Your First-Party Data Is Worthless Until You Use It
Marketers have spent billions building sophisticated customer-data infrastructure. Far too little of that intelligence ever reaches the platforms where their advertising dollars are actually spent.
This Signal & Noise exclusive is brought to you by AUDIENCES.
The Valuable Data Marketers Barely Use
Over the past five years, marketers have developed a serious first-party data dependency. To satisfy this addiction, they’ve bought CDPs, cloud data warehouses, data clean rooms, identity graphs, and consent management platforms galore. They’ve unified customer profiles, compressed records, mapped journeys, created golden records, and sat through more presentations featuring the phrase “single source of truth” than any reasonable person should have to endure—I know, because I have been one of the people giving these presentations.
The result is an extraordinary—dare I say unprecedented—amount of customer intelligence. Brands not only know who buys from them, but they know who buys repeatedly, who hasn’t bought in six months, who has a high lifetime value, and who returned three pairs of shoes after wearing them to a wedding. They can identify their best customers, their worst customers, their newest customers, their former customers, and the people who are so unhappy should never receive another acquisition ad from them again.
Then the paid media campaign launches... and much of that incredible intelligence stays exactly where it was. The CRM knows one thing, the CDP knows another, and the analytics team has a beautifully governed dataset in the cloud. Meanwhile, the media team is targeting a platform-defined audience called something like “Urban Lifestyle Enthusiasts” and hoping for the best. Sound familiar?
Truth be told, this has been one of paid media’s most persistent (and costly) problems for years. I saw it in my earlier work with Amazon and Amazon Marketing Cloud, where a shockingly low number of brands were actually uploading any first-party data into their clean room, and I’ve encountered it repeatedly since. While nearly everyone agrees high-quality first-party data can improve targeting, suppression, lookalike modeling, optimization, and measurement, actually getting that data into paid media—safely, legally, accurately and fast enough to matter—is where the wheels come off.
In many cases, the obstacle is technical. The systems don’t connect, identifiers don’t match, or the data arrives through a manual upload that is already woefully out of date by the time someone finds the correct CSV. In other cases, the root cause is organizational. The people who actually control the data sit across IT, analytics, CRM, InfoSec, or legal, which means we're several meetings—and at least one minor turf war—away from the people buying media. And sometimes the problem is governance. Customer data is valuable precisely because it is sensitive, which of course means activating it requires clear permissions, careful controls, and agreement among people whose professional incentives range from “move faster” to “absolutely do not move anything.”
End of the day, every one of those concerns is legitimate. But together, they help explain why so little paid media is informed by one of the most valuable assets a brand owns. The opportunity cost begins with the simplest of use cases. For example, a brand keeps paying to target people who are already happy long-term customers. Or it retargets a person who just purchased the exact same product being advertised. Believe it or not, some brands will spend equally to target customers worth $40 and ones worth $4,000 because the media platform cannot see the difference.
The missed opportunity gets larger from there. First-party data can help brands build models from their highest-value customers rather than from everyone who happened to click “buy.” It can connect campaign exposure to real business outcomes, and it can tell marketers which audiences produce lasting customer value instead of cheap conversions that look impressive in a weekly report and disappear by quarter’s end.
No, none of this makes first-party data magical. Bad data is still bad data, even after an expensive technology platform gives it a nice dashboard or ad hoc connections to various platforms. But clean, consented and strategically useful customer data should play a far larger role in paid media than it does today. Brands have spent years building this asset. The urgent question now is whether anyone can get the treasure to the media team without putting it in a CSV, attaching it to an email, and asking legal to please respond before the campaign ends.
The Last Mile Is Where Data Strategies Go to Die
On paper, first-party data activation sounds almost insultingly simple. Identify an audience, send it to a media platform, run a campaign, and measure the results. Anyone who has actually attempted this knows that “send it to a media platform” is not simple and conceals an extraordinary amount of work.
