Recommendation Marketing requires five conditions before an agency can honestly share performance risk: a written outcome, measurement established before the work, bounded claims, defined obligations and exposure on both sides, and a scorecard neither party can rewrite after the result is known.
The discipline organizes the work around whether a business becomes easier for buyers, and the AI systems assisting them, to understand, verify, and recommend. Recommendation Measurement governs the evidence for selected parts of that standing. Shared accountability and performance-linked compensation come afterward.
Last issue diagnosed four places agency accountability can break down:
- output becomes the target,
- reporting favors what the agency controls,
- attribution remains ambiguous,
- and channel ownership fragments a buyer decision experienced as one whole evaluation.
The next step is not a guarantee. It is a set of conditions that must exist before shared performance risk becomes credible.
TL;DR
- Many marketing arrangements require the buyer to fund activity before the intended outcome can be observed. That does not make the arrangement dishonest, but it leaves the buyer carrying much of the outcome risk.
- Recommendation Marketing defines the shared outcome. Recommendation Measurement establishes the governed record. Only then can responsibility and compensation be connected to measured movement.
- Five conditions make shared performance risk credible: a written outcome, measurement agreed before the work, bounded claims, real obligations and risk on both sides, and a governed scorecard neither side can rewrite afterward.
- Missing any one condition can produce a guarantee nobody can keep, a retainer wearing a performance label, or a dispute waiting for its trigger.
- Right Angle is building toward this order, measurement first, accountability second, pricing third, but is not announcing a public performance offering in this issue.
The Payment Inversion
Start with the sequence of payment and value. In some exchanges, they occur together. Buy a hot dog and you receive the product as you pay. In many business transactions, delivery is inspected and the invoice is paid later. Marketing often works differently: the buyer funds people, media, systems, and production before the intended business effect can be observed.

That sequence is not proof of failure or bad faith. Agencies incur real costs before results appear, and many outcomes depend on variables outside an agency’s authority. Historically, contracts therefore attached payment to things both sides could define and inspect: time, scope, campaigns, content, media, traffic, leads, and other forms of activity or output.
The World Federation of Advertisers reported in September 2026 that labor-based payment was the standard model for 19% of the multinational companies in its study, down from 54% in 2011. Fixed-fee or output-based models rose from 20% to 33%. The industry is changing how work is purchased, but changing the unit of payment does not automatically change the outcome for which anyone is accountable.
The harder question is not whether marketers deserve to be paid. They do. It is whether the commercial arrangement can distribute risk more honestly once a limited, governed outcome can be observed. Better measurement does not remove uncertainty. It does make “we could not measure anything meaningful” a less complete answer.
Why I Am Inside This Proposal
I have spent more than thirty years asking clients to pay agencies before anyone could know exactly what the work would produce. I have written the scopes, defended the retainers, carried the payroll, managed the teams, and tried to connect the reports to business results when the causal line was anything but clean.
I do not think agencies created this problem. I think we inherited a model built around what could be counted and contracted. Recommendation Marketing gives us an opportunity, and an obligation, to organize accountability around something more consequential: whether the business becomes easier to understand, verify, and recommend.
Why Performance Arrangements Fail
Performance pricing is not new. It can still fail in three familiar ways:
The guarantee nobody can keep. An agency promises an outcome it does not control, and the arrangement fails when reality exceeds the promise.
The retainer in costume. The fee is renamed, but the triggers remain loose enough that compensation is effectively unconditional.
The dispute waiting for its trigger. Both sides agree to “results” without first agreeing on the definition, baseline, conditions, evidence, responsibilities, or decision rule.
All three begin with the same sequencing error: compensation is designed before measurement. Shared accountability is not primarily a pricing innovation. It is a measurement discipline with pricing consequences.
What Recommendation Marketing Requires Before Risk Can Be Shared

The outcome is written down and accepted by both sides
Recommendation Marketing begins by defining the shared outcome. Not “more visibility” or “better pipeline,” but whether the business is becoming easier to understand, verify, and recommend for the buyer questions that matter. The agreement must identify which questions, which observed systems, what counts as appearing, what counts as accurate representation, what evidence will be recorded, and what business signals remain outside the scorecard.
Alok Kiran Bhalke made the problem plain in the discussion after the last issue: the outcome can live inside an assumption nobody wrote down. Naming the outcome is the first act of sharing accountability because it forces both parties to discover whether they mean the same thing.
Measurement exists before the work begins
Recommendation Measurement must exist before accountability can attach. A Frozen Question Set is selected before anyone knows whom the answers favor. A baseline is recorded before improvement work begins. Conditions are declared, observations are repeated, and normal variability is characterized. Only then can a later comparison show bounded movement rather than invite a negotiation over the past.
Recommendation Measurement does not observe the buyer’s entire decision, prove why a sale occurred, or eliminate model drift. It creates a governed record of selected answer-level standing. That is narrower than full attribution and more useful than pretending no meaningful comparison is possible.
The claims stay bounded in the agreement
Recommendation Marketing does not claim control over the recommendation. No agency controls what an AI system will say, what a buyer will believe, or whether a sales team will convert interest. The accountable claim must therefore be narrow: movement on agreed questions, within declared conditions, over a defined period, with the limits written beside the measure.
A credible agreement states what the measurement can show, what it cannot show, how model or market shifts will be investigated, and which decisions require human judgment. An arrangement that requires certainty will break. One designed around disclosed uncertainty has a chance to hold.
The economics create real exposure and obligations on both sides
Recommendation Marketing is a shared discipline, not an agency promise performed in isolation. If the agency’s compensation cannot change when the agreed measure does not move, performance risk has not been shared. But the client’s obligations must be equally explicit: access, approvals, evidence, operational follow-through, sales reporting, and the authority to make changes the work requires.
Starting exposure also matters. Foundational work has real cost, so the honest standard is not “free until it works.” The standard is proportion: the initial commitment should be visible and justified, the measurement should begin early, and the performance-linked portion should remain genuinely contingent. Otherwise, the old allocation of risk survives beneath new language.
Neither side can rewrite the scorecard after the work begins
Recommendation Measurement protects the discipline from convenient interpretation. The client opens its attribution picture honestly, including evidence that credits nobody. The agency accepts a governed record it cannot quietly edit: questions frozen, baseline preserved, conditions declared, and movement claimed only within the agreed decision rule.

