Data-driven decisions have been one of the most important promises of digitalization for years – yet only a few companies implement them consistently. In the podcast, Max from Scavenger AI and Florian, Head of Data & Analytics at the TWT Group, discuss this gap: Why data projects fail, how to make them successful, and how AI is changing the entire work with data.
TWT Group: Where Data, Marketing, and Technology Meet
The TWT Group operates between Professional Services (software development, UX, project management) and agency business (marketing, social, SEO, AI). Florian leads the Data & Analytics department and works closely with performance, social, and SEO teams.
His ambition is clear:
“We think data-driven and user-centered – and always under the credo: effective.”
Why Data? The Fascination of Objective Truth
Florian originally comes from performance marketing. What excites him about data is its radicalness:
“I have always been intrigued by objective truth – not the gut feeling of whether an ad is nice.”
Data enables decisions that are directly linked to business goals. This creates clarity: What really works? Which measures contribute to revenue, CLV, leads, or customer engagement?
Why Many Companies Fail Despite Technology
Many companies have good tools and infrastructure. Nevertheless, they struggle to leverage the potentials. For Florian, this has little to do with technology but rather with organization and culture.
“It’s the connection between the technical side and the business side. Data work is always a change topic.”
Often decades of experience (“we’ve always done it this way”) clash with hard facts (“the ad is nice, but it doesn’t perform”). This break is organizationally demanding – and rarely moderated cleanly.
What Works in Practice: Start Small, Scale Quickly
Many data initiatives fail due to their size. Companies want to develop a three-year data strategy right away, define roles, reorganize teams – and lose pace and motivation in the process.
Florian recommends the opposite:
start with small projects
set clear goals
deliver a clean POC
make successes visible and communicate
According to Florian, low-hanging fruits include:
first, simple A/B tests
rethinking marketing efficiency (not just clicks & impressions, but value contributions like CLV)
visualizing the customer journey for the first time
automating standard reports – ideally without Excel
retrieving first insights via AI
These manageable steps create quick successes and political acceptance.
How Successful Data Projects Are Anchored in Companies
A common mistake: A good POC is packaged in a 60-page presentation and gets lost in approval loops. Florian recommends instead:
“A one- or six-slider is enough. What was the question? What came out of it? What does it bring for the business?”
The technical depth is irrelevant at this point. What matters is that stakeholders understand the added value and become curious.
Florian therefore swears by an Open Door Policy: Anyone interested should come and be allowed to ask questions. This ensures a natural, organic flow of knowledge – far more effective than top-down coercive measures.
Data Ownership: The Pivot of Every Data Strategy
The most important sentence in the entire conversation is probably this:
“All discussions stop when you name a person or team that is responsible.”
Without clear data ownership, there are:
endless KPI discussions
differing definitions of the same metric (“What is a lead?”)
poor data quality
lack of decision-making authority
delays in analyses
Data ownership defines:
What does a metric mean?
Who is responsible for quality?
Who makes decisions based on the data?
Flo always boils it down to the point: Nothing is more important if a company wants to work seriously in a data-driven manner.
What an Ideal Reporting Setup Looks Like
Florian thinks in three levels and links them clearly together:
1. Strategic Level (C-Level)
The “North Star” – a metric that describes whether the company is growing healthily.
“If this one number grows, the company is healthy.”
2. Tactical Level (Management)
OKR tracking, performance monitoring, achievement of individual team goals.
3. Operational Level (Departments)
Detail metrics necessary for daily management.
(CPC, CTR, scroll depth, cart abandonment, etc.)
This structures data work clearly and removes its diffuse character.
What AI Can Realistically Do Today – And What It Cannot
AI is massively changing data work. For Florian, AI is primarily an accelerator, not a replacement.
“I no longer need to be an analyst today to answer initial questions.”
AI can:
automatically derive initial analyses
optimize campaigns
vary creatives
synthesize data
automatically generate standard reports
But: AI is extremely dependent on the existing data base.
“With poor data, AI becomes a chaos amplifier.”
Tools do not replace a clean data strategy. Poor data leads to poor recommendations – whether human or machine.
Looking to the Future: Agentic Systems & Invisible BI
Florian is convinced that we will see a massive leap in the next three years:
Agentic systems that handle operational marketing tasks themselves
prescriptive analyses that provide concrete action recommendations
Invisible BI that proactively sends insights to teams
This means:
On Monday mornings, a product manager automatically receives a summary of the most important KPIs including action suggestions – before he has even asked for it. BI becomes invisible and yet omnipresent.
What Companies Should Change Tomorrow
If companies could improve just one thing immediately, it would be this:
“Define data ownership. That’s the basis for everything.”
With clear responsibility, misunderstandings, KPI discussions, and reporting chaos disappear. Teams can finally make autonomous decisions – and work in a data-driven manner.