The most valuable dataset I have ever worked with was not scraped, bought, or inferred. It was typed, by customers, on purpose. A quiz, a preference form, a "tell us what you are trying to fix" box. Thousands of people voluntarily describing their own intent with a precision no identity graph on earth could match. The media plan built on it embarrassed everything we had run on modeled audiences. That experience turned me from a data volume guy into a data provenance guy, and I have never gone back.
The industry calls this zero party data: information a customer intentionally and proactively shares. The label is clumsy but the category is the future, because it is the only class of data whose supply grows as privacy regimes tighten. Third party data is dying by law. Inferred data is degrading as its inputs get cut off. Volunteered data has no such ceiling. Its only constraint is whether people want to tell you things, which means its economics are the economics of trust, and trust is a market most marketers have never had to price.
This paper is about that pricing. Not the soft "build trust" sermon, but the actual economic machinery: why declaration outperforms inference, what it costs to acquire, how it depreciates, and why the brands treating volunteered data as a balance sheet asset will spend the back half of this decade beating the brands still hoarding surveillance exhaust.
Key Findings
- Volunteered data beats inferred data on the three dimensions that matter commercially: accuracy, permission, and durability. Inference only ever won on volume, and volume without accuracy is a cost center.
- Every act of declaration is a purchase. The customer spends effort and disclosure, and expects a return. I formalize this as the Reciprocity Ledger: an unwritten account every customer keeps of what they told you versus what they got for it.
- The yield on volunteered data collapses when brands collect without visibly using. Unused declarations are not neutral. They are withdrawals from the ledger.
- Declared data depreciates. Intent statements age in weeks, preferences in months, identity facts in years. Most brands store all three in one field and refresh none of them.
- By 2030 I expect declared data capture to be a designed, budgeted discipline with its own funnel metrics, as standard as email capture became in the 2010s.
- Inventory your declared data separately from everything else. Most companies cannot answer "what have customers explicitly told us, and where does it live?" Answer it. That inventory is your most defensible asset.
- Fund the exchange. Every ask needs a visible give: better recommendations, real utility, early access, saved time. Budget declared data capture like paid acquisition, with a cost and a conversion rate per field.
- Close the loop within one session where possible. The fastest way to build the Reciprocity Ledger is to use a disclosure the moment it is made, in front of the customer.
- Put depreciation dates on declared fields and build refresh moments into the lifecycle. A preference without a timestamp is a rumor.
- Never use a declaration in a way the customer would not recognize as connected to why they shared it. One creepy reuse can zero out a ledger it took years to build.
Declaration Beats Inference, and Not by a Little
Start with the comparison the adtech industry spent twenty years avoiding. Inference guesses what you want from what you did. Declaration asks. The guess is cheap per record and wrong at a rate nobody likes to publish. The answer is expensive per record and correct almost by definition, because the customer is the world's only authoritative source on their own intent.
Consider the difference in kind, not just degree. An inference engine sees a user browsing strollers and concludes "expecting parent," a category that will chase that user for a year, long after the stroller was a gift for a cousin. A declaration says "shopping for my sister's baby shower, budget around two hundred dollars, event on the twelfth." One of these is an audience segment. The other is a brief. You cannot buy the second one from a broker at any price, because it exists only inside a relationship.
The strategic point: as covert collection gets regulated and blocked, inference quality falls while declaration quality holds. The gap widens every year in favor of asking. The foundational layer underneath all of this is still your owned data infrastructure, and I laid out that architecture in my first party data strategy guide. Zero party is the premium floor built on that foundation: first party tells you what people did, zero party tells you why, and why is where the margin lives.
The Reciprocity Ledger
Here is the mental model I want to install. Every customer maintains a Reciprocity Ledger with your brand: an informal running account of disclosures made versus value received. Tell you their size, their goals, their budget, their birthday, and each entry sits on the ledger waiting to be honored. Honor it, by making the next experience visibly smarter, and the customer's willingness to disclose again goes up. Ignore it, and something worse than nothing happens: the customer learns that talking to you is pointless, and the ledger closes.
This explains the pattern I see constantly in audits. A brand runs a beautiful onboarding quiz, captures rich preference data, and then sends the same batch and blast email to everyone. Six months later they wonder why quiz completion is falling. It is falling because the ledger is in deficit. Customers are not privacy absolutists. They are accountants. They will tell you almost anything if the telling visibly pays.
The ledger also explains why volunteered data cannot be shortcut. Buying "declared" data collected by someone else transfers the record but not the relationship. The disclosure was a deposit in another brand's ledger. Used by you, a stranger, it reads as surveillance, which is exactly how customers experience it. Provenance is not a compliance detail. It is the product.
The Depreciation Schedule Nobody Runs
Volunteered data is an asset, and like any asset it depreciates, on schedules that vary wildly by type. Intent declarations, what I am shopping for right now, decay in days to weeks. Contextual preferences, sizes, dietary needs, communication choices, hold for months to a couple of years. Identity facts, name, birthday, household composition, decay slowly but do decay, because lives change.
Most CRMs treat all of these as permanent facts, which produces the uniquely self defeating experience of a brand confidently personalizing on a declaration that expired two years ago. Stale personalization is worse than none, because it demonstrates that you were listening then and stopped. On the Reciprocity Ledger, it books as a broken promise.
The operational answer is refresh loops: lightweight, well timed re asks built into natural moments, post purchase, seasonal, milestone. Done right, refresh is not friction. It is the brand visibly keeping the file current, which itself signals that disclosure matters. The brands that master conversational refresh, increasingly through AI assisted interfaces where a customer can just tell the brand things in plain language, will hold declared datasets that compound instead of rot.
Volunteered Data in an Agent Mediated World
Now the forward edge. Over the next few years, a growing share of declaration will not be typed into your quiz. It will be transmitted by the customer's own AI agent: preferences, constraints, and intent shared programmatically with brands the customer authorizes. This is declaration at machine scale, and it changes the capture game entirely. The brand skill shifts from designing quizzes to being the kind of entity an agent is instructed to trust with its principal's brief, and to answering machine questions well when asked. I have written about marketing to AI agents as its own discipline, and volunteered data is where it collides with this series.
The economics of that collision deserve respect. When agents can carry a customer's declared profile between brands under the customer's control, declaration becomes portable, and portability creates market pressure to reward it properly. Whoever offers the best return on disclosure wins the profile. The adjacent question, whether customers start charging outright for access, is the subject of Paper No.34.
My falsifiable prediction: by 2029, at least one major commerce platform will support customer controlled, portable preference profiles that brands can request access to, and declared data acquisition cost will be a standard line in CAC reporting. If declared capture is still a side quest owned by nobody in 2030, I was wrong about the decade.
What I Would Do About It
Surveillance data told you what people looked like from behind. Volunteered data is people turning around and talking to you. The entire skill of the next five years is being worth talking to.