Superweek 2026 - Day Two
- Dan Truman
- Feb 3
- 5 min read
Efficiency, Intent, and the Limits of Cleverness
If Day One was about existential questions – AI, agency scale, purpose and trust – Day Two was about application. How money actually gets spent. How decisions really get made. Where theory breaks once it meets humans, platforms and incentives.
Across MMM, solo consulting, intent, server-side tagging, experimentation and AI agents, a consistent thread emerged: optimisation without judgment is just automation of past mistakes.
Here’s how Day Two unfolded.
Making More Money Without Spending More
Gabriele Franco – Cassandra
How to Make More Revenue with the Same Marketing Budget
Gabriele came armed with numbers, not vibes. Cassandra manages over $500m in ad spend, and the promise was bold but precise: follow the process and brands can see 20–30% ROI improvements without increasing budget.
The deceptively simple questions every marketer asks:
Where should we invest?
When should we invest?
…are almost impossible to answer reliably with last-click data, broken attribution, and a bottom-funnel obsession.
Gabriele reframed the problem.
Strategy and planning don’t come from conversion tracking. They come from incrementality.
Marketing Mix Modelling (MMM) One statistical model across all channels, regardless of funnel stage, using historical investment and sales data to understand influence.
Simulation and Forecasting Budget allocators embedded directly in the model to simulate revenue outcomes.
Incrementality Testing Regional A/B tests layered on top to calibrate and correct MMM bias.
The key insight wasn’t the tooling – it was the loop. MMM gives direction. Incrementality grounds it in reality. Feed those results back into the model and suddenly planning becomes far less speculative.
Seasonality, lag, and carryover effects all matter. Spend windows and effect windows are rarely aligned, especially around moments like BFCM. Strategy lives in those gaps.
The demo landed the point cleanly: compare business-as-usual against a deliberately planned strategy, and the difference is stark.
Efficiency isn’t about cutting spend. It’s about earning the right to spend.
Should You Go Solo?
Anna Lewis – Polka Dot Data
A Candid Look at Independence
Anna’s talk was one of the most human sessions of the day.
Celebrating ten years of Polka Dot Data, she didn’t romanticise independence. Instead, she framed it as a journey – literally – using a flying metaphor.
Pilot: the founder
Cabin crew: your team
Mechanics: freelancers and specialists
Passengers: clients
Her core message was refreshingly grounded: businesses are built on people, not structures.
She talked openly about hiring early, valuing transferable skills, and the reality that confidence fluctuates. What mattered most, according to the room, wasn’t perfection or productivity – it was connections and determination.
Other hard-won truths:
Most work still comes from word of mouth
Speaking and community matter more than marketing funnels
Progress beats perfection – aim for “perfect enough”
Anna also pushed people to be intentional about how they want to work. Freelancer. Contractor. Agency. Product. Each spreads risk and responsibility differently.
Independence isn’t for everyone. And that’s okay.
From Clicks to Causes
Steen Rasmussen
From What to Why? Deconstructing User Intentions
Steen went straight for a long-standing failure mode in analytics: we are excellent at describing behaviour and terrible at explaining motivation.
Clicks, pageviews and conversions tell us what happened. Stakeholders want to know why.
His framing was disarmingly simple.
Around 85% of users arrive knowing why they’re there:
Buy
Research
Find contact details
Compare options
If we can’t articulate the dominant intentions on a site, no amount of event tracking will save us.
Momentum and intention are distinct. Confusing them leads to dashboards that look busy but answer nothing.
The challenge isn’t data scarcity. It’s clarity.
(Also: Steen promised everyone a curling ball in 2027. Zoli agreed.)
Reading the World, Not Just Your Website
Daniel Waisberg – Google
Google Trends at Scale
Daniel reminded everyone that not all insight starts on your domain.
GA tells you what happened on your site
Search Console tells you what happened in Google Search
Google Trends tells you what’s happening in the world
The most important clarification: Google Trends is not volume. It’s a probabilistic index of relative interest.
Interest ≠ demand.Interest ≠ clicks.Interest ≠ revenue.
But interest often signals intent before volume appears.
The new Trends API standardises scaling across time and queries, making proper comparison possible. Still, client education is critical. Peaks matter. Context matters.
Trends tell you when to look deeper, not what to conclude.
One excellent line to steal:
Interest suggests intent. Volume confirms it.
Also confirmed: pineapple on pizza is acceptable to Daniel.
A Reality Check on Server-Side Tagging
Simo Ahava – Simmer
Server-Side Shenanigans
Simo did what Simo does best: calmly dismantled an industry sales pitch.
Server-side tagging is often sold as a way to “get the data back”.
Simo’s position was blunt: that framing is wrong, and often vendor-centric.
The real benefits of server-side have little to do with quantity:
Performance
Security
Enrichment
Data minimisation
Durability in first-party contexts
He raised uncomfortable but necessary questions:
Can implementations actually be audited?
Who truly has control?
Are we optimising for vendors or users?
“Control,” he argued, “is an illusion.”Browsers, ad blockers, vendors, agencies, users – all exert influence at different layers.
His challenge to the community was clear:
Stop parroting vendor talking points
Read vendor requirements critically
Design for transparency, control and agency
Prioritise users without pretending clients don’t matter
Quality over quantity isn’t a slogan. It’s a responsibility.
When Personalisation Actually Matters
Matt Gershoff – Conductrics
When 1 + 1 ≠ 2
Matt closed one of the loops opened on Day One: complexity for its own sake is rarely justified.
His talk focused on interactions – between tests, and between users and tests.
Sometimes effects are additive. Sometimes they’re multiplicative. Often we assume the latter without evidence.
His practical test:
Run a simple (additive) model
Run a complex (interaction) model
If the error reduction isn’t meaningful, complexity isn’t worth it
Truth matters less than utility. If the simpler model explains the world well enough, use it.
This was a quiet but powerful reminder: sophistication should be earned, not assumed.
AI Agents, Again – But This Time for Real
Jordan Peck and Jon Su – Snowplow
Tracking AI Agents in 2026
The day ended where many talks now begin: AI agents.
Jordan and Jon made a useful distinction:
Internal agents (high control, high visibility)
Customer-facing agents (low control, low visibility)
The latter are the problem. ChatGPT, Gemini and others act as opaque intermediaries.
You don’t control the experience. You barely see the data.
Their conclusion was refreshingly honest: there is no clean, universal solution. Tracking only works when you own the interface. Otherwise, you’re inferring behaviour without context – and “why” remains elusive.
At least they didn’t pretend otherwise.
Day Two, Distilled
Day Two wasn’t about shiny futures. It was about discipline.
Spend efficiency comes from planning, not dashboards
Independence is a people problem, not a structure problem
Intent beats interaction counts
Trends signal curiosity, not certainty
Server-side is about responsibility, not recovery
Complexity must justify itself
If Day One asked “what kind of industry are we becoming?”,Day Two asked “are we actually equipped to run it well?”
Tomorrow brings the answers.
But first - songs.