The dual-PM split Ben is in is the named anti-pattern, and the drift is the design working

The question

"What frameworks exist for a dual-PM split where a front-end PM holds the sponsor relationship and a technical PM holds the roadmap substance, without the front-end PM drifting into roadmap ownership?"

Context: Ben is technical PM for CAF inside phData's DIE org; Todd is "front-end PM" and a direct report to Vincent (exec sponsor). The split is still TBD. The CAF roadmap v0 pack pre-mortem flagged exactly this drift risk twice (2026-07-05, 2026-07-06) and it is still unresolved.

What we already know (from the vault)

• The risk is already named in our own pre-mortem, verbatim: "Todd reports directly to Vincent and the PM split is still TBD. If the split lands wrong, 'front-end PM' can drift into owning the roadmap surface." Our stated mitigation was "circulate L1 jointly with Todd (3.4 makes him a co-owner of a lane early)" — [[2026-07-06-caf-die-roadmap-v0-pack]]. This brief concludes that mitigation is probably backwards. See Convergences.
• We already proposed the correct-shape fix without knowing it matched the literature: "Clean substance cut: founder owns the engine (CAF, brigade pattern, contracts, backend); Todd owns the user-facing front end (Snowflake app surface). Founder defines it because he understands what there is to divide." — [[2026-07-05-caf-technical-pm-role-framing-and-roadmap-kit]].
• The asymmetry is explicit in the vault: Ben's reporting line runs Taylon → Elizabeth, not Vincent. Todd's runs direct to Vincent. Ben holds substance; Todd holds the short line to the sponsor ([[2026-07-05-caf-technical-pm-role-framing-and-roadmap-kit]]).
• Ben's roadmap thesis is seam-shaped, not spoke-shaped: "the platform IS its seams. Spokes have owners; the seams are where value crosses and where nobody is accountable today. This roadmap assigns every seam a contract, an owner, and a date." ([[2026-07-06-caf-die-roadmap-v0-pack]]). This turns out to be the strongest structural defense available to him — see Synthesis.
• Timing pressure is real and cuts against slow org-politics fixes: Anderson's Snowflake site gates money release, 321GO is already deployed at two clients, and the pre-mortem separately flags "the hub gets claimed de facto" if a running implementation ships before the Fabric spec circulates ([[2026-07-06-caf-die-roadmap-v0-pack]]).

What the web says

• The exact split described in the question has a name and it is an anti-pattern, not a framework. Marty Cagan describes the "two in a box" PM as "a product manager that interacts with customers and stakeholders, but a different person (e.g. product owner) that interacts with the engineers," and calls splitting responsibility for the three constituencies "a very serious mistake." His stated principle: "you need to protect the product manager's direct access to these three constituencies" (users/customers, business stakeholders, engineers). svpg.com/two-in-a-box-pm
• Cagan endorses no structural device that makes the split work. His prescribed remedies for PM overload are all load-shedding that preserves access: hire project managers, product designers, user-research admin support, QA engineers, or right-size the team's scope. He does not offer a decision-rights artifact, meeting design, or escalation path that rescues the constituency split. svpg.com/two-in-a-box-pm
• Cagan's related PM-vs-PO writing is weaker than expected as a source here. In "Product Manager vs. Product Owner Revisited" he argues "the product manager is also the product owner" and that "the product owner responsibility is just a very minor part" of the job — but the piece does not discuss SAFe's organizational model, and does not delineate what a PM may delegate vs must personally own. The dispatch brief's expectation that this article addresses SAFe's split is not borne out. svpg.com/product-manager-vs-product-owner-revisited
• The decision-rights frameworks (RAPID, and its DACI/RACI cousins) are the only named frameworks that speak to the mechanism — and they are declarative, not enforcing. Rogers & Blenko's RAPID (Bain, June 28 2006) assigns Recommend ("making a proposal or offering alternatives"), Agree (veto), Input ("relevant information"), Perform (execution), and Decide — "the single point of accountability who commits the organization to action." The framework's core move is consolidation: "it comes down to one person who must decide." Veto power is deliberately rationed: "only a few should have such veto power: legal counsel, for certain decisions, or the head of an affected unit." bain.com/insights/who-has-the-d
• RAPID's evidence base, read at the primary source, is thinner than its citation frequency implies. The only quantitative claim in the article is that "only about 15% of companies practice effective decision making," from a survey of 350+ global organizations. There is no data in the piece linking decision-rights clarity to financial or delivery performance. Treat RAPID as useful vocabulary, not as an evidenced intervention. bain.com/insights/who-has-the-d
• Amazon's single-threaded owner model is the strongest structural counter to any dual-PM split, and it is a rejection of the premise. The principle is that for each initiative there is one leader whose focus is that initiative alone, empowered to decide without navigating approvals. Secondary sources only — I did not reach an Amazon-authored primary text in this pass, so treat the specific phrasing as unverified. pedrodelgallego.github.io/blog/amazon/single-threaded-model, mariogerard.com/single-threaded-leader
• "Two in a box" also has a legitimate positive lineage, but it is the opposite structure. Search results attribute the model to Intel under Andy Grove, applied to product as "pair PMing," where co-PMs make decisions together. The distinction that matters: pair PMing is symmetric (both people hold all three constituencies jointly); the PM/PO or sponsor/substance split is asymmetric (each person holds a subset). Unverified — I did not fetch the Trexin or Medium sources and did not find an Intel-primary account. trexin.com/two-in-a-box, medium.com/@arielyoffie

