The Future of the Connected Customer Stack | Zack Wenthe & Adrian Bolosan | Signal & Noise LIVE at Advertising Week 2026

Full analysis

This is a vendor-sponsored panel from Advertising Week, so the loudest claims come from people with something to sell. Zack Wenthe runs product marketing at Tealium, a tag management company. Adrian Bolosan runs ISV go-to-market at Databricks. Strip away the sponsorship and one real argument remains: where should a live marketing decision get computed, and what does AI do to the economics of buying identity data. That's it. Nothing here moves a market this week.

How hard is this to undo? Nothing to undo. This is a read on where enterprise marketing stacks are drifting, not a buy decision. You can act on any of it incrementally.

What's actually being decided: Whether to keep paying for bulk identity-graph subscriptions and per-seat SaaS, or shift toward on-demand, use-case pricing. And where to put your real-time personalization logic.

What sets the deadline: Nothing. No contract date, no shutdown, no price change named. Move when your renewals come up.

The Skeptic

Wenthe says one-to-one personalization is "bullshit" and cohorts are enough. He's right, and he's been right for a decade. What's new? Nothing. Bolosan's counter, that "agentic" finally gets us to treating a person as a person, is the exact pipe dream Wenthe just buried, dressed in a 2026 word. Watch the incentives: Wenthe sells the data-feeding layer, so he wants you to believe cohorts win and the signal quality matters. Bolosan sells the warehouse, so he wants you to believe you need deep historical intelligence married to the edge. Both can't be the center of your stack. The "OpenAI decisions" product Wenthe name-dropped doesn't appear to exist. Treat the whole panel as positioning.

The Builder

The one genuinely useful claim: decisioning round-trips to a data warehouse are too slow for in-session personalization, so put the millisecond decision at the ad platform or email tool and let the warehouse enrich the next interaction. That's correct and it's actionable. If you're waiting on a warehouse query to decide what to show someone mid-session, you've already lost the session. Build the fast path at the channel, feed history back asynchronously. The "marketing engineer" framing is real too: the person who wires agents, models and data together is now a distinct role from the marketer who just queries data in plain English. If you're hiring, that split is worth naming in the job description.

The Compute Pragmatist

The identity economics shift is the part with actual dollars attached. Match-rate testing, where you check how many of your customer records a data provider can match to their graph, used to take weeks of bake-offs. Bolosan says it now runs in minutes on a platform like Databricks. If true, switching between Experian, Acxiom, and TransUnion gets cheap, and cheap switching kills the bulk license-everything-upfront model. You stop buying the whole graph and start paying to fill specific gaps on demand. CPG brands with thin first-party data still need the big graphs. Everyone else gets pricing leverage they didn't have. That's a real move in how identity gets bought.

The Open-Source Advocate

The "SaaSpocalypse" debate is where the panel is softest. Wenthe compares SaaS to bread: we can bake it, we still buy it. Fine, but that dodges the live question. Per-seat licensing is under real pressure from single-purpose tools built in hours with AI-assisted coding. The panel admits it, then retreats to "enterprise governance and security keep the big platforms safe." Both things are true and they don't contradict. Salesforce survives for the system of record. The reporting dashboard, the campaign one-off, the internal tool you'd have bought a seat-based add-on for? That gets vibe-coded and thrown away. The seat count shrinks at the edges while the core platform holds. Nobody on this panel wants to say the number, because both guests sell platforms.

Where they actually disagree

The real split is Wenthe versus Bolosan on where intelligence lives. Wenthe wants decisioning at the channel because latency rules everything. Bolosan wants the warehouse married to the edge because history makes the next decision smarter. For a buyer this isn't academic: it decides whether you invest in warehouse-side modeling or channel-side speed. The panel landed on "composable decisioning, different decisions in different layers," which is correct and also the answer that sells both their products. Convenient.

The second split: does cheap match-rate testing actually break identity pricing, or do the big three graph providers just repackage into usage tiers and hold their margin? Faster testing lowers switching cost. It doesn't automatically lower what Experian charges. The provider response is what settles this.

What it hinges on

Two things. One, is match-rate bake-off time really collapsing from weeks to minutes across providers, not just inside one demo on one platform? If yes, buyers get leverage at renewal. Two, does the per-seat model actually erode, or do vendors just rename seats as "capability" and keep the revenue? Before you act on either, do one thing: at your next identity renewal, ask for on-demand, gap-fill pricing and see if the provider blinks. That test costs you nothing and tells you whether the economics have really moved.

Prediction: At least one of the three major consumer-identity providers (Experian, Acxiom, TransUnion) will publicly launch an on-demand or usage-based identity-matching product, priced per query or per match rather than as a bulk annual graph license, before the end of Q3 2027.

Confidence: Medium — the switching-cost collapse is real, but incumbents move slowly and may defend bundles.

Why: Bolosan, who worked at Experian, says match-rate testing has collapsed from weeks to minutes on cloud data platforms, and he specifically cites Experian already offering accuracy and reach "dials" to clients. When the cost of testing a rival's match rate drops to near zero, the bulk upfront license loses its main defense, which was that switching was too painful to bother. Providers facing that pressure either cut price or unbundle into pay-per-use tiers to keep customers from shopping around every quarter, and unbundling protects margin better than an across-the-board cut. The opposite outcome, all three holding firm on bulk-only licensing, requires them to ignore that their biggest customers can now comparison-shop in an afternoon.

Revisit by 2027-09-30: We're right if Experian, Acxiom, or TransUnion announces a per-query or usage-based identity-matching offer by then. We're wrong if all three still sell identity resolution only as bulk annual graph subscriptions.