Let’s say, for example, a retailer wants to reach high-value customers who purchased from a particular category more than six months ago but haven’t returned since. The company has the transaction history, customer profiles, consent records, product taxonomy, and lifetime-value calculations needed to build that audience. The use case is clear, and the potential value is obvious.
Now the fun begins. Someone has to define the segment correctly, determine which systems contain the required fields, confirm the data can legally be used for advertising, extract the records, normalize the identifiers, remove ineligible customers, format everything according to the destination platform’s specifications, and transfer it without creating a security incident or triggering a legal panic attack.
Then the platform has to match those records to its own users.
A customer file containing one million people does not become an addressable audience of one million people. Records may be incomplete, outdated, incorrectly formatted, duplicated, or connected to identifiers the platform cannot recognize. Customers may have used a different email address, changed phone numbers, opted out, or simply disappeared inside the platform’s identity machinery. Each step quietly chips away at the audience.
By the time the segment becomes available for activation, it may be considerably smaller than expected—and several days, weeks, or even months older. That last point matters more than many marketers realize. Customer data has a shelf life. A list of people who abandoned a shopping cart yesterday may be extremely useful, but the same list activated three months later is basically a historical document.
Yet many activation workflows still operate like a clunky accounting process. Teams define a segment, open a ticket, wait for approval, export a file, upload it somewhere, wait for it to process, check the match rate, discover a formatting problem, fix the file, and upload it again. Eventually the audience appears in the buying platform, where somebody has to remember why it was created in the first place.
This workflow can technically be described as "data activation," but then again so can FedExing a floppy disk.
The real goal for data activation should be a continuously operating connection between customer intelligence and media execution. When a customer buys, churns, upgrades, opts out, becomes more valuable, or stops qualifying for an audience, that change should flow into campaign decisions quickly and reliably. Unfortunately, that rarely happens today. Instead, campaigns continue targeting yesterday’s customer. Suppression lists go stale, and high-value audiences are refreshed sporadically. Media teams optimize against platform events while richer business outcomes sit in another system. The company may possess a detailed, current understanding of its customers, but its advertising platforms receive an occasional update.
AI raises the stakes considerably. An intelligent media system can make thousands of decisions about targeting, bidding, creative, and budget allocation, yet those decisions will only be as useful as the signals reaching it. Feeding an AI system delayed or incomplete customer data simply allows it to make outdated decisions much faster and at greater scale.
The current state isn't working and the future of first-party data activation has to involve more than moving larger audience files. Marketers need faster signal delivery, persistent connections, automated updates, clear permissions, reliable identity resolution, and feedback loops that connect media exposure to real customer outcomes.
This is what I'm referring to as the "last-mile problem." Most brands have already built the roads, warehouses, and distribution centers. The package is sitting on the truck, but they just can’t seem to get it to the front door.
Stop Paying to Advertise to the Wrong People
The easiest way to understand the value of first-party data is to start with the least exciting use case: knowing who NOT to target. It's not a sexy topic and audience suppression will probably never headline a keynote at Cannes. Let's face it, nobody is going to stride dramatically across a stage and announce that the future of marketing is showing ads to fewer people. But suppression may be one of the fastest and most reliable ways first-party data can improve media performance. It's low-hanging fruit, people.
Consider the customer who just bought a refrigerator. Unless your company has developed a refrigerator that people collect like sneakers, continuing to serve that person refrigerator ads is unlikely to generate any additional sales. This practice wastes money, irritates the customer, and demonstrates that basic intelligence supposedly flowing through the marketing organization has failed to reach the advertising campaign. Epic fail.
The same problem appears everywhere you look. Banks market introductory offers to existing cardholders. Subscription companies chase people who already subscribe. Retailers promote products customers just purchased, and travel brands retarget people for trips they have already taken. Occasionally, brands even spend money trying to reacquire customers they recently removed for fraud or nonpayment. Every one of those impressions may look perfectly legitimate inside a media platform. It was delivered to a targetable person, counted correctly, and included in the campaign report. It was also a waste of money and effort.
