The attribution peace treaty: ending the sourcing war between marketing and sales
On this page
The sourcing war between marketing and sales ends with a treaty, and the treaty has five terms: one shared number with visible provenance, explainability privileged over accuracy, scheduled model changes, explicit influence and sourcing accounting, and a permissions annex naming who may change the logic. Sign nothing less. B2B marketing attribution did not cause the war by being wrong. It caused the war by being unable to prove anything either way.
The war is structural, and the buying data explains why. Gartner's research puts 6 to 10 decision-makers inside a complex B2B purchase, spending about 17 percent of their total purchase time with suppliers of any kind, with roughly 80 percent of B2B sales interactions now running through digital channels. Ten hands touch a deal, mostly out of view. Under those conditions no positional rule can settle who made the deal happen, so credit is contested by design. The IAB's State of Data 2026 research adds the mood: 75 percent of US buy-side leaders say measurement is broken. Broken measurement does not stay neutral inside a company. It picks fights.
Why credit is contested by design
Sources: Gartner B2B buying research · IAB State of Data 2026 via eMarketer, Feb 2026
What follows is the operating manual: the war documented in the field's own words, why fighting is the rational strategy under untrusted numbers, the five treaty terms drawn from teams that ended it, and the de-escalation sequence for organizations already at war. It is the companion piece to the attribution trust gap, which documents the measurement failure underneath the fight.
The war, in the field's own words
The most direct damage assessment in our field record connects a revenue miss to the measurement system itself.
We've had a really bad year in sales and I personally think that it's a direct result of this sourcing competition. People are spending more time trying to figure out who owns the opportunity than how we all work together.
a marketing operations leader at a cloud storage company
The rest of the record fills in the picture. A marketing leader at a healthcare benefits company: "the way we leverage attribution has been more to divide, rather to unify." A marketing leader at a cybersecurity company, on the intramural version of the same fight: "the events team is going to tell me how important their events are, and the product marketing team is going to tell me how important their steps are. Nobody's going to agree." And the executive form of the conflict, in one sentence:
As soon as you present a slide that shows a different number than what sales sees, it becomes a war of dashboards.
a CMO, from the founder's published field notes
The front line does not stop at the marketing and sales border. It runs inside marketing too. A marketing leader at a contract lifecycle management company found her own reporting picking winners among her own teams: "only the contact-us form is getting credit because it's MQL first touch." When the measurement cannot allocate credit between events and product marketing, or between a form and the channel that filled it, every team learns to fight for the touch it controls. The war is fractal. Wherever an untrusted number meets a budget line, a front opens.
The war also has casualties, and they cluster on one side of the org chart. Spencer Stuart's 2025 study puts Fortune 500 CMO tenure at 4.3 years against a 4.9-year C-suite average, with only 66 percent of the Fortune 500 keeping a dedicated top-marketing role at all, down 8 points in a year. And replacing the general does not end the war. One CEO in the founder's published record replaced his CMO twice in 18 months and got the same result both times, because the infrastructure both CMOs inherited was broken. The fight survives every personnel change, which is the strongest evidence that personnel were never its cause.
The fight is rational
The standard diagnosis of the sourcing war is cultural: silos and egos. The field record supports a harder reading. Under untrusted measurement, fighting is the correct strategy.
Consider the position of a sales leader handed a marketing-sourced pipeline number he cannot audit. The number will shape budget shares and the narrative of the year. If it cannot be verified, accepting it is negotiating against himself, and contesting it costs one meeting. He contests. Marketing, holding a model it cannot fully explain either, escalates with a second dashboard. Both sides are behaving reasonably, and that is what makes the war stable.
The numbers keep supplying ammunition. A marketing leader at a sales compensation software company described the model sensitivity that turns every review into a hearing: "depending on the model that you chose, that same event, you would be sharing a different ROI value." At one enterprise we documented in the trust-gap report, the same email channel measured $16M of influenced pipeline on a linear model, $1.8M on an AI model, and $18K the next quarter. A number that volatile cannot arbitrate anything. It can only arm both sides. Deepinder Singh Dhingra, RevSure's founder, states the endpoint: "No company dies from lack of data. They die when data becomes political instead of analytical."
Notice that the distrust is bipartisan. A growth leader at a compliance software company threw out a number that flattered his own side, an email channel showing $5M of pipeline at zero cost: "I would love it if our email marketing channel actually produced 5 million in pipeline cause it costs us $0. So I just don't think that's accurate." Leaders on both sides reject even favorable numbers when the machinery is untrusted, which means the war is epistemic rather than tribal. Epistemic wars do not end with team-building. They end with evidence.
The practical consequence is that treaties beat pep talks. An alignment offsite adjusts attitudes toward a number nobody trusts, which lasts about one QBR. The durable fix changes what the number is and how it may be changed, which is what the rest of this playbook specifies.
The treaty terms
Five terms, drawn from what actually held inside enterprise revenue teams. Each carries the failure mode it replaces.
- One shared number with visible provenance.
Less effective: marketing reports from the automation platform, sales reports from the CRM, finance keeps a spreadsheet, and every review opens with a reconciliation fight that pre-empts strategy.
