Pitch a Partnership When the Value Arrives Unevenly

A partnership slide usually has two columns. Left: what we do. Right: what they do. The columns sit at the same height, so the work looks symmetrical. It rarely is. "Commit two staff for six weeks" and "run the collection through the pipeline" can share a row and be nothing alike. One is a present, mostly irreversible cost. The other might be a few engineering days and an email.

That mismatch is survivable. What is not survivable is a diagram that also flattens time — when one side spends now and is paid later, in a currency it does not control, while the other side spends a little now and keeps something useful either way. At that point the slide is not simplifying the deal. It is hiding it.

The work here is to put two timelines on the table before anyone says "shared success," and then to compare the structures that change the exposure. That assumes the proposed work and handoffs are already described somewhere. If they are not, write that first; a timeline of obligations built on vague obligations is just a prettier guess.

Who commits before anything is learned

Take an illustrative case, invented and deliberately free of any real figures or valuations. Fernwood is a small community archive: roughly four thousand oral-history transcripts, nine hundred finding aids, and a decade of scanned municipal minutes. Kestrel runs a general retrieval service that is strong on web text and weak on small, idiosyncratic archives. On such material, the service tends to segment long transcripts mid-thought, merge personal names, and "correct" local place names into typos.

A trial is proposed. Fernwood prepares the collection; Kestrel runs it through its pipeline and reads the output.

Put the two timelines side by side, and keep the rows honest about time:

Two things stand out, and neither is about fairness. The first is reversibility. Six weeks of two people cannot be unspent. Sharing transcripts with a third party can be discontinued, but the learning drawn from them cannot be unheard. The second is who retains value if the trial stops. Fernwood's durable benefit is downstream and shared. Kestrel's is present and kept. That asymmetry — certain and retained on one side, contingent and deferred on the other — is the actual subject of the pitch, and it is invisible in a two-column diagram.

Note that "largely not recoverable" is doing real work. Some of Fernwood's preparation is reusable no matter what: a rights list keeps its value. So the cost is not all sunk. The point is not that Fernwood gets nothing. It is that Fernwood's best outcome depends on a decision Kestrel has not made.

Benefits arrive in different currencies

The next move is to stop summing. Learning, visibility, distribution, and revenue are different currencies. Kestrel's operational learning is not convertible into Fernwood's revenue, and a good shared outcome does not settle who did the present work. Averaging them into one "mutual value" figure produces a number that answers no question anyone asked.

Sort each possible benefit along three axes, and keep them separate:

• Kind. Operational learning, distribution, credibility, revenue. Fernwood's hoped-for benefit is discovery for its visitors; Kestrel's immediate benefit is product knowledge. Different currencies.
• Timing. Now, after a second stage, after adoption. Fernwood's discovery benefit lives in the last bucket. Kestrel's learning lives in the first.
• Certainty. Observed, proposed, or hoped. Nothing in this example has been observed — no trial has run and no partnership exists. Kestrel's expected retrieval failures are a proposal, not a finding. Fernwood's later discovery benefit is hoped for, conditioned on two decisions Fernwood does not make.

Here is the part worth sitting with. Of the three axes, negotiation can move only one cleanly: timing. You can stage a commitment, delay a cost, or attach a later benefit to a named condition. You cannot negotiate kind — learning stays learning. And you cannot manufacture certainty by wanting it; the only way to raise certainty is to have someone observe something, which means running part of the work.

So when a partner proposes to "balance" an uneven deal, ask which axis is actually moving. If the answer is "we'll describe the future benefit more warmly," nothing has moved. If the answer is "the second stage will not begin until we have seen a result," timing has moved, and that is a real change.

Four structures that change the exposure

The point of staging is not virtue. It is exposure management. Compare these four honestly, because each buys something and costs something.

A smaller first activity. Fernwood prepares two hundred transcripts and fifty finding aids — one person, one week — instead of the full collection. This answers a narrower question: do Kestrel's failure modes show up even at small scale? It can establish that much. It cannot establish retrieval quality across four thousand items, whether the cleaning burden is proportional at full scale, or whether Kestrel will build archival support. A smaller sample narrows the question; it does not answer the big one.

A staged commitment. Stage one is the sample. Stage two — the full six-week preparation — begins only if a named condition is met first. This can produce a genuine go/no-go. It cannot change direction: Fernwood still goes first, so staging shrinks the bet without changing who is exposed. And it collapses into theater if the condition is automatic — "we mutually agree to continue" is not a condition, it is a calendar. For staging to mean anything, the trigger has to be something that might fail.

A changed contribution. Kestrel writes and runs the extraction and normalization; Fernwood handles only rights and output verification, which only Fernwood can judge. This moves mechanical work to the party with the tooling and can establish whether the cleaning burden was ever Fernwood-specific. It requires Kestrel's real agreement and real capacity. If Kestrel declines, this option is simply not on the table, and the others must be judged without it.

No collaboration in the current form. If none of the above yields enough usable return or information for Fernwood to justify the present commitment, the answer in the current shape is to decline or redesign.

Published collaboration programs stage participation in roughly these terms. Microsoft's co-sell overview for partners describes joint opportunities and partner-to-partner activity as tracked, stageable routes rather than one all-or-nothing agreement. That is useful for mechanics — it shows that a staged structure is ordinary, not a sign of distrust. It is not evidence that any particular split of contribution and benefit is reasonable. The overview does not address timing asymmetry, and it could not, because whether a given arrangement is fair is a property of the specific parties, not of a program.

