Just get on the plane. ✈️ We'll take it from there. We'll whisk you straight from the airport to the conference venue where 🌓🌕🌕🌕🌕🌗 days of pure data awaits - packed with speeches, conversations, unexpected adventures, late-night discussions, and all the other things SUPERWEEK has become synonymous with.

🛎️ RECEPTION WITH CONDUCTRICS

OPENING

Iqbal Ali
Conductrics - Ilford, UK

Matt Gershoff
Conductrics - Austin (TX), USA

A THOUSAND AI ANALYSTS. A MIND OF YOUR OWN.

Ibrahim Elawadi
Philips - Amsterdam, The Netherlands

Last year at Superweek I gave the same question to 1,000 AI analysts and looked at what came back. What surprised me most was how closely they mimicked recognisable human behaviour.

This year I'm going back to the same 1,000 analysts with a harder question: how do you know if what they give you is any good? I'll share what I've been testing since then, what worked, what didn't, and what it says about how we should be checking AI work.

I'll also look at the human side. Working with AI every day changes how we think, and not always in the direction we'd like.

You'll leave with a clearer way to judge your own AI analysts, and a few uncomfortable questions to ask yourself.

WE BROKE GOOGLE'S NEW API SO YOU DON'T HAVE TO

Dan Truman
Duga Digital - Sevenoaks, UK

Doug Hall
Duga Digital - Casefabre, France

[This talk will be a 2 person tag team cabaret by Dan and Doug.]

In this session, we invite all walks of life, every level of technical expertise is welcome, from those only fluent in corp-speak, to hard-core, dark mode Omarchy pioneers. We'll explain how we developed something quite special with the Data Manager API - we're handling online, offline, lead gen, soft and hard conversions with attribution capabilities never before seen.

And it's been quite a journey.

Doug will present the technicalities, Dan will provide the human readable version. This is important as everyone needs to understand the impact, the how, and the why. In doing so, we want to show how Duga works, how we vibe, how we work to each other's strengths.

It'll be a laugh riot - see you there.

Karolina Wrzask
Google - Lodz, Poland

DAY ZERO: A BOOTSTRAP SYSTEM FOR ANALYTICS DOCUMENTATION & GOVERNANCE

Lukas Grebe
UNICEF - Remscheid, Germany

Every new client, new employer, or new stack starts the same way: no measurement plan, no data dictionary, no naming conventions, and a tracking setup nobody can explain. Over the past decade — across agency, in-house, and freelance work — I've refined a bootstrap system for fixing that in the first weeks, before the first "why don't these numbers match?" lands.

This session walks through the actual kit: what I set up on day zero, in what order, what earns its place and what I've learned to skip. The spine is the Analytics Topic Pyramid — Define, Capture, Report, Analyze, Test, Predict — with documentation as the connective tissue between layers.

One useful side effect: if you write this documentation in structured markdown, it's natively readable by AI. Governance done properly a decade ago turns out to be retroactively AI-ready.

MAN VS. MACHINE: A FOREGONE CONCLUSION?

Marie Fenner
Piano - London, UK

Every wave of automation arrives with the same warning: the machines are coming for our jobs. It was true of the looms of the First Industrial Revolution, and it's the headline again with AI, now with new names attached. "Digital dementia" warns that outsourcing our thinking will leave our brains sluggish. FOMO has made way for FOBO: the fear of becoming obsolete.

What has changed is not the fear but the line. We used to know what humans did best and what machines did best. AI is blurring that line. This session looks at real contests between human and machine, what they reveal about where the line now sits, and what a winning formula looks like for analytics leaders and their teams.

PROVE IT: A FIELD GUIDE TO EVERYTHING THAT'S QUIETLY WRONG WITH YOUR DATA

Sayf Sharif
DataTrue - Philadelphia (PA), USA

Les murs ont des oreilles... the walls have ears. From Catherine de' Medici to European Black Chambers, four hundred year old Swedish shipwrecks, American Bootleggers, and an English antique dealer in Dorking, Surrey, UK. Privacy and good data have been quietly failing the people who trusted them for a very long time, long before anyone had heard of a website.

This talk walks through four categories of how your data and your privacy compliance actually break today, almost always invisibly, as well as presenting a framework to prove they haven't and keep them from doing so.