Recommended: a single pipeline and attribution number that both sides consume, where any figure decomposes on demand into the deals and touches behind it and the rules that credited it. The field description of the after-state is blunt about what actually ends arguments:
RevSure being multi-touch and a trusted brand, people just accept the value... it's harder for people to poke holes in and I for one enjoy that.
a demand gen leader at a customer success software company
Harder to poke holes in is the design goal, stated exactly. The number does not need to be beyond dispute in theory. It needs to survive dispute in practice, with its evidence attached.
- Explainability privileged over accuracy.
Less effective: the most sophisticated multi-touch attribution model available, re-tuned as the vendor ships, defended in meetings with "the algorithm weights it."
Recommended: the model both lieutenants can walk their CEO through unaided. Where explainability and marginal accuracy conflict, the field's calibrated operators choose explainability, with reasons:
I would rather have a worse model that I can actually explain to people if they ask. That was the whole problem with the W thing: the data changes all the time. It's impossible to explain.
a marketing operations leader at a software testing company
An attribution number's function inside a treaty is adjudication. A ruling nobody can follow settles nothing, however correct it is.
- Scheduled model changes with change control.
Less effective: re-weighting the model mid-quarter because a new feature shipped, silently moving every team's sourced number while their targets stand still.
Recommended: the model freezes within the quarter and changes ride a published calendar with notice. Targets move with the model, or the model waits. The reason is fairness, and fairness is load-bearing for peace. A marketing leader at a cybersecurity company, on why her team freezes the model quarterly: "If we change that all of a sudden, and it's out of their control, holding them responsible to that number is also hard." People will accept a number that moves against them. They will not accept one that moves by surprise while their targets stand still.
- Explicit influence and sourcing accounting.
Less effective: one blended "sourced pipeline" number that every team has learned to farm, with renewals, SDR meetings, partner deals, and event follow-ups claimed by whoever reports first.
Recommended: two published views with the edge cases written down in advance. Sourcing answers who originated the opportunity. Influence answers what moved it. The patterns that kept the peace at multiple companies in our record: renewals stay in the influence view and out of the marketing ROI view, so expansion revenue stops being a land grab, and SDR-sourced meetings are deliberately separated from marketing attribution, so the SDR team's number is its own. The specific choices matter less than their explicitness. Every unwritten edge case is a future border dispute.
- The permissions annex.
Less effective: attribution logic lives wherever admin rights happen to sit, so the number changes whenever someone with a login has a strong opinion.
Recommended: a short annex naming who may change attribution logic, who may re-categorize an opportunity's source, who approves the definition of a marketing-influenced renewal, and where each change gets logged. This term is about to matter more, because agents are joining the terrain. Deepinder's formulation: "Your context graph becomes what your agents see. Your permission graph becomes what your agents can do. If both are ambiguous, your agents will fight the same fight your humans are fighting now, without anyone in the room to slow them down." The treaty you write for humans is the permission structure your agents inherit.
The war and the treaty
The operating state untrusted attribution produces, against the terms that ended it in the field
The war
The treaty
Source: RevSure working sessions across enterprise GTM teams, quoted at descriptor level
The de-escalation sequence
For organizations already at war, order matters. Five moves, run in sequence.
-
Declare the ceasefire. Freeze the current model, whatever it is, and grant amnesty on history: no retroactive re-crediting of past quarters. Relitigating old credit reopens every wound and changes no budget that still exists.
-
Reconcile the substrate before debating models. Most sourcing fights are secretly data fights: the same account exists in four systems with different owners and stages, so each side's number is computed from a private version of reality. Until identities, stages, owners, and funnel definitions reconcile, model debates are fought over unstable ground. This is the job of the Full Funnel Data Graph: one substrate both sides' numbers are computed from. The substrate work also surfaces why the old fights were unwinnable. At the same compliance software company, 95.5 percent of deals carried a lead source of "cold call" because it was the CRM default, applied for years. Their growth leader's verdict on the cold-call motion: "It's just not a thing we use." Two functions had been arguing sourcing splits over a field that described neither of them.
-
Concede the model to win the process. This is the move senior marketers resist, and the field's pragmatists make it deliberately. When trust is at zero, a CMO who insists on the sophisticated model is asking the organization to trust her twice, once on the number and once on the machinery behind it. Concede the machinery first:
the problem is reporting to the higher ups, you need something that is very solid... they can stick with last touch, whoever it is
a marketing leader at a cybersecurity company
Last touch undercounts marketing, and she knows it. She trades reported credit for process legitimacy: a solid, simple number that survives the executive meeting, inside a change-control process she helped write. The model can be upgraded later, through the calendar, with notice. The process, once lost, is the thing that cannot be cheaply rebuilt. Say the cost out loud: this playbook asks marketing to give up real reported credit in the near term, and that concession is what makes the treaty a treaty rather than a victory demand.
