is this a good investment?

Here's what I learned about the difference between a board that funds AI and a board that kills it.

BOARDS THAT FUND AI hear about revenue first and technology never. At C.H. Robinson, I showed that our AI prediction platform generated $150M in new business. That got me $20M in multi-year investment. I never once showed a model performance chart.

BOARDS THAT KILL AI hear about potential and possibility. Words like "enable" and "optimize" and "leverage" are death sentences in a board room. If you can't put a dollar amount on it, don't bring it up.

The mistake most AI leaders make is thinking the board presentation is about getting buy-in for the technology. It's not. It's about proving you're running a business, not a science experiment.

At Olo, the board cared about one thing during IPO prep: does our data strategy hold up under public market scrutiny? I didn't talk about algorithms. I talked about risk, revenue, and growth opportunities. Six presentations. All funded. The company went public at $3.6B.

Three tactical things that work in every board presentation.

Lead with the business problem, not the solution. "We're losing $40M in pricing efficiency" hits different than "we built a machine learning model for pricing."

Compare your cost to the alternative. Boards love this. "We built it for $3.2M in 90 days. Vendors quoted $20M and 18 months." That's a story they'll retell.
End with what you need, not what you did. Boards want to make decisions. Tell them the decision you need them to make and give them the numbers to make it.

If you're presenting AI to your board soon, what's the one number that would make them say yes?

I am here to help companies make money from AI.  Check out my profile and site for more details.

#AIStrategy #DataLeadership #GenAI

taplio

The Battle

Armin Kakas

Edward Chenard

Followers

Posts

Engagements

Likes

Performance ScoreEdward Chenard's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Armin Kakas

Edward Chenard

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);}Edward Chenard's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreEdward Chenard's LinkedIn Performance