MONDAY EVENING FIRESIDE PUB QUIZ WITH CONDUCTRICS

Ezequiel Boehler
Conductrics - Buenos Aires, Argentina

Matt Gershoff
Conductrics - Austin (TX), USA

Blair Ault
Conductrics - Austin (TX), USA

They will be your quiz masters for a fun filled night of quizzical conundrums, bragging rights, and prizes.

IT DEPENDS: THE TRADEOFFS BEHIND EVERY AI DATA TASK

Juliana Jackson
Jellyfish - Malmo, Sweden

Every question about using AI for data work gets the same answer: it depends. This talk is about what it depends on.

Which model to use, how much data you need, how accurate is accurate enough, who checks the output. None of these has a best answer. Each one is a tradeoff, and most of us make it without noticing.

In this talk, I will take 4 tasks you already do or are about to do (classification, prediction, semantic analysis, and analysis and visualisation) and run each one through the same questions: what you need to consider, which tradeoffs matter, what they cost, and how to evaluate whether you chose well. In painful detail, as always.

You will find out when to give up data volume for quality, when precision matters more than recall and when it is the other way round, when a large model is worth the money and when a small one will do, and how much human review the output really needs.

By the end my talk, you will be able to look at any AI data task and see which tradeoff you are about to make...and wether it's worth it.

Denis Golubovskyi
Stape.io - Kyiv, Ukraine

SIGNAL HARVESTING REVISITED: DESIGNING FOR UNCERTAINTY

Astrid Illum
TV 2 Danmark - Copenhagen, Denmark

A few years ago at Superweek, I talked about "signal harvesting": the idea that sometimes the best way to understand behaviour is not to observe people more closely, but to create something meaningful for them to respond to.

Since then, I've had the chance to return to that idea in a very different setting: account sharing in streaming.

This talk is about what happened when we stopped trying to prove exactly what people were doing, and instead started asking what we could reasonably respond to and what we might learn from that response.

It is about friction, feedback loops and product interventions as a way of creating better signals. And, ultimately, about what happens when the thing you build to change behaviour also becomes the thing that helps you understand it.

THE ANALYTICS SYSTEM THAT DOESN'T LIE TO YOU

Ivan Burban
Coupler.io - Krakow, Poland

Those who implemented AI in analytics workflows faced common issues with accuracy and trust. In this session, Ivan will explore how the Coupler.io team resolved these challenges and turned interactions with AI into a proven approach. Ivan will discuss the role of the Analytical Engine, the lifecycle of BI and AI solutions, what the context layer must include, and how to compile these elements into a system for business decisions, for both internal and client needs.

AI ONTOLOGY

Vlad Flaks
OWOX - Toronto, Canada

Not every color pattern on a Rubik's cube can be solved, and not every combination of words carries meaning, even when an LLM delivers it with total confidence. As the industry moves from AI that answers to AI that acts, that gap becomes a real business risk. Interest in ontology has skyrocketed over the past year for exactly this reason.

An ontology defines what exists in a domain, what each thing means, and how everything relates. It goes beyond a data model by representing data, logic, and actions together. That structure is what lets AI search, classify, connect, infer, and reason about your information reliably.

In this talk, Vlad walks you through the most promising ways to make your databases not just queryable, but meaningful from a business perspective. He has advocated ontology models in analytics since 2014 and holds a 2018 patent on implementing ontological models for natural language queries.

Jordan Peck
Snowplow - Leeds, UK

AI READINESS, AI DELIVERY, AND AI USAGE

Yehoshua Coren
Analytics Ninja - Jerusalem, Israel

A presentation about AI, made by AI, for AI, and delivered by AI. For the first time in the history of Superweek, the "Epic Legend" Rabbi Yehoshua Coren will be using military grade Boston Dynamic robot dogs to deliver a revolutionary presentation about AI powered entirely by BigQuery using Pub/Sub reverse ETL streaming realtime messaging. All metrics presented will be exclusively AI driven with the strictest of guardrails employed via AI generated semantic layer. (It is recommended not to bring magnets in the conference during the presentation).

