Why Channel-Centric Data Is Costing You Customer Moments

Your customer just browsed a product on your app, abandoned a cart, received a retargeting ad, clicked through to your website, and then messaged your support team — all within 90 minutes. Your data infrastructure logged five separate events across four separate teams. None of them talked to each other.

The Channel Model Was Never Built for the Customer

As Tealium’s Zack Wenthe articulates pointedly: customers don’t think in channels. They move fluidly across touchpoints, experiencing a single relationship with your brand — while your org chart experiences five parallel campaigns with five separate KPIs. The result is a data architecture that is operationally tidy and strategically blind.

This matters beyond UX. When your first-party data is siloed by channel, consent management fractures too. A customer who opts out of email comms in your CRM may still be targeted via your paid media pixel, because the opt-out never propagated. In Southeast Asia, where PDPA in Thailand, PDPC in Singapore, and Indonesia’s PDP Law each carry their own enforcement posture, that fragmentation isn’t just an experience failure — it’s a compliance exposure.

The fix isn’t a new tool. It’s a different organising principle: build your data layer around moments, not channels.

What Moment-Centric Architecture Actually Looks Like

A moment is any interaction where a customer signal carries intent — a search, a scroll, a purchase, a complaint, a survey response. When you unify these signals at the event level, in real time, rather than aggregating them into channel-level reports after the fact, you get something genuinely useful: context.

Kyle Salomon’s candid retrospective on his time as a head of analytics at dbt is instructive here. One of his core admissions is that his team built reporting infrastructure that answered the questions leadership asked, rather than the questions that would have changed decisions. Channel-siloed data does exactly this — it answers “how did email perform?” when the real question is “what does this customer need next?”

In practice, moment-centric architecture requires three things: a unified identity layer that resolves the same customer across app, web, and offline; event schemas that capture behavioural intent signals rather than just clicks; and a consent graph that travels with the identity, not the channel. Brands running on Grab’s or Shopee’s ecosystem already see this model in action — those platforms activate personalisation at the moment level because their identity resolution is native to the product, not bolted on.

---

Consent as a Data Quality Signal, Not a Legal Checkbox

Here’s the reframe that changes everything: when you build your data architecture around moments, consent stops being a compliance overhead and starts being a data quality filter.

Consider what happens when a customer actively opts into SMS communications. SurveyMonkey’s new Salesforce integration — enabling automated, personalised SMS survey invites triggered directly from CRM events — is a small but telling example of this. The brands getting signal quality from that channel aren’t the ones with the biggest lists; they’re the ones whose customers chose to be there. Opted-in SMS respondents in the context of a post-purchase survey aren’t just legally safer to contact — they’re meaningfully more likely to give you data you can act on.

This is the consent-as-signal principle: high-consent data is higher-quality data. It represents people who have a relationship with your brand, not people who were scraped, assumed, or inherited from a third-party list. For Southeast Asian markets where third-party cookies have been unreliable for years (Safari ITP killed them quietly; Chrome’s deprecation just formalised the funeral), first-party, consent-rich data is already the only durable foundation.

The Hallucination Problem in Your Analytics Stack

There’s a subtler risk worth naming. As more teams use AI-assisted tools for data interpretation, translation, and insight generation — particularly in multilingual markets like the Philippines, Indonesia, or Malaysia where customer feedback arrives in Tagalog, Bahasa, and English simultaneously — the accuracy of what your models tell you about customer intent can quietly degrade.

Research published on Towards Data Science by Aleksandr Gapchenko highlights a low-cost method for detecting hallucinations in neural machine translation using attention misalignment — essentially flagging where a model’s token-level confidence diverges from the source signal. The strategic implication for analytics teams is broader than translation: any model that summarises or interprets customer signal without surfacing its own uncertainty is introducing noise into your insight layer. Teams relying on LLM-generated summaries of customer feedback, support tickets, or survey responses should be asking whether their tooling flags low-confidence outputs — or simply presents them as fact.

Moment-centric data architecture doesn’t just help with activation. It creates the ground truth that makes AI-assisted interpretation more reliable, because the underlying events are clean, timestamped, and consented.

Key Takeaways

• Restructure around moments, not channels: Map your data architecture to the sequence of customer intent signals — not to your org chart — and consent propagation becomes a single solved problem rather than a per-channel afterthought.
• Treat opted-in data as a quality tier, not just a compliance requirement: High-consent touchpoints like triggered SMS surveys consistently produce higher-signal feedback; prioritise building those flows before scaling reach.
• Audit your AI interpretation layer for uncertainty transparency: If your analytics or insight tools don’t surface confidence levels when summarising customer data — especially across languages — you’re making decisions on hallucinated ground truth.

---

The channel model persists because it’s easy to govern, not because it’s right. As real-time data infrastructure matures and Southeast Asian regulators tighten enforcement around consent, the brands that have already reorganised their data thinking around customer moments will find compliance, personalisation, and trust compounding together — rather than pulling in opposite directions. The question worth sitting with: if your customer experienced your brand as one continuous relationship, would your data be able to tell that story?

---

At grzzly, we help brands across Southeast Asia build first-party data programmes that are consent-first by design and moment-ready by architecture — not retrofitted for compliance after the fact. If your current stack is channel-organised and you’re feeling the activation ceiling, we’d enjoy that conversation. Let’s talk