Posts Tagged ‘DataStrategy’

Building High-Performance Data Teams Starts With People, Not Tools

In the world of data engineering, we often obsess over the "plumbing." We talk about ETL latencies, Snowflake clusters, and the latest vector databases. But according to Sam Wrench, Lead at Reality Mine and former GB Dodgeball coach, we’re often looking at the wrong part of the machine. The most efficient data pipeline in the world is useless if the people at either end of it—the engineers building it and the stakeholders consuming it aren't in sync. To Sam, data engineering isn't just a technical challenge; it’s a high-performance team sport.

The Economics of the Modern Data Stack

Who captures the value now? The modern data stack still works. The problem is not capability. The problem is economics. On Friday, we hosted a LinkedIn Live to talk about something most teams feel but rarely articulate clearly: the incentives behind the modern data stack have shifted, and those shifts are starting to shape architecture, pricing, and leverage in ways that matter over the next 12–24 months. This was not a tools debate. It was a discussion about who captures value, who carries risk, and why “best-of-breed” no longer feels neutral. Joining the conversation were: Maxime Beauchemin, creator of Apache Airflow and Apache Superset. Taylor Murphy founder of Meltano and Arch, now at Astronomer. Aaron Phethean, Founder and CEO of Matatika. What follows is not a summary. It’s what actually matters.

Baidu ETL Connector: How MVF Solved an Unsupported Data Source

Today, Matatika is the only ETL provider offering a fully supported, production-grade Baidu connector. While most ETL platforms support mainstream paid media sources like Google Ads and Facebook Ads, Baidu often sits outside standard connector catalogues. For teams running marketing activity in China, this creates a familiar problem: critical data exists, but there is no clean, supported way to ingest it. MVF ran into exactly this issue.

Why marketing data connectors quietly inflate your ETL costs

And how MVF consolidated paid media data without losing insight Marketing data rarely breaks loudly. It degrades quietly. Pipelines keep running. Dashboards still load. Spend gets approved. But somewhere between your tenth and thirtieth connector, the economics stop making sense. Teams often assume marketing connectors are cheap because each one looks small in isolation. A Google Ads connector here. A LinkedIn Ads connector there. Another for TikTok. Another for reporting exports. Each one feels justified. Together, they quietly become the most expensive part of the data stack. MVF learned this the hard way.

🔔 Celebrating a Most Eventful Matatika Year! 🔔

The Matatika Year in Review: On the Twelfth Day of Data... As the year wraps up, we’re taking a lighthearted, musical look back at the incredible journey we’ve shared! Thanks to the energy and support of our amazing community, 2025 has been an absolutely unforgettable year of growth, connection, and major milestones. Grab a hot drink, and join us as we sing the praises of the Matatika year that was!

 

Stop Manually Uploading Spreadsheets: 5 High-Impact Use Cases (and How Resident Advisor Fixed It)

If your week still includes exporting CSVs and uploading them into dashboards, you’re paying a “data tax” in delays, context switching, and stale numbers. Here are the top five use cases where teams should replace manual uploads with a real Google Sheets → warehouse pipeline plus a quick look at how Resident Advisor (RA) made this work in production.

Build Strategy First. Choose Technology That Deserves It

Introduction Every organisation has lived this story. A new data platform or AI tool promises to fix everything. Budgets are approved, contracts are signed — and a few months later, the same problems remain. Dashboards don’t align, pipelines still break, and confidence in the numbers keeps slipping. For Dylan Anderson, Head of Data Strategy at Perfusion, the real issue isn’t the technology itself — it’s where teams choose to begin. “Vendors are great at selling the idea that their product will solve all your data needs,” he says. “But strategy isn’t about buying more tools. It’s about helping the business reach its goals using data, technology, and AI in the right way.” In this episode of Data Matas, Dylan joins host Aaron Phethean to explore why strong data strategies always start with purpose, and how honest, outcome-driven technology earns its place in the stack.

How high-performing data teams align business goals, data strategy, and technology to build trust that lasts

How to Scale Mixpanel Data Efficiently Without Spiralling Costs

Your product's growing. More users, more events, more insights flowing through Mixpanel. That's exactly what you want. The problem? As your Mixpanel event volume increases, your data infrastructure costs often grow faster than your revenue. High-volume, append-only event data breaks traditional ETL pricing models. Every duplicate event costs money. Every unchanged property gets billed. Growth becomes a financial penalty. There's a smarter way to handle event data at scale.