Make "no" a real alternative

State the condition under which the present form asks one side to commit without enough usable return or information to judge the next step. In the example: if Kestrel will not take on more of the preparation, will not accept a stage condition that could fail, and will not commit the deferred benefit to anything concrete, then Fernwood is spending staff weeks on a result it does not control.

That is a proposal problem, and it should be described as one. It is not an accusation about motives. Kestrel may fully intend to build archival support. Intention is not commitment, and the pitch must treat "Kestrel ships this" as a proposal to be made rather than a fact to be relied on. Nor does either organization owe the other a speculative future reward to make the diagram balance. You do not need to invent a downstream prize for Fernwood to justify a deal that Fernwood cannot evaluate.

Declining is not a failure of the relationship. It preserves the collection and the option to try again under different terms. It also has a cost for Kestrel, which learns nothing. That two-sided cost is worth naming, because it is exactly the reason Kestrel has an interest in a bounded or staged trial rather than an all-or-nothing one. Disguising the mismatch under a shared-success headline helps neither party; it just postpones the conversation to a point where it is more expensive.

Name who decides, and what would justify the next stage

Staging only works if you say what the next stage is waiting on, and who gets to say whether the wait is over.

A milestone should retire the specific uncertainty that made staging useful. Elapsed time is not a milestone. "Three months have passed" answers nothing. The stage-one uncertainty in the example is narrow: does Kestrel's retrieval actually improve on this material? A real condition might be that Fernwood's reference staff run twenty queries whose answers they already know — including ten items the current catalog search misses — against both Kestrel's output and Fernwood's existing search, and the next stage begins only if Kestrel surfaces a clear majority of those ten.

Two cautions. First, until the trial runs, that condition is a proposal, not a result. Nothing here has been measured. Second, it answers only a retrieval question. It does not establish that visitors will find more, which is a later and separate question with a different owner and its own evidence.

Name the decision-makers as roles, not predictions. Fernwood's director and head of collections decide whether to commit the remaining preparation. Kestrel's product lead decides whether archival support enters the roadmap. Naming a role is not claiming the person will say yes.

Everything that involves money — compensation, valuation, ownership of data or rights, binding obligations — belongs to the commercial and legal owners. The proposal's job is to expose the decision clearly enough that they can make it. It should not claim that a named partner will accept it, or that the arrangement is objectively fair. It should not try to.

A proposal each side can answer on its own timeline

What goes to Kestrel is short. Three lines: what Fernwood will do now (bounded, one person, one week, two hundred transcripts); what Fernwood needs to see before committing the rest (the retrieval condition above, judged by reference staff); and what happens if either side stops.

The stopping route is the line most proposals omit, and it is the one that makes the rest credible. If Kestrel declines the second stage, Fernwood keeps a rights list for the sampled series and a cleaner catalog for those items, and Kestrel keeps its learning. If Fernwood declines, Kestrel keeps the learning too. Neither outcome is a catastrophe, which is precisely why a yes can be trusted. A deal that is only safe when it succeeds is not safe.

Notice that even the stop is uneven — Fernwood would have spent a week and kept something modest, while Kestrel would have kept product knowledge. That is another reason to bound stage one tightly. The right size for the first commitment is not the size that feels fair. It is the size that survives the deal not continuing.

Shared purpose is compatible with unequal value. Kestrel can genuinely want the collaboration and still gain earlier and more surely than Fernwood does. Fernwood can genuinely want it and still be the side that commits first. The pitch's job is not to make that inequality disappear. It is to make it visible enough that both sides can see what they are choosing — and choose freely, including the choice to stop.

Frequently asked questions

Why can a two-column partnership slide hide the deal?

The columns sit at the same height, so the work looks symmetrical even when it is not. The slide can also flatten time: one side may spend now and be paid later in a currency it does not control, while the other spends a little now and keeps something useful either way. Two timelines should be on the table before anyone says "shared success."

What asymmetries matter in the Fernwood and Kestrel example?

Reversibility matters: six weeks of two people cannot be unspent, and while sharing transcripts can be discontinued, learning drawn from them cannot be unheard. Who keeps value if the trial stops also matters: Fernwood's durable benefit is downstream and shared, while Kestrel's learning is present and retained. That is an asymmetry of certain and retained value on one side versus contingent and deferred value on the other. "Largely not recoverable" matters because Fernwood's best outcome depends on a decision Kestrel has not made, even though some preparation, such as a rights list, keeps value.

Which axis can negotiation move cleanly, and what does that mean?

Negotiation can move timing: you can stage a commitment, delay a cost, or attach a later benefit to a named condition. It cannot negotiate kind, because learning stays learning, and it cannot manufacture certainty without having someone observe something by running part of the work.

What four structures can change the exposure?

A smaller first activity narrows the question without answering the big one. A staged commitment can produce a real go/no-go, but it shrinks the bet without changing who is exposed, and it is theater if the condition cannot fail. A changed contribution moves mechanical work to the party with tooling, but requires real agreement and capacity. No collaboration in the current form is the option to decline or redesign.

What makes a stage condition real, and what would justify declining?

A milestone should retire the specific uncertainty that made staging useful; elapsed time is not a milestone. The condition has to be something that might fail, with decision-makers named as roles rather than predictions. If a partner will not take on more preparation, will not accept a stage condition that could fail, and will not commit the deferred benefit to anything concrete, the current form asks one side to spend staff weeks on a result it does not control. That is a proposal problem to describe, not an accusation about motives; declining preserves the collection and the option to try again under different terms, though Kestrel would learn nothing.

Further reading

• Your Pilot Succeeded With Heavy Support. What Does That Prove About a Wider Rollout? — Before treating a small first stage as proof, read this on separating the tested arrangement from the proposed one and stating what the wider offer would remove.