As organisations race to embrace real-time capabilities, data streaming is increasingly seen as a cornerstone of modern digital infrastructure. However, beneath the momentum lies a critical question if enterprises are truly ready for streaming, or are they setting themselves up for failure?
In conversation with iTNews Asia, Andrew Sellers, Head of Technology Strategy, Confluent shed light on why the biggest risk in data streaming today is not underinvestment, but doing too much, too soon. According to him, one of the most dangerous assumptions enterprises make is believing they must replicate their competitors.
“They see what others have built and think, ‘I need to do all that right now.’ So they stand up a centre of excellence, invest heavily, and build a centralised service before the first use case even exists.” This approach, he explained, often leads to failure because the platform is forced to serve too many needs at once.
Instead of large-scale rollouts, Sellers strongly advocated a use-case-first strategy. “Find the first really great data product, the first really great use case, and prove value there.” This allows teams to build familiarity with streaming technologies, which are fundamentally different from traditional architectures. Once early success is demonstrated, adoption tends to spread organically within the organisation.
Sellers emphasised that sustainable data streaming adoption is not driven top-down, but through demonstrated value. “The enterprise grows itself. It happens organically once people see what’s possible.”
This gradual expansion reduces risk, avoids over engineering, and ensures that investments are aligned with real business needs.
Limitations holding enterprises back
A persistent myth around data streaming is that it is too expensive compared to traditional batch processing. Sellers argued that this perception is outdated. “If you look at total cost of ownership today, you can actually do streaming more effectively than batch,” he added.
However, he warns that delaying adoption can create bigger financial challenges later. “The real cost comes when you build everything in a batch first and then try to switch. That’s when it becomes hard and expensive.”
Even with the right intentions, enterprises can misuse data streaming technologies. “Any distributed system becomes counterproductive if you use it in a way it wasn’t designed for,” he added.
Sellers noted that teams sometimes force patterns or architectures that conflict with how streaming platforms are meant to operate, leading to inefficiencies. This reinforces the importance of starting small and learning through real use cases rather than attempting large-scale implementations upfront.
Beyond overbuilding, Sellers highlighted another common pitfall, fragmentation. Some organisations allow multiple teams to independently deploy their own streaming systems, creating governance and cost challenges later. The ideal approach lies in balancing a centralised platform with organically growing use cases.
While many enterprises track technical metrics, Sellers believes success is better measured through usage.
If people are producing and consuming data, the initiative is probably going to work. If consumption is low, that’s when you’re in trouble.
- Andrew Sellers, Head of Technology Strategy, Confluent
He recommended focusing on adoption to measure whether streaming is delivering actual value rather than just existing as infrastructure.
The bigger picture: Streaming as a foundation
Pilot projects play a crucial role in validating data streaming strategies. “We use pilots to create technical wins showing customers something they couldn’t do before,” he said.
These early successes often become the foundation for broader enterprise adoption, strengthening both the technical and economic case for scaling.
For enterprises exploring data streaming, the message is clear: resist the urge to build for the entire organisation from day one. Start with a meaningful use case, prove its value, and allow momentum to drive expansion. Because in data streaming, success isn’t determined by how much you build upfront, but by how effectively you grow what works.