Convergences and contradictions

• Convergence — our engine/front-end cut was right and the literature backs it. The 2026-07-05 proposal (Ben owns the engine, Todd owns the user-facing surface) splits by product surface, which preserves each PM's access to all three constituencies for their surface. That is the one split shape Cagan's principle survives. Our instinct was correct; we just didn't know it was the load-bearing choice.
• Contradiction — our own stated mitigation is the drift mechanism, not the defense. [[2026-07-06-caf-die-roadmap-v0-pack]] proposed "circulate L1 jointly with Todd" and making him "co-owner of a lane." Under RAPID's consolidation logic and Cagan's access logic, joint circulation of the roadmap artifact is precisely how the person with the shorter sponsor line acquires roadmap standing: the sponsor sees two names on the roadmap and routes roadmap questions to the one he already talks to. Co-authorship of the artifact is the drift. Making Todd co-owner of a lane (a surface) is fine; making him co-signer of L1 (the roadmap) is not. This is a direct walk-back of our prior recommendation.
• Contradiction — no framework answers the question as posed. I found no named framework whose purpose is "keep the split while preventing relationship-holder drift." The literature's answer is uniformly: the drift is not a failure mode of the split, it is the split's natural equilibrium, so change the split. Any consultant-style artifact that claims to solve this (a RACI matrix pinned to a wall) is folk advice dressed as structure — it declares who decides without giving anyone an interest in honoring the declaration.

Synthesis for RDCO

The uncomfortable finding first: in Cagan's model, Ben is the product owner, not the product manager. The question is framed as "how do I stop Todd from drifting into roadmap ownership," which presumes Ben currently holds it. But the structure described — one person facing the sponsor, another facing the engineers and the substance — maps onto the PM/PO anti-pattern with Ben in the PO seat. Organizational gravity in that shape runs one direction: roadmap authority migrates to whoever the decision-maker actually talks to. Todd is three days in and greener, so today the substance asymmetry masks the structural one. That is a temporary condition with a predictable expiry. The drift the pre-mortem worried about is not a risk of the split landing wrong; it is what the split does when it lands as described.

No structural device with evidence behind it prevents this while preserving the split. Ranking the candidates honestly: (1) Surface-based split — each PM owns a whole surface end-to-end including its own sponsor line — is the only device that changes the equilibrium rather than fighting it, and it is what both Cagan and the single-threaded-owner model converge on. (2) Explicit decision rights (RAPID's "D") is real and worth doing, but it is declarative and its primary-source evidence base is a single survey statistic; more importantly, decision rights are granted, not claimed. Ben cannot assign himself the D over the CAF roadmap. Vincent can. Any RAPID exercise that isn't ratified by the person both PMs answer to is theater. (3) Single-writer artifact ownership — one author of record on the roadmap, circulated under one name — is a real device but it is my inference from RAPID's consolidation logic, not a cited finding; treat it as a reasonable operational rule, not evidence. (4) Meeting design, joint circulation, regular syncs, "make him a co-owner early" — this is the folk-advice tier. No source in this pass offers evidence for it, and per the contradiction above at least one variant is actively counterproductive.

The seam thesis is a better defense than any org chart Ben can negotiate for. [[2026-07-06-caf-die-roadmap-v0-pack]] already commits to claiming the unowned interfaces: Fabric port-set spec, Build Manifest as the CAF→DCA contract, the Evals seam definition. This matters structurally, not just technically. A seam contract determines what both sides of the seam can build and in what order — so whoever authors the contracts controls sequencing regardless of who owns the "roadmap" document. And a contract is an artifact a front-end PM cannot credibly co-author; it requires the substance Ben has and Todd doesn't. This is roadmap authority acquired through capability rather than granted through org design, which is the only kind that doesn't evaporate when Todd's sponsor line gets used. Caveat: it is defensive-in-depth, not a substitute for asking Vincent to rule on the split, and it has an expiry too — the pre-mortem's "hub gets claimed de facto" risk (Murray's Neptune graph "a weekend project away") is the same clock running on the same wedge.

