Response Memo · Prepared for John Bond

The technical read on Vision Bridge Prism, corrected.

Horizon's internal assessment is accurate on commercial structure and honest on two real limitations. It is wrong on the four claims that matter most: attribution methodology, deduplication, measurement basis, and scalability ceiling. This memo answers each directly, with what's actually in the stack.

SubjectVision Bridge Prism — Horizon Assessment
Attendees referencedMichael Cohen, Kevin McGehee, Katie Comerford
Prepared forBill, Bob, John Bond
01 — Where the assessment is right

Two limitations, conceded without argument.

No defensiveness on these. They're real, current, and worth naming plainly before the corrections below.

Conceded

No offline or physical-retail conversion path today.

A genuine gap for omnichannel brands with in-store sales. Not disputed.

Conceded

Agentic optimization currently recommends. It does not yet execute buys automatically.

Publisher and deal-ID moves are surfaced to a human, not auto-executed. Roadmap item, not shipped.

Accurate description

The commercial model is CPM buying with a CPA guarantee, margin captured on the spread.

Correctly described. Standard for guaranteed-performance CTV. Nothing to correct here.

These two gaps are exactly why this is a partnership and investment conversation, not a vendor pass/fail.

A company with the measurement infrastructure shown in Sections 02 through 06, that is still early on offline conversion capture and automated execution, is a growth-stage build, not a ceiling. The right capital and the right enterprise partner close exactly those two gaps. That is the opportunity in front of Horizon, not a reason to walk.

02 — Where the assessment gets the mechanism wrong

Four claims, corrected against what's actually built.

Each pairing below is the memo's language on the left, the actual mechanism on the right.

Horizon's claim

"Vague and arbitrary methodology, limited statistical validation."

Treats VisionBridge as a single self-reporting layer built only to justify the CPA buy.

Corrected

Media mix modeling and registered holdout tests are already in the stack.

Attribution is one of three documented pillars, not the whole product. A statistical MMM baseline allocates credit independent of the CTV buy, and registered holdout tests measure incremental lift against a true control group — the exact causal-inference discipline the memo says is missing. If it wasn't shown in the room, that's a gap in the meeting, not the platform.

Horizon's claim

"Coordination and deduplication break down without independent enterprise measurement."

Frames dedup as a capability that fails once media mix broadens.

Corrected

Deduplication is the core mechanism, not a missing one.

Cross-channel suppression against Meta, Google, and existing-customer history, before any CTV credit is issued, is the entire reason the product exists. Double attribution is the problem it was built to solve. It's the first check on every order, not a limitation that appears at scale.

Horizon's claim

"Heavy reliance on UTMs and log files, fragile and identity-gap prone."

Implies the guarantee ledger is built on the same fragile layer as journey visualization.

Corrected

Billing runs on exact order-ID reconciliation, not UTMs.

UTMs and log files support journey visualization only. The guarantee ledger is built on deterministic order-ID matching against the client's own commerce data — a separate, harder layer than the one being critiqued.

Horizon's claim

"Commercial packaging rather than proprietary technological breakthroughs."

Frames the offering as pricing structure without a real technical edge.

Corrected

A live integration collapses conversion time by an order of magnitude.

Impression-level CTV exposure streams directly into Meta's retargeting layer. That collapses average time-to-conversion from 50-plus days to under 3. That's a running pipeline, not a pricing structure — and it's the one differentiator the assessment didn't mention.

03 — Scalability and enterprise fit, corrected

The $2–5M ceiling reflects go-to-market, not infrastructure.

Horizon's own enterprise-complexity framing, reworked against what's actually true of each factor.

Enterprise factorHorizon's claimCorrected
Broader media mixes & brand investment VisionBridge only justifies programmatic media, doesn't evaluate holistic incrementality MMM baseline plus registered holdout tests already measure incrementality independent of the CTV buy — see Section 02.
Multi-agency ecosystems Coordination and dedup break down without independent enterprise measurement Cross-channel dedup against Meta, Google, and customer history is the founding mechanism, not a scale failure point.
Offline sales & omnichannel tracking UTMs and log files can't support offline or physical-retail paths Conceded as-is. Real gap today. Billing itself does not run on UTMs (see Section 02), but offline conversion capture is not yet built.
Measurement philosophy Causal inference and MMM directly conflict with self-justifying attribution MMM and holdout testing are the measurement philosophy already in place. The conflict the memo describes doesn't exist in the current build.
04 — The differentiator not in the memo

A running pipeline, not a pricing structure.

