Share with your CMO
Most martech stacks are over-engineered for a speed no one actually needs. Writing at CMSWire, analyst Scott Brinker-adjacent contributor argues that “real-time data” is a vendor-coined term borrowed from 1990s computing, and that marketers have been paying a compounding infrastructure tax ever since. The actual requirement for most use cases, from retargeting to lifecycle segmentation to weekly campaigns, is data that arrives before the decision moment, not data that arrives instantly. Only a narrow set of cases, fraud detection, cart abandonment, in-session personalization, and consent or suppression signals, genuinely require near-instant data pipelines.
What this means for your business
Whether this argument hits your budget depends on how your data infrastructure was sold to you. If your CDP, customer data platform, or data warehouse vendor pitched “real-time” as the headline capability during procurement, there’s a reasonable chance you’re running streaming pipelines, continuous identity resolution, and always-on transformations across use cases that a nightly or hourly batch would serve just as well. That gap between architectural ambition and actual decision latency is where the budget leak lives, and it compounds quietly.
The “just-in-time” framing the piece advances is genuinely useful, even if it comes from a practitioner with obvious affinity for tidy frameworks, which can tilt toward clean taxonomies over the messy reality of mixed-latency data environments. The harder operational truth is that most organizations don’t have a clean map between use cases and data pipelines. The streaming infrastructure often exists because it was easier to build one fast system than to govern a tiered one. Migrating to freshness-based service level agreements, essentially a contract per use case specifying how stale data can be before a decision degrades, requires data team discipline that is frequently absent.
The CMO who should be worried isn’t the one running a lean stack on a CDP with hourly refreshes. It’s the one who approved a “real-time” upgrade eighteen months ago, saw the vendor demo the fraud and in-session use cases, and then watched the rest of the organization treat every pipeline as equally urgent. If your next vendor renewal includes a line item for streaming data infrastructure, the right question to walk in with is: which of our active use cases actually fail with hourly data? If the answer is fewer than three, you’re probably funding a vanity architecture.
Concept deep-dive: Data Freshness SLA
A data freshness SLA, service level agreement, is a per-use-case contract specifying the maximum acceptable age of data before it degrades a decision. Think of it like a sell-by date applied not to groceries but to a customer record: a fraud signal might spoil in two seconds, while a lifecycle segment might stay accurate for 24 hours. The business value is that it lets architecture teams tier infrastructure by actual need rather than building the whole stack to the tightest possible tolerance.
Based on reporting from The Real-Time Data Myth Draining Marketing Budgets, originally published 2026-08-31 20:02:00.