What this means for the ask Ben should actually make. Not "clarify the RACI." The ask is a surface cut plus a named D, from Vincent, in writing, this month: Ben owns the platform/engine surface (Fabric, contracts, CAF backend, seam roadmap) end-to-end including his own direct line to Vincent on it; Todd owns the client-facing app surface end-to-end including its sponsor line. Two single-threaded owners, not one owner and one interface. The failure mode to name explicitly when proposing it is the one the literature documents — a split by constituency rather than by surface produces exactly one outcome, and it costs the org the person who understands the system.

Why this is in the vault

This resolves the open "Todd/Vincent dynamics" pre-mortem item in [[2026-07-06-caf-die-roadmap-v0-pack]] and overturns its stated mitigation (joint L1 circulation), which changes what Ben should do with the roadmap circulation step that is still pending. It also converts the still-TBD PM split into a specific, citable ask to Vincent rather than a boundary Ben negotiates bilaterally with Todd.

Open follow-ups

• Does the consulting/agency "engagement lead vs delivery lead" split actually work, and if so what makes it different? Ben's situation is a professional-services org, not a product org, and services firms run relationship-holder / substance-holder splits durably at scale. My hypothesis is that it works because the relationship-holder owns commercial scope and never owns the technical roadmap — but I found no primary source on this in this pass and did not test it. Genuine RESEARCH question. This is the sharpest gap: the whole product-management literature above may be answering for the wrong org type.
• What did Vincent actually intend by creating a "front-end PM" role three days before Ben asked about the split? Whether Todd was hired to own the app surface or to be Vincent's roadmap proxy determines whether the surface-cut ask lands as helpful clarification or as a territorial move. BUILD/TEST task — answerable only by asking Vincent or Taylon, not from public sources.
• (Lower priority, not proposed for the backlog) Whether Amazon has published a primary-source account of single-threaded ownership; all sources reached here were secondary.

Related

• [[2026-07-06-caf-die-roadmap-v0-pack]]
• [[2026-07-05-caf-technical-pm-role-framing-and-roadmap-kit]]
• [[2026-07-06-die-fabric-hub-spoke-map-and-roadmap-implications]]
• [[2026-05-27-analytics-career-ladder]]

Sources

Vault

• 01-projects/phdata/2026-07-06-caf-die-roadmap-v0-pack.md — [[2026-07-06-caf-die-roadmap-v0-pack]]
• 01-projects/phdata/2026-07-05-caf-technical-pm-role-framing-and-roadmap-kit.md — [[2026-07-05-caf-technical-pm-role-framing-and-roadmap-kit]]
• 01-projects/phdata/2026-07-06-die-fabric-hub-spoke-map-and-roadmap-implications.md — [[2026-07-06-die-fabric-hub-spoke-map-and-roadmap-implications]]
• 01-projects/phdata/roadmap-v1-work/roadmap-v1-founder-notes.md (founder-only layer; same Todd/Vincent risk text)

Web (fetched and read)

• Marty Cagan, "Two in a Box PM," Silicon Valley Product Group — https://www.svpg.com/two-in-a-box-pm/
• Marty Cagan, "Product Manager vs. Product Owner Revisited," Silicon Valley Product Group — https://www.svpg.com/product-manager-vs-product-owner-revisited/
• Paul Rogers & Marcia Blenko, "Who has the 'D'?", Bain & Company, June 28 2006 — https://www.bain.com/insights/who-has-the-d/

Web (search-result level only — claims above marked unverified)

• https://pedrodelgallego.github.io/blog/amazon/single-threaded-model/
• https://www.mariogerard.com/single-threaded-leader/
• https://www.trexin.com/two-in-a-box/
• https://medium.com/@arielyoffie/2-styles-of-product-management-pair-pming-vs-swim-lanes-5ee7e6f4623b
• Matt Wheal, "How to Split the Product Manager Role — and why PM/PO is an anti-pattern," LinkedIn (not fetched; LinkedIn article, treated as inaccessible)

Not reached / flagged

• No paywalled source was hit in this pass. The original HBR print version of "Who has the 'D'?" (HBR, January 2006) was not fetched; the Bain-hosted version was used instead.
• No Amazon-authored or Intel-authored primary text on single-threaded ownership or two-in-a-box was located within the source cap.