Demand you cannot reach without another organisation's continuing permission is not an asset you hold. It is an arrangement you are being allowed to continue.
A growth paper reaches the board with a good number in it. Acquisition volumes are up, cost per acquisition is holding, and the channel mix shows a single platform supplying most of the new demand. The discussion moves to whether to spend more there.
Put the same concentration on the supply side and the conversation changes entirely. One vendor, no qualified alternate, no agreement of substance, no notice period: that is logged as a single-source exposure, given an owner and a mitigation plan, and revisited until a second source exists. Demand-side concentration attracts none of that discipline, because it arrives dressed as marketing expenditure rather than as a dependency.
Two different failures hide behind that costume. The first is concentration — the enterprise has one route to its customers and does not control it. The second is control in the legal sense: the assets that would let it rebuild a route are frequently held by someone else, and nobody has checked whose name is on them.
The Strategic Context
For a great many enterprises the demand-generation asset is no longer a sales force, a shelf position or a branch network. It is a position inside another organisation's ranking, matching or recommendation system. That position is real, it is valuable, and it is not owned. It is granted, priced by the grantor, and revisable without consultation.
The allocation consequence is where the damage compounds. Rented demand is an operating cost with immediate, attributable return. Owned demand — direct relationships, a domain, first-party records, a reason for customers to arrive without being intercepted — is an investment with a long payback and weak attribution. Put the two in the same business case and the rented channel wins every time, on the merits as measured. The measurement system therefore defunds the only asset that reduces the exposure. That is not a marketing failure. It is a capital allocation failure wearing a marketing label.
Access Is Being Read as Ownership
Daily use feels like possession. The team logs in each morning, the invoices are paid, the revenue arrives, and the relationship acquires the texture of property. Three misreadings follow.
An audience is treated as a relationship. Reach supplied by a platform is borrowed attention, and a follower count is an index of the grantor's willingness to show you to people — not a list you can address.
Silence is treated as stability. No dispute has occurred, so rights are assumed to exist. Standard platform terms are generally unilateral and revisable, and an appeal against an automated enforcement decision is usually not an appeal to a person with discretion.
Spend is treated as leverage. A large account makes an enterprise a valued customer of the platform. It does not make it a counterparty with remedies.
Reframing the Issue
Most leadership teams carry two categories: things that are ours and things that are somebody else's. The useful model has three. Owned — the enterprise holds legal title and can operate the asset without anyone's permission. Contracted — someone else performs, but under a bargained agreement with notice, cure rights and a remedy when it fails. Granted — access continues at the discretion of the grantor, revocable, without notice and without a forum.
Most digital demand sits in the third category and is governed as though it sat in the first. The right measure is therefore not how much traffic a channel supplies. It is what proportion of demand the enterprise could still reach on the morning the grantor says no — a portability question, and one that belongs to the board rather than to the marketing function.
Demand Concentration Deserves the Discipline You Give to Supply
Procurement has a settled method for single-source exposure and it transfers without modification. What is the substitution lead time? What does qualifying an alternate cost, and who holds that budget? What happens to margin if the effective price rises materially and there is no credible threat to leave? What fails simultaneously if it stops?
The last question separates a rented channel from a single supplier. A supplier failure interrupts one input. A change to a ranking or matching rule moves every product, every region and every segment at the same moment, because all of them depend on the same mechanism. The exposure is correlated, and correlation is what makes a risk expensive rather than merely large.
The same question is now arriving one layer beneath the channel, where enterprises build products on capability they do not own either — though whether that capability is something you sell or something you use is a separate decision, with different economics and different owners [Related article: Is Your AI a Feature You Sell, or a Tool You Use?].
Control Is a Register Entry, Not a Habit
Ask who the registrant of record is for the primary domain. Ask which payment method renews it, which mailbox receives the renewal notice, who holds administrative access to the advertising accounts, the analytics property, the application listing and the customer records, and what happens to each when that person's engagement ends.
