Why Salesforce cannot be your marketing source of truth

Every VP of GTM has watched a proposal go out of Salesforce with a price the deal desk retired last quarter. The rep pulled it from a quote template that nobody re-synced. The CRM was the system everyone trusted, so nobody thought to check the number inside it.

The Commercial Truth manifesto argues that marketing has never had infrastructure. Engineers commit code to Git; finance posts every entry to a ledger. The CRM gave the deal a system of record decades ago — but the claims that close the deal were left as prose scattered across docs, decks, and Slack threads.

So the instinct to make Salesforce the source of truth for marketing is natural. It is already the spine of revenue operations, so why not let it hold what the company says about itself too? The answer is structural, and it is worth naming plainly, because this is the same movie GTM leaders have watched five times (Source: Assay “what we’re not” positioning canon).

The promise Salesforce made

In 1999, customer relationship management was not a software category; it was a set of behaviors. Salespeople kept contacts in Rolodexes, deals in spreadsheets, and meeting notes in their heads (Source: “What CRM Did for Pipeline, Commercial Truth Does for Knowledge”, Assay canon).

Salesforce did not just build a database. It made the argument that the deal was a structural asset of the company that deserved governed infrastructure, and it moved customer data “from Heads to Infrastructure” (Source: same canon). That argument was correct, and it built a category on it.

The promise the buyer hears today is the extension of that win. If the CRM became the single place revenue lives, surely it can become the single place the truth lives — pricing, positioning, the competitive line, the proof point. One system, one login, one record everyone reads from.

What it actually delivered

What Salesforce delivered was a system of record for the deal, not for the claim. The Assay canon states it directly: the CRM is “optimized for the customer record, not the commercial claim” (Source: Assay “what we’re not” positioning canon).

Look at where the commercial truth actually lives in a Salesforce-centric org. Pricing sits in a separate spreadsheet or a price book that the quote tool reads at a single point in time. Positioning lives in marketing docs, and competitive intel lives in someone’s Slack DMs (Source: Assay “what we’re not” positioning canon).

The CRM tracks the deal stage with precision. It does not track the truth status of the claims being made inside that deal (Source: “What CRM Did for Pipeline, Commercial Truth Does for Knowledge”, Assay canon). Salesforce can tell you exactly where an opportunity sits in the funnel and stay silent on whether the price on the attached proposal is the one the deal desk approved this quarter.

The same canon draws the division of labor cleanly. The CRM owns the “who” and the “when”; it does not own the “what” — the factual content of what the company says about its pricing, its security posture, its competitive position right now (Source: same canon). When everyone owns the what, nobody does, and that is the knowledge layer gap the modern revenue stack inherits.

The structural reason it cannot be the substrate

The failure is not configuration. You cannot fix it with another custom object or a tighter validation rule. There are four structural reasons, and each is a property of the data model, not the discipline of the admin team.

1. It is modeled on the customer object, not the claim. Accounts, contacts, opportunities — the schema is built around the relationship and the deal. A price, a competitor capability, or a positioning statement has no native home as a first-class, governed object. It rides as a field on a quote line or a note on a record.

2. There is no propagation. Edit a price in the price book and nothing else moves. The decks, the proposal templates, the email sequences, and the AI SDR that all reference that number stay stale. The cascade that should fan a change out to every downstream surface does not exist, so version skew is the default.

3. There is no source type or confidence signal on the claim. A price keyed in by an admin two years ago renders identically to one verified by the deal desk yesterday. The record carries a “last modified by” — useful for tracking blame, useless for tracking truth (Source: Assay “what we’re not” positioning canon). There is no enforced confidence ceiling separating canon from guess.

4. It governs the relationship, not the assertion. The CRM is excellent at the question where does this deal stand? It is structurally silent on the question is the claim inside the proposal still true today? — the same blind spot every predecessor system shares (Source: Assay “what we’re not” positioning canon).

Underneath all four is one root cause. The CRM “was designed for commercial claims as a typed, versioned, sourced, testable artifact” — that is precisely what it was not designed for (Source: Assay “what we’re not” positioning canon). Salesforce is a system of record for the deal asked to be a system of record for the claim, a function it was never schema’d to hold.

The claim-level fix

The fix is not a better CRM. It is a shift in the unit of storage — from the customer record to the commercial claim. As the typed-graph argument lays out, each fact becomes a typed node: a price, a competitor capability, a positioning statement, each a first-class object rather than a field.

Every node carries five things the record never could — a source type, a confidence score with an enforced ceiling, a version history, a cascade map of every downstream surface, and an audit-log entry on each change (Source: Assay “what we’re not” positioning canon). Confidence is stated as a calibrated interval, and unverified data is barred by ceiling from overriding canon.

Now a change behaves like infrastructure. Edit one pricing node and the cascade traces its dependencies, flags the proposal templates, decks, and agents that reference it, and routes them through a review pass. The number of downstream surfaces a single edit touches is not yet measured publicly; the point is the mechanism, not a headline figure.

Salesforce does not get deleted in this picture; the CRM still owns the customer record and the deal stage, and the claims the proposal references become governed objects in a layer above it (Source: Assay “what we’re not” positioning canon). As the what CRM did for pipeline argument frames it, the CRM owns the “who” and the “when,” and the Truth Graph owns the “what.” The infrastructure layer sits one level above the system of record — the substrate it reads from when it needs to know what the company is currently saying.

Closes / opens

Closes the LSO §F.10 predecessor-comparison cluster for the Salesforce query — the promise the CRM made, the deal-record it actually delivers, the structural reason a customer-object schema cannot be the claims canon, and the claim-level substitute. The full negative case, including every other system pressed into this role, lives in the what we’re not canon.

Opens the integration question: when Salesforce stays the system of record for the deal, what is the cleanest way to feed it canonical, current claims so the proposals and quotes leaving the CRM are grounded in verified positioning rather than a stale price book?

The methodology Assay is developing for the Commercial Truth Index measures exactly this capacity — whether an organization can ground, score, version, and propagate its commercial claims across every surface, human and AI, rather than bolt them onto a system built to govern the deal.

This essay is grounded in the Assay “what we’re not” positioning canon and the “What CRM Did for Pipeline, Commercial Truth Does for Knowledge” essay. Methodology for the Commercial Truth Index is in development.