50+ → <3
Days to conversion, before and after impression-level CTV exposure is streamed into Meta's retargeting layer

This is the technical breakthrough the assessment says doesn't exist. CTV impressions aren't a silo that gets reported on after the fact — they feed Meta's audience layer in real time, collapsing the gap between exposure and purchase from a month and a half down to under three days. Every other claim in this memo is defensible on paper. This one is visible in the account the moment it's turned on.

05 — Live proof point: Boost Mobile

This is not theoretical. It is running in an active Boost Mobile pilot now.

The Boost Mobile x VisionBridge launch plan and attribution methodology, dated August 11, 2026, puts the operating rules, validation dependencies, and platform plumbing in writing before scale.

Horizon’s claim

“Self-justifying attribution” and a vague, arbitrary methodology.

The concern is that the system can loosely credit itself after the fact, with no defined basis for the decision.

Live pilot rule

Attribution is pre-click only, with a rules-based model open to client input.

Credit is allocated only to the pre-click touchpoint: the impression that ran before a platform click. Anything after that click receives zero attribution credit. The final thresholds, including haircut percentages and credit-share splits, are configurable and open to Boost’s input before launch. That is a conservative, defined, negotiable ruleset, not an elastic one.

Horizon’s claim

Independent enterprise measurement is absent, and the product has a $2–5M ceiling.

The assessment treats external validation and cross-platform operating capacity as missing prerequisites for serious scale.

Live pilot plan

Independent validation is a stated dependency, and the API pipeline is ready across four major platforms.

MRC accreditation and independent LTV benchmarking are named dependencies before spend scales. At the same time, platform connector access across The Trade Desk, Meta, Google Ads, and DV360 is confirmed and ready to run through the API pipeline. This is a national telecom pilot trafficking programmatic activity across four major platforms simultaneously, not a small-DTC-only workflow.

Horizon’s claim

The causal-inference discipline is missing.

The assessment assumes untested audience theories are being presented as proof.

Live pilot discipline

The plan separates hypothesis from proof and names its quality checkpoints.

Device-tenure targeting is explicitly labeled a hypothesis to be tested, not a proven segment. Formal three-month and six-month post-acquisition checkpoints measure retention, product attach, and upgrade or downgrade signals. Labeling an unproven segment as unproven is the methodological rigor and intellectual honesty the assessment says is missing.

06 — Independent third-party validation, already delivered

Horizon says independent measurement is missing. Here it is, live, with two outside firms.

Frøya's CTV lift study is not a VisionBridge self-report. It ran a clean 37-state versus 13-state holdout, then extended it to a 90-day national rollout, cross-validated by two independent geo-lift methodologies: Haus.io and Measured.com. Full report: froyatest.visionbridgetv.com

Frøya CTV Lift Study executive overview dashboard
Phase 01 · 46 days

2.66x Per Capita iROAS, 99.7% confidence.

37 test states vs. 13 holdout states, no VisionBridge ads. Two independent methodologies, Haus.io Per Capita and Measured.com Revenue Share, both confirmed incrementality.

Phase 02 · 90 days

Signal held at 1.59x Platform ROAS, 2x the spend, every month.

Same design carried through the pre-national window. February peaked at 1.78x platform ROAS at 99.6% confidence, the cleanest single read of the study.

Phase 03 · National rollout

1.52x ROAS held the moment holdout markets went live.

Vision Bridge attribution independently validates a ~4.5 day consideration cycle inside the 7-day attribution window, across 1,556 tracked conversions.

Four independent measurement frames that converge on the same answer is the causal-inference discipline the assessment says is absent. It is running, published, and dated.

Closing

One meeting is a thin sample for a strategic-fit call.

The commercial description in Horizon's assessment is accurate, and the two conceded limitations in Section 01 are real, and exactly why this is a partnership and investment conversation. The corrections in Sections 02 through 06 aren't a matter of interpretation. They're what's running in the account today, independently validated by two outside measurement firms. The right next step is putting the measurement lead in a room with Michael, Kevin, and Katie to walk through the holdout design, the dedup logic, and the MMM layer directly, rather than settling it in writing.

Sean Clayton Singulariti