Three Things Every Data Leader Should Kill Before Building Another Dashboard

Most organisations are drowning in dashboards that no one trusts. In this Data Matas episode, former Worldpay and FIS data leader Phil Thirlwell explains why the key to better decisions isn’t building more it’s stopping first. He breaks down how dashboard sprawl, KPI overload, and service-desk habits create chaos, and how treating dashboards like products can rebuild trust. Phil shares practical ways to simplify metrics, prioritise outcomes, and run data teams with purpose. The takeaway: fewer dashboards, clearer decisions, stronger alignment between data teams and the business.

 How to kill dashboard sprawl, service-desk habits, and KPI overload to rebuild trust in data.

How Hypebeast Reached 97% AI Adoption Without Fear or Layoffs

At Hypebeast, 97% of staff now use AI daily not out of fear, but choice. Director of Data & AI, Sami Rahman, reframed AI as a creative ally, not a threat. By focusing on practical wins, like speeding up research and cutting drudgery, he built trust and curiosity. Instead of pushing tools, he created demand through scarcity, measured impact rigorously, and deleted underused agents without sentiment. The result: adoption that stuck, creativity that flourished, and teams that saw AI as empowerment, not replacement. A playbook for leaders who want AI adoption to last built on trust, not hype.

AI adoption is at the top of every data leader’s agenda. Yet most attempts stall. Leaders flood their teams with new tools, staff get overwhelmed, and adoption drops. In some cases, AI is seen as a threat rather than an enabler.

At Hypebeast, it’s different. 97% of staff now use AI agents in their daily work. Not because they were forced, but because they wanted to.

In this episode of Data Matas, I spoke with Sami Rahman, Director of Data & AI at Hypebeast, about how he made AI adoption stick. His story offers grounded lessons for any data leader trying to balance hype with reality.

Meet Sami Rahman

Sami leads Data & AI at Hypebeast, the global fashion and lifestyle brand. His career spans psychology, counter-terrorism, and data science — giving him a rare perspective on how people, systems, and trust interact.

That unusual career path means he doesn’t just see AI as technology. He sees it as part of a wider human system — where behaviour, incentives, and culture matter just as much as code.

“We’re a creative company. We don’t want to replace journalists or designers. But we can use AI as a force multiplier — speeding up research, consolidating information, and helping people make decisions faster.”

That human-first but pragmatic outlook shaped every decision in Hypebeast’s adoption journey.

Why Most AI Adoption Efforts Fail

AI is no longer optional. Boards and executives expect adoption. But many teams fail to deliver value. Why?

Sami points to fear — not of the technology, but of being abandoned.

“The reason people are fearful around AI isn’t the tech. It’s because they don’t trust governments or institutions to look after them if jobs disappear.”

This distinction matters. If AI is framed as replacement, staff feel threatened. If it’s framed as empowerment, they engage.

Psychology backs this up. People resist change when they fear loss of control or status. The solution isn’t just better tech — it’s better framing. Data leaders need to talk about AI as a tool that supports their teams’ value, not one that makes them redundant.

Lessons from Hypebeast’s Adoption Journey

1. Frame AI as a Force Multiplier

At Hypebeast, AI is not a substitute for creativity. It’s an assistant. Tasks like research, summarisation, and trend monitoring were made faster and easier, while final judgement stayed human.

“We’re not trying to replace jobs. We might automate manual tasks, but we won’t remove the human side.”

This framing reassured staff that their value remained central — and made AI a welcome tool, not a competitor.

Implementation guidance:

  • What to do first: Communicate clearly that AI enhances, not replaces.
  • Tools: Introduce AI where speed and consolidation matter (e.g. research, summarisation).
  • Watch-outs: Don’t oversell — focus on real, modest gains.
  • Benefit: Higher trust and willingness to experiment.

2. Focus on “Unsexy” Use Cases

Flashy AI demos rarely translate into real value. Sami leaned into the unglamorous but high-impact tasks: scanning social feeds, packaging intelligence, automating logistics and finance.