Reliable suppression changes the economics before a marketer makes a single improvement to targeting. It removes obvious waste, reduces unnecessary frequency, and leaves more budget available to reach people who might actually buy something. It can also radically improve the customer experience, because let's face it no one likes it when a brand behaves as if every interaction is the first time it has met you. But suppression is only the starting point. Once marketers connect current customer intelligence to paid media, they can begin making meaningful distinctions among the people they want to reach. A customer who purchases twice a year should not necessarily receive the same offer, bid, or message as someone who purchases twice a week. A customer with a high probability of churning may require a different strategy from one who has steadily increased spending for three years.
Customer Lifetime Value, or CLV, makes those differences especially important. Most advertising platforms optimize toward the events they can observe: clicks, leads, app installs, purchases, and other immediate actions. CLV is defined as the total amount of money (or profit) a company expects to make from a single customer over the entire time they stay with the company. Connecting CLV to paid media helps reveal whether a campaign acquired loyal, profitable customers, or attracted people who bought once with a discount code and were never seen again.
First-party data can provide this missing context. It can help marketers build seed audiences from genuinely valuable customers, create exclusions based on recent activity, adjust bids according to predicted value, and develop lookalike models based on the people the business actually wants more of. This is a meaningful improvement over selecting an off-the-shelf audience based on a vague combination of browsing behavior, inferred interests, and whatever else went into producing “Affluent Lifestyle Influencers.”
First-party data is equally important after the impression. Connecting media exposure to transactions, renewals, retention, and CLV allows marketers to measure performance against business outcomes rather than platform-defined proxies or vanity metrics. A campaign that produces inexpensive conversions may look outstanding... until the customer data reveals those buyers rarely return and have low CLV. Another campaign may appear more expensive yet consistently acquiring high-value customers who remain loyal for years. Without this feedback, the algorithm optimizes toward the easiest observable action, and unfortunately, the easiest action to generate is not always the one that makes the company money in the long run.
Suppression, value-based targeting, lookalike modeling, and closed-loop measurement are not esoteric ideas. Most marketers already understand them and have been discussing them for years. The remarkable part is how rarely they operate together using current, high-quality data. The value is there, the use cases are well understood, and the media budget is already being spent. The missing piece is the connection.
AI Makes the Last Mile Much More Important
For years, marketers could get away with treating first-party data activation like a periodic exercise.
New campaign coming up? Great, refresh the audience, upload a new suppression file at the beginning of the month, or maybe rebuild the lookalike model. When all else fails, send an updated customer list whenever somebody realizes the current one is way out of date. This was never a particularly good way to operate, but it was survivable when media decisions moved at roughly human speed.
AI changes the equation. Advertising platforms already use machine learning to make enormous numbers of decisions about bids, audiences, placements, frequency, creative, and budget allocation. Tomorrow's agentic systems will take this considerably further. Instead of waiting for a media buyer to adjust a campaign, an AI agent may continuously analyze performance, shift spending, test audiences, alter bids, recommend creative, and negotiate inventory across multiple platforms.
That sounds impressive—and it can be. But an intelligent system operating with stale customer data is still operating with stale customer data. The underlying problem is speed itself. Because AI can make more decisions in an hour than a media team could make in a week, if the inputs are outdated, incomplete, or disconnected from actual business outcomes, the system can also compound mistakes with breathtaking efficiency.
Imagine a customer purchases a product on Monday, but the suppression audience does not refresh until Friday. A human-managed campaign might waste a few impressions during that gap. An AI system optimizing aggressively across channels could retarget the customer dozens of times, increase the bid because the person recently demonstrated purchase intent, and decide the campaign is performing beautifully because the transaction appeared after an ad exposure. The machine may even slap itself on its metaphorical back to congratulate itself for causing a sale that had already happened.
The same issue applies to customer value. An AI system can identify patterns, model behavior, and optimize spending far faster than a human team. But if it receives only platform conversion data, it will optimize toward the customers who are easiest to acquire—not necessarily the customers the business should be trying to acquire. A cheap conversion is catnip to an algorithm. Whether that customer stays, returns the product, defaults, cancels after the promotional period, or spends thousands of dollars over the next five years is invisible to the system.