The measurement instrument may be operated by one party or a technical provider. What matters is governance: the question set, baseline, records, and interpretation rules are visible to both sides and cannot be revised merely because the result is uncomfortable. A shared-risk arrangement that could never cost the agency anything was never sharing risk.
The Buyer’s Checklist
Before entering any shared-risk marketing arrangement, require:
- The intended outcome in writing.
- The question set, baseline, conditions, and decision rule before improvement work begins.
- The limits of the measurement stated inside the agreement.
- Agency compensation that can genuinely change with the agreed measure.
- Client obligations and dependencies stated with equal specificity.
- Starting exposure proportionate to the foundational work and the uncertainty involved.
- Access to the underlying measurement record, not merely a favorable summary.
Any provider unwilling to meet those requirements, including us, has told you something useful.
What This Proposal Does Not Mean
- It does not mean guarantees. Neither an agency nor a measurement system controls the buyer, the market, or an AI model.
- It does not mean marketing becomes free until it works. Foundational work has real cost, and starving the work is not shared accountability.
- It does not mean every engagement should be performance-priced. Diagnostic, foundational, or not-yet-measurable work may require a different commercial structure.
- It does not mean attribution ambiguity disappears. The ambiguity is bounded, disclosed, and governed rather than left unspoken.
- It does not mean answer-level movement equals revenue impact. Recommendation standing is one observation layer; sales, financial, brand, customer, and qualitative evidence remain necessary.
Where Right Angle Stands
Right Angle is building toward this model deliberately. The measurement order comes first: define the Frozen Question Set, record the baseline, preserve the conditions, publish what the instrument can and cannot show, and accept the result whether it is favorable or not.
The commercial model must come later. I am not announcing a productized performance offering today because doing so before the measurement has proved itself on us would reverse the discipline this series has argued for: measurement first, accountability second, pricing third.
What I can offer now is the conversation that precedes any arrangement: is your business measurable in the way condition two requires? That question can be examined without a guarantee and without pretending the answer is already known.
That is where Right Angle stands today. Recommendation Marketing defines the destination. Recommendation Measurement establishes the record. Shared accountability becomes possible only when both are strong enough to withstand an unfavorable result.
Thinking about your AEO Strategy? Sightline by Right Angle
The Invitation
If you run an agency, which of the five conditions would you refuse, and why? If you buy marketing, which condition would your CFO add?
Comments and direct messages both work. The disagreement is part of the test.
Coming next: your questions. The October 27 issue will be built from what readers have been asking about Recommendation Marketing in comments, messages, and the conversations this series continues to start.
Frequently Asked Questions
What does shared accountability mean in Recommendation Marketing?
It means defined agency compensation can be connected to governed movement on an agreed Recommendation Marketing outcome, such as selected answer-level standing on a Frozen Question Set. The measurement, limits, responsibilities, and decision rule must exist before the work begins. It does not mean the agency guarantees revenue or controls buyer choice. This information is not legal advice and does not create an attorney-client relationship.
Is shared performance risk the same as pay per lead?
No. Pay per lead prices a particular unit, often one the provider can influence directly. Shared performance risk begins by defining the broader outcome, the governed measurement, the obligations of both parties, and the limits of attribution before compensation triggers are designed. This information is not legal advice and does not create an attorney-client relationship.
Can an agency guarantee results under this model?
No. A guarantee would ignore variables the agency does not control. The honest commitment is bounded: defined work, governed measurement, disclosed uncertainty, and compensation that responds to the agreed measure under the stated conditions. This information is not legal advice and does not create an attorney-client relationship.
Why does the cost of starting matter?
Because a large, unexplained upfront exposure can leave the buyer carrying nearly all the risk before the first measured cycle. The initial commitment should reflect legitimate foundational work while preserving a genuinely contingent portion of compensation. This information is not legal advice and does not create an attorney-client relationship.
What should a business do before discussing shared performance risk?
Become measurable. Identify the buyer questions that matter, freeze the question set, record a baseline, state the conditions, and understand which business signals can and cannot be attributed. If those foundations do not exist, the commercial trigger is premature. This information is not legal advice and does not create an attorney-client relationship.
B Randall Willis
Founder, Right Angle
Visibility + Engagement = Opportunity