“We leaned into use cases that aren’t super sexy but free up time.”

By cutting drudgery, staff had more time for meaningful creative work.

Implementation guidance:

  • What to do first: Audit manual processes that drain time.
  • Tools: Simple AI agents for monitoring, reporting, logistics.
  • Watch-outs: Avoid over-investing in “showcase” projects.
  • Benefit: Faster results, more trust in AI.

3. Create a Curiosity Gap

Perhaps the boldest move was delaying access. For 10 weeks, Sami drip-fed teasers: short demos showing what AI could do.

“It was 10 weeks before we gave anyone access — on purpose… By the launch, adoption went from 3% to 95% in three weeks.”

Scarcity created FOMO. Instead of pushing adoption, Hypebeast created pull.

This contrasts with most change management programmes, which over-prepare staff with slide decks, training, and handholding. Sami flipped the playbook — and in his industry, it worked.

He’s quick to note it wouldn’t fit everywhere. In banking or pharma, where regulation and compliance demand rigour, leaders may need a heavier hand. But in fast-moving creative industries, curiosity was the lever.

Implementation guidance:

  • What to do first: Hold back, and drip-feed examples.
  • Tools: Short demo videos or snippets to spark curiosity.
  • Watch-outs: Don’t launch too early before demand builds.
  • Benefit: Rapid, voluntary adoption.

4. Kill Zombie Agents Without Sentimentality

Hypebeast set strict benchmarks: daily or weekly agents had to hit 80% usage. If they weren’t used, they were deleted.

“If an agent isn’t used, we delete it. No sentimentality. It’s not failure — it’s iteration.”

That unsentimental approach kept adoption high and avoided wasted energy.

This is another overlooked lesson. Too many teams keep “zombie tools” alive because someone invested time or money. Sami’s product mindset — measure, test, delete — freed his team to focus only on what added value.

Implementation guidance:

  • What to do first: Define usage thresholds per agent type.
  • Tools: Usage dashboards, adoption metrics.
  • Watch-outs: Don’t cling to underused tools.
  • Benefit: Consistently high adoption, leaner portfolio.

Measuring Adoption: Beyond Usage

Usage was Sami’s primary metric, but adoption measurement can and should go deeper. Data leaders can track:

  • Time saved on manual tasks
  • User satisfaction with AI support
  • Error reduction in workflows
  • Frequency of repeat use across teams

By triangulating usage with impact, leaders can see not just whether AI is being used — but whether it’s making a difference.

This level of measurement is critical for building trust with executives and avoiding the “we bought AI, but what did it achieve?” backlash.

Real-World Impact

Hypebeast reached 97% adoption within three weeks of launch. Staff now use AI agents daily across journalism, retail, logistics, and design.

The before state was one of downsizing, high pressure, and manual workloads. The after state is a team with more time for creativity, backed by systems that take care of drudgery.

Instead of creating fear or resistance, the approach built curiosity and trust. AI is now embedded in workflows, freeing staff to focus on creative and strategic tasks.

Putting It All Together

Hypebeast’s success was not built on hype or heavy-handed change management. It came from reframing AI as support, solving practical problems, creating demand through scarcity, and cutting what didn’t work.

The results speak for themselves: adoption consistently above 90%, and a team that sees AI as an enabler, not a threat.

For data leaders, the playbook is clear:

  • Start with framing — AI enhances, it doesn’t replace
  • Automate boring tasks first
  • Create demand with curiosity, not forced adoption
  • Be ruthless with underperforming tools

Your Next Move: A Leader’s Checklist

Before your next AI rollout, ask yourself:

  1. Have I made it clear AI is here to support, not replace?
  2. Am I solving everyday pain points first, not chasing flashy demos?
  3. Can I create demand by showing value before rolling out access?
  4. Do I have clear benchmarks for adoption and usage?
  5. Am I willing to cut what doesn’t deliver?

Tick those five boxes, and you’ll be far closer to adoption that actually sticks.

Dig Deeper

This article is based on Data Matas Episode [X] with Sami Rahman, Director of Data & AI at Hypebeast.

📺 Watch the full conversation: https://www.youtube.com/@matatika
🎙️ Listen to the podcast: https://www.matatika.com/podcasts/