First-party data supplies the context required to distinguish activity from value. Its signals can tell the system that a purchase happened, but also what happened after the purchase took place. It can reveal whether the customer returned, upgraded, churned, became profitable, or turned into a long-term advocate for the brand. For AI-driven media to deliver on its promise, these signals cannot arrive via a monthly file upload. They need to move continuously along with the permissions, definitions, and governance necessary for the system to use them appropriately.
This last qualification matters. Giving an AI agent access to more customer data does not give it unlimited authority to use that information. I wrote about this recently in my article titled "The Consent Advantage: Permission Is the Foundation of AI-Driven CX." The crux of the argument is the data may be approved for measurement, but not targeting. It may be usable in one market, but restricted in another. A customer may revoke consent, a campaign may involve a regulated category, or a sensitive attribute may require additional controls.
A functioning activation layer therefore needs to communicate more than audiences. It must also communicate purpose, permission, recency, provenance, and the rules governing how each signal may be used. This is where the conversation about AI in marketing often gets ahead of reality. We spend plenty of time debating what increasingly autonomous systems will do, but we spend far less time asking whether those systems can access the right data, understand its meaning, and use it within the boundaries the organization (or the individual) has established.
Without this foundation, agentic marketing becomes a very fast driver navigating with an old map.
I suspect the AI era will reward companies that shorten the distance between what they know about their customers and what their media systems can act upon. The competitive advantage will come from the speed, quality, and governance of that connection. In other words, real-time intelligence is only valuable when it can lead to real-time action.
Closing the Gap Is Not Really About Tech
The way most marketing departments respond to issues like this is to buy something, usually another tool. Is there a gap between customer data and media activation? Great, let's create a vendor shortlist and schedule some demos, assemble the procurement team, and prepare for a series of presentations in which every slide is labeled “AI-native.”
Jokes aside, technology obviously matters here to a certain extent. Let's face it, nobody is going to solve continuous data activation with positive thinking and more cross-functional meetings. But most brands already own plenty of sophisticated marketing technology, including at least several platforms that ostensibly can unify, transform, govern, enrich, and activate customer data. But adding yet another box to the reference architecture diagram does not necessarily make the data move any faster.
Closing the last mile requires a functioning operating model connecting the systems where customer intelligence lives with the platforms where media decisions happen. That connection not only has to do several things well, but it also has to do them continuously.
First, marketers need to begin with specific business decisions rather than an abstract ambition to “activate more data.” Let's start with concrete use cases. Which customers should be suppressed? Which behaviors indicate churn? What separates a valuable customer from a cheap conversion? Or which signals may affect a bid, audience, message, or budget decision? Without this level of clarity, a company can spend months integrating systems and still produce little more than a larger collection of targetable lists.
Audience definitions themselves also need to remain consistent. “High-value customer” cannot mean one thing in the CRM, another thing in the analytics environment, and whatever the media agency inferred from a spreadsheet it sent over six months ago. The logic used to create an audience should be transparent, reusable, and connected to the business definition that gave the audience value in the first place.
Then comes the connection itself. Data needs a reliable path connecting the systems where it is maintained to the destinations where it can be activated. That path has to account for different identifiers, schemas, APIs, refresh schedules, authentication methods, and platform requirements. It also needs to handle additions and removals. Adding a newly qualified customer to an audience matters, but removing someone who purchased, opted out, or stopped qualifying may matter even more.
This is where many activation strategies reveal their age. Many clients I have worked with have processes in place that were designed to periodically move a complete file instead of continuously communicating what changed. To solve for this, a more useful model is event-driven. When a meaningful customer event occurs—a purchase, renewal, cancellation, consent change, loyalty milestone—the relevant media systems should receive an updated signal. The entire customer database does not need to take a field trip every time someone buys a pair of pants.
