- Real-time infrastructure costs meaningfully more to build and maintain than batch pipelines.
- Most business decisions don't actually require sub-minute data freshness to be effective.
- Operational, in-the-moment decisions are where real-time genuinely earns its cost.
- Batch reporting run nightly or hourly is often sufficient, and dramatically simpler.
Real-time is a genuine engineering investment
Building real-time data pipelines requires infrastructure — streaming platforms, event processing — that costs more to build and maintain than a batch pipeline that runs on a schedule. That cost is worth paying for the right use-case, but it's not free, and it shouldn't be the default without a reason.
Most decisions don't need sub-minute freshness
A weekly revenue review, a monthly board report, a quarterly retention analysis — none of these need data that's seconds old. Batch reporting, refreshed nightly or a few times a day, is completely sufficient for the majority of business reporting use-cases, at a fraction of the infrastructure cost.
Operational decisions are where real-time earns its keep
Fraud detection, live inventory management, an in-the-moment customer support alert — these are decisions that lose most of their value if the data is even an hour old. That's the category where real-time infrastructure investment clearly pays for itself.
The middle ground: near-real-time
Many use-cases that feel like they need real-time actually just need 'fresher than daily' — hourly or every-few-minutes refresh, which is much cheaper to build than true streaming infrastructure. Checking whether near-real-time actually satisfies the need avoids over-engineering a batch problem into a real-time one.
Match the investment to the decision speed
The right question isn't 'should we have real-time analytics' in the abstract — it's 'which specific decisions in our business lose value if the data is an hour old.' Answering that narrows the real-time investment to where it earns its cost, and leaves everything else on a much simpler batch schedule.