The failures those questions find belong to one class: assets the enterprise depends on that were never brought under its legal control. A mark used commercially for years and never registered belongs to the same class. So does supplier-developed source code where escrow was discussed, agreed in principle and never documented. Each works perfectly until the relationship ends, and each is discovered at the moment of dispute, which is the most expensive moment available.
The remedy is unglamorous and takes a fortnight: a register naming every asset, the legal entity holding it, the renewal date, the people with administrative access, and the succession path for each. Which entity ought to hold them, and what that structure does and does not shield, is a separate governance question [Related article: What Limited Liability Actually Excludes]. Protecting them once title is clear is a third discipline again, and one routinely mistaken for a technical matter rather than a decision about risk appetite [Related article: The Padlock Is Not the Control].
Decision Framework
| Tier | What you hold | If it ends | What remains next morning | Governance it earns |
|---|---|---|---|---|
| Owned | Title; operable without permission | Your decision, your timing | The asset and its output | Asset register, named custodian, renewal discipline |
| Contracted | A bargained agreement | Notice, cure rights, a remedy | Time to substitute, and a claim | Contract review, counterparty concentration limits |
| Granted | Access at another's discretion | Immediately, without a forum | Only what you built elsewhere | Risk register entry, board-set exposure limit, tested alternative |
Three tests keep the classification honest. The withdrawal test: run the scenario at executive level rather than inside the marketing function, and require the answer in revenue and in weeks. The register test: the enterprise should produce registrant records, renewal paths and access lists within a working day, and an inability to do so is itself the finding. The substitution test: what it costs and how long it takes to stand up an alternate route, decided before it is needed, because optionality is only ever available for purchase in advance.
From Strategy to Execution
Immediately, recover title and close key-person access. This is administrative work of no strategic interest whatsoever, which is precisely why it is still outstanding, and it is the only item here that can be finished this quarter.
Over the medium term, build addressable demand and treat portability as a standing metric rather than a campaign. It will not survive a return-on-investment comparison against the rented channel and should not be asked to. It is bought for optionality, which means it needs a named benefits owner, a protected budget line, and a sponsor senior enough to defend it from that comparison.
In the long term, channel architecture becomes a position rather than an accident. An enterprise that can leave a platform negotiates differently from one that cannot, and it can accept a dependency deliberately, on stated terms, having priced what withdrawal would cost. Dependence chosen and funded is a strategy. Dependence discovered during an incident is a consequence.
Signals to Monitor
Direct and branded demand flat while total demand grows. Effective acquisition cost drifting upward with no explanation you can contest. Changes to platform terms, attribution rules or category policy — particularly the platform entering your category. Renewal notices arriving at a mailbox nobody reads. An agency or contractor that resists transferring administrative access. Credentials held by one individual. And the quiet one: a growth plan whose next stage requires the grantor to keep behaving exactly as it does now.
Questions for the Leadership Team
- If our principal demand channel withdrew tomorrow, what share of next quarter's revenue could we still reach, and how quickly could we evidence that figure?
- Who is the registrant of record for our domains, and which entity's name is on the customer records we describe as ours?
- What did we spend this year building demand we can reach without permission, as a proportion of what we spent renting it?
- Which executive is accountable for the demand-generation asset itself, as distinct from the campaigns run through it?
- What would we need to observe in a platform's behaviour before we funded an alternate route, and have we written that threshold down anywhere?
Closing Perspective
Ownership is not a feeling produced by daily use. It is an entry in a register and a clause in an agreement, and an enterprise that can produce neither is renting while reporting as though it owns.
The choice facing most leadership teams is not between platforms and independence; refusing the scale a platform supplies is rarely the commercially serious answer. It is between dependencies chosen, priced and funded, and dependencies inherited by default and discovered under pressure. The second kind is not cheaper. Its cost is simply deferred to a date the enterprise does not select.
About the author
Kevin Jogin is Founder & Principal Advisor at EraNorth. Meet the Founder.