VIBE CODING NIGHT WITH OWOX 🧑‍💻✨

Vlad Flaks
Toronto, Canada

Ilya Chuklyaev
Mijas, Spain

Join Ilya and Vlad on Tuesday evening for an evening of mind-bending vibe coding, data experiments and probably a few things nobody planned. 🚀

IF YOU CAN TRACK AN AGENT, YOU CAN TRACK A BOT: HOW AGENTIC TRAFFIC IS THROWING A WRENCH INTO OUR ANALYTICS

Jason Packer
Quantable - Columbus (OH), USA

Throughout the history of web analytics, we've tried (and frequently failed) to filter out bots and automated traffic. Now AI agents are on our sites too: ChatGPT, Claude, Meta Muse et al. researching and even buying on behalf of people. With automated traffic now more than half of all web traffic, "filter it out" no longer cuts it.

So we now have three types of traffic: human, agentic and bot. Is it possible to tell them apart? And if we can, what are the use cases for that data? The idea for Telltale Bot, Jason's new bot and agent detection service, actually came out of a conversation last year at Superweek. He'll talk about how detection works, and why bot identification is more relevant than ever.

PROOF IT BEFORE YOU BUILD IT: A STATISTICAL METHOD TO ANALYSE DATA, SEARCHING FOR THE BEST FIT - THROUGH MCP, CLAUDE, AND MIXPANEL

Fosca Fimiani
Multiminds - Aalst, Belgium

A process tried and abandoned ten years ago with UA and Excel, too time-consuming. Today, picked back up as a proof of concept for something bigger: a data offload and warehouse, not something the client would agree to without proof the method worked.

The idea and the process, shown here to be replicated with any set of tools.

WHAT THE HELL HAPPENED BETWEEN WASP AND MANTIS? 20 YEARS OF MEASUREMENT, SURVEILLANCE, AND QUESTIONABLE DECISIONS

Stéphane Hamel
Immeria - Montréal (QC), Canada

20 years ago I created WASP, the very first tag auditing tool designed to reveal what analytics technologies were doing inside the web page. Back then some industry leaders thought web analytics was hard: tags were multiplying, implementations were messy and analysts needed a better way to understand what was happening under the hood.

Fast-forward twenty years. The humble tracker has evolved into an extraordinarily complex ecosystem of advertising platforms, consent systems, identity technologies, server side tracking, data layers, pixels, APIs, browser protections, privacy regulation and, increasingly AI.

So I visited WASP and MANTIS was born.

Looking at today's web through MANTIS feels strangely familiar. The technology has changed dramatically. Some of the problems haven't. And somewhere along the way, an industry built to understand what people did on websites became very good at following people around the internet.

This talk is a personal, technical and occasionally uncomfortable journey from WASP to MANTIS. We'll look at what actually changed, what didn't, what we fixed, what we made considerably worse, and what twenty years of tracking technology tells us about the future of digital analytics.

No vendor bashing. No privacy theatre. No AI-generated predictions about the future of analytics.

Just twenty years of evidence, a browser, and the few uncomfortable questions about the industry we helped build.

Peter O'Neill
ZHS Orchards - Haarlem, The Netherlands

REVERSING DECISION LAUNDERING - THE PATH OF LEAST RESISTANCE

Steen Rasmussen
Steen.fyi - Copenhagen, Denmark

Organizations approach to data are in a lot of cases actually blocking them from getting even a fragment of the value out of the data.

But it doesn't have to be this way. Once we can identify the problems we can push for a solution that works.

THE HANDOVER PROBLEM: BUILDING CLIENT CAPABILITY OR CLIENT DEPENDENCY?

Eva González Vior
Datola - A Coruña, Spain

A new analyst arrives, the agency changes, and rebuilding begins: new tools, a fresh implementation, and another promise that soon we’ll start putting the data to work. Each team improves things in its own way, while the client goes back to square one.

This talk turns a critical eye on our own industry: how our technical preferences, incentives, and project delivery practices can create dependency, even with the best intentions. When is rebuilding necessary, and when are we simply reluctant to understand what we’ve inherited?

A reflection on continuity, knowledge, and autonomy, built around one question: if the client can carry on without us, have we lost an account or have we done our job well?

WHAT'S ENTERTAINMENT GOT TO DO WITH ANALYTICS, ANYWAY?

