Why 3 Identity Paths & $100 Orders Redefine SaaS CDPs

SaaS platforms and customer data infrastructure teams tend to think in events and schemas, but Rokt mParticle's latest piece argues the real product challenge is identity: ecommerce has authenticated accounts, subscriptions have identity by design, and physical channels rely on loyalty. For CDP and personalization builders, the article maps three identity paths to different data activation and recalculation requirements—for example, turning a missing weekday purchase into a cadence-break signal.

On August 24, 2026, MarTech and Search Engine Land published a contributed piece from Rokt mParticle arguing that performance marketing's reliance on transaction-level conversion data misses the most important variable: the history of the customer relationship. The article opens with a simple comparison. A $100 order from a first-time buyer and a $100 order from a customer who purchases every two weeks look identical in a conversion feed. Yet they represent fundamentally different retention economics, lifetime value, and appropriate marketing response. More data fields—product choices, basket value, discounts, channel, location—do not solve this unless the transaction is tied to a persistent profile.

“

A $100 order from a first-time buyer and a $100 order from a customer who purchases every two weeks look identical in a conversion feed.

The framework splits identity challenges into three categories. Ecommerce retailers can authenticate accounts. Subscription businesses have identity by design because the relationship requires an account. Physical retail, grocery, and quick-service restaurants face the harder problem: transactions complete whether or not the customer identifies themselves. There, loyalty programs serve as the bridge that attaches an in-store or drive-through purchase to a known individual. However, Rokt mParticle cautions that loyalty program enrollment is not the same as identity coverage. The metric that matters is the share of transactions that arrive with a known customer attached, not raw membership counts.

This distinction has practical consequences for marketing and data teams. A brand could have millions of loyalty members but still lack coverage on most purchases, meaning most transaction signals remain unusable for personalized paid media. Derived attributes should not all be refreshed or activated in the same way. Purchase cadence is an illustrative case. Knowing that someone bought coffee this morning is useful for a short window; knowing they usually buy every weekday morning creates a different signal. When that customer skips three days, no new event arrives, yet the absence itself is the meaningful signal—for a lapse campaign, a win-back offer, or a reprioritized audience segment. Cadence signals only work if they are recalculated often enough to capture the non-event.

The strategic implication is that connected history transforms the unit of analysis from the transaction to the person. This aligns with the broader movement in martech toward first-party data and customer data platforms, but the article frames it specifically for performance marketing: paid media decisions should be based on what a conversion means in the context of prior behavior, not on conversion value alone. It suggests data engineering and marketing teams need to collaborate on identity resolution, event schemas, and attribute refresh SLAs. The absence of an event becomes a first-class signal only when the data layer understands the customer's normal pattern.