- Publish the treaty on one page. Signed by the CMO and CRO as a pair, structured so nothing important stays verbal.
| Treaty section | What it fixes in writing | Who owns it |
|---|---|---|
| The two views | Sourcing and influence definitions, with the renewal and SDR rules | CMO and CRO jointly |
| The model | Which model is primary, and why it can be explained unaided | RevOps |
| The change calendar | When rebalancing happens, notice period, how targets adjust | RevOps, approved by both |
| The permissions annex | Who may change logic, re-categorize sources, redefine influence | Named roles, audited |
Publication is the point. A treaty that lives in one leader's head binds nobody, and a treaty on one page gets enforced by everyone who can read it.
- Upgrade the model inside the process you won. With change control live and provenance visible, moving from last touch to multi-touch, and from touchpoints to Decision Attribution, stops being a political event and becomes a scheduled improvement. That is where the sophistication belongs: last, on settled ground.
The measurement that survives cross-examination
What changes after the treaty is the subject of the meetings. The cloud storage leader's diagnosis ran "people are spending more time trying to figure out who owns the opportunity than how we all work together." Reverse that sentence and you have the after-state: the sourcing question gets answered by the system, with provenance, and the meeting moves to where the next dollar goes.
The war also has a payroll cost that ends with it. Across three companies in our record, marketing attribution reporting consumed roughly 100, 50, and 40 hours per quarter before a shared number existed, hours spent largely on manufacturing ammunition for the next reconciliation meeting. Reclaimed, those hours became analysis, which is the work both sides claimed to want the whole time.
RevSure's role in the after-state is the layer underneath the terms. Decision Attribution credits the decisions that moved a deal rather than whichever touch happened to sit first or last, computed on the Full Funnel Data Graph so both sides' numbers share one substrate. Every credited decision carries a Decision Trace, the auditable evidence chain that makes poke-holes sessions survivable. And model and permission changes run through the GTM Harness loop, Propose, Approve, Commit, Roll back, which is the change-control term implemented as software. A CFO once told Deepinder that "marketing is the only department that grades its own homework." The treaty retires that sentence, because the homework gets graded by evidence both departments can audit.
None of this softens the underlying measurement problem. With 6 to 10 people deciding a purchase across mostly digital channels, attribution stays genuinely hard, and the treaty does not promise a perfect number. It promises a number whose imperfections are documented, and whose rules stay stable enough to govern by.
The consequence to close on: the sourcing war is rational behavior under untrusted measurement, so it ends when the measurement stops being contestable, and at no other time. Reorgs move the front line. New CMOs restart the fight from inherited positions. The war ends when the number survives cross-examination, because at that point contesting it stops paying. Peace, in measurement as anywhere else, is an equilibrium, and equilibria are built from incentives rather than intentions.
What to do next quarter
- Count the versions. How many definitions of sourced pipeline exist across your systems, and how many recurring meetings exist to reconcile them. That count is your war-intensity index.
- Draft the one-page treaty before touching any model settings: two views, change calendar, permissions annex, signatures.
- Freeze the model for the quarter and announce the freeze. Stability first, sophistication later.
- Write the edge-case rules explicitly: renewals in influence and out of ROI, SDR meetings separated, partner-sourced deals assigned, CRM defaults cleaned. Publish them where both teams report from.
- Pick one segment and run the shared-number pilot with full provenance, then grade it on a single criterion: did it survive cross-examination in the QBR without a second dashboard appearing.
- If trust is at zero, concede the model. Take last touch and the process, and schedule the upgrade.
Where this comes from
Built from our work inside enterprise GTM teams: indexed, verbatim working sessions with marketing, RevOps, and revenue leaders, quoted here at descriptor level with permission discipline. The treaty terms generalize from teams that had already invested in measurement and wanted it to hold, which is a selection effect worth knowing about. One limit: where compensation plans pay directly on contested sourcing, measurement reform alone will not end the war. Comp design is outside this playbook's scope.
Frequently asked questions
Why do marketing and sales fight over attribution?
Because the number decides budgets and careers, and most attribution cannot survive an audit. Gartner puts 6 to 10 decision-makers in a complex B2B purchase, so credit is contested by design, and when changing the model changes the answer, contesting the number is rational. The fight is a measurement problem wearing an org-chart costume.
What is the difference between sourced and influenced pipeline?
Sourced pipeline credits the team that originated an opportunity; influenced pipeline counts every interaction that moved it. Peace requires publishing both views with edge cases decided in advance. Field patterns include keeping renewals in the influence view but out of marketing ROI, and separating SDR-sourced meetings from marketing attribution entirely.
How often should you change a B2B marketing attribution model?
On a published calendar, never silently mid-quarter. Targets are set against the model, so unannounced re-weighting breaks fairness. A marketing leader at a cybersecurity company freezes the model each quarter because holding people responsible to a number that changed outside their control is hard. Change control is a treaty term, and it protects trust.
How do you get sales to trust multi-touch attribution?
Provenance beats persuasion. A demand gen leader at a customer success software company described a multi-touch number teams accept because it is 'harder for people to poke holes in.' Make every figure decompose on demand to the deals and touches behind it and the rules that credited it. RevSure records a Decision Trace for each credited decision, so the number survives cross-examination.