Josh Silverbauer
Intrepid Digital - Philadelphia (PA), USA

Timo Aden
Digitl - Hamburg, Germany

THE BROKEN BUSINESS OF ANALYTICS CONSULTING

Damion Brown
Melbourne, Australia

Analytics consultancies and agencies tend to be built by excellent practitioners: the technical work is strong, but the commercials and "business stuff" is often an afterthought. All too often, consultancies and agencies find their involvement with a client ends too soon, their engagements aren't consistent or profitable enough, and they're chasing flickers of pipeline, and being reactive without really having a plan for growth. That won't stay viable forever; as an industry, we can and must do better at the business of analytics consulting.

Damion's session looks at where consultants and agencies commonly get things wrong on their approach, their pricing, their messaging and the value they deliver to their clients. Drawing on his experiences of starting, scaling and selling the Australian analytics agency Data Runs Deep, Damion works through the structural reasons analytics is hard to sell profitably, the traps he fell into and that you don't need to, and how to think of your consultancy as more like a business.

ANALYTICS AS CODE: WHY YOUR DATA GOVERNANCE BELONGS IN GIT

Julien Coquet
Montélimar, France

We treat infrastructure as code. We use Git for everything. But analytics strategy? Still living in Slack threads and outdated spreadsheets or Google Docs.

This talk shows why your tracking plan belongs in Git: version controlled, reviewable, and traceable. Using a YAML-based governance approach, we'll walk through a practical framework for treating analytics like the infrastructure it is.

During this presentation, Julien Coquet will show you how to:

Kelly Wortham
Analytics Ninja - Springfield (MO), USA

IT'S THE PROCESS: HOW FIXING BROKEN WORKFLOWS TURNED ANALYTICS FROM REPORTING INTO REAL BUSINESS IMPACT

Anna Bettina Kiss
TESCO - London, UK

In this practical case study from Tesco, Anna shows how the real bottleneck isn't data quality or tools, but broken processes between analytics, product, and engineering and how consistent workflows, clear ownership, and better prioritisation unlocked faster delivery and better impact.

Most data problems aren't really data problems - they're process problems. Analytics teams often feel stuck - slow delivery, unreliable data, and insights that go nowhere.

If your team produces great insights but struggles to get them acted on, this session will help you identify what's truly holding you back, and how to fix it.

YOUR DATA HAS A SECOND LIFE - AND YOU MAY NOT LIKE IT. STOLEN DATA, DARK WEB MARKETS AND AI-ASSISTED CRIME

Andras Tevesz
Allstars - Gyor, Hungary

Szilard Torok
Allstars - Budapest, Hungary

You spend years building customer knowledge. Someone with stolen access can start making sense of it in minutes.

A leaked customer list, a stolen browser session, a compromised inbox: minutes later, an attacker's AI is mapping relationships, spotting valuable targets and planning the fraud. This session follows your data beyond the breach.

Using documented cases, we follow stolen data from breach to underground market to attack. We separate what AI agents truly change from where claims of autonomy outrun the evidence. And we show how defenders can use the same tools to investigate abuse.

The fallout reaches far beyond security. Customer trust, account integrity and bot-distorted signals all shape the metrics your business relies on. Defenders can use these same tools to investigate suspicious activity and interrupt the chain. In a way, this session continues where Andras' DEF CON 34 workshop left off.

Leave knowing what makes your data valuable to strangers, what access an AI agent should ever have and what proof you need before trusting a metric or an alert.

Ezequiel Boehler
Conductrics - Buenos Aires, Argentina

#SPWK JoftY: SUPERWEEK JAM OF THE YEAR

Josh Silverbauer
Intrepid Digital - Philadelphia (PA), USA

Work with Josh Silverbauer on a custom SUPERWEEK song throughout the week, then perform it together on Thursday. The performance will be recorded.

🏆 THE XIII. GOLDEN PUNCHCARD PRIZE

The rules are simple: demonstrate a digital analytics solution, method, hack or piece of wizardry that goes way beyond the defaults. ⚡ Bring the weird stuff. 👀 The clever stuff. The stuff you'd think you probably shouldn't have spent that much time building. 🤓 The audience decides who gets the true-to-scale Golden Punchcard.

Send your nomination to punchcard@superweek.hu 📩, even in the last minutes. ⏳