Signals also need rules attached to them. Teams should know where the data came from, when it was updated, where it may be used, and when permission expires. Otherwise, faster activation simply creates a faster route to compliance issues down the road. Someone must also be able to tell whether the process is working. This means marketers need visibility into audience size, match rates, refresh times, delivery failures, consent status, and the destinations in which each segment is active. If an audience unexpectedly loses half its members due to unsubscribes or fails to update for two weeks, this should trigger an immediate alert—not an awkward discovery during the quarterly business review.
Finally, the connection must work bi-directionally. This means customer intelligence should inform media execution, and media activity should flow back into the company’s measurement environment in a closed loop. Marketers need to understand which audiences were reached, what they were shown, what actions followed, and whether those actions created business value. Without this feedback loop, activation remains a dumb outbound delivery exercise. Media platforms receive customer data, make decisions inside their environments, and report performance using their own definitions of success. The brand contributes its most valuable intelligence to the platform, but the platform keeps much of the learning.
That is data donation, not a closed loop. A legitimate data activation capability connects audience creation, permission, identity, delivery, optimization, and measurement as parts of the same overall process. It replaces occasional uploads with persistent connections, and turns audience management into a living system rather than a collection of files.
The brands that solve this will be the ones making their customer intelligence usable at the moment a media decision is made. This is what closing the last mile actually looks like.
Data Should Never Leave Home
For decades, the standard approach to data activation has been to move the data from one place to another. Extract it from one system, transform it into the required format, send it somewhere else, and hope the receiving platform can match enough of it to make the exercise worthwhile. If the audience changes, repeat the process. If another platform needs the same audience, make another copy and send that one somewhere else too.
The problem is every data movement creates additional work and incremental risk. Files need to be generated, encrypted, transferred, processed, reconciled, monitored, and eventually deleted. Each destination has its own requirements, match logic, permissions, and failure modes. Every new copy creates another place where data can become stale, fall out of sync, or be used in a way nobody originally intended.
The irony is many brands have spent years centralizing customer data inside secure cloud environments and hardening their IT infrastructure precisely so they can govern it more effectively. They have invested in access controls, encryption, clean data models, consent records, identity management, clean rooms, and carefully monitored infrastructure. Then, when it is time to activate the data, they export it and ship it somewhere.
A better approach begins by treating the company’s existing cloud environment as the source of truth. Audience logic can operate where the data already lives, using the current customer records, business rules, and permissions. The activation layer can then communicate the required audience signals to media destinations without creating an unnecessary new repository for the underlying customer data.
This is a different paradigm that changes more than just the plumbing. When audience definitions remain connected to the original data, they update as the data changes. This means a customer who purchases is removed from a prospecting audience, and someone whose lifetime value crosses a threshold enters a high-value segment. A person who withdraws consent is removed from eligible activation. These changes should happen through an ongoing connection rather than waiting for the next file refresh.
This approach also creates a cleaner division of responsibilities. The brand maintains control of its customer intelligence, business definitions, and permissions, while the activation layer handles the yeoman's work of translating those decisions into the schemas, identifiers, and technical requirements native to each destination. Media platforms receive the signals required to execute the campaign without becoming the primary home for the brand’s customer strategy.
In coming years, this distinction will become increasingly important as AI systems assume a larger role in media execution. As tasks are increasingly delegated to AI, organizations will conclude that an agent does not need unfettered access to an entire customer database to decide whether someone belongs in a particular audience. All it needs is an accurate signal, the context required to interpret that signal, and clear rules governing what it may do with it. Giving every marketing system access to everything may eliminate some integration problems, but would certainly open up other proverbial cans of worms, creating exciting new opportunities for lawyers, regulators, and incident-response teams.
Data minimization implies a more sustainable model. This means moving the smallest amount of information necessary to accomplish a defined purpose, keeping sensitive customer intelligence inside the environment designed to protect it, and preserving a record of which audiences were activated, where they were sent, and what permissions were applied at the time. This does not, however, eliminate every challenge. Media platforms still rely on identity matching, destination APIs still change, consent requirements differ across markets, and of course closed-off platforms retain considerable control over what advertisers can send, observe, and retrieve.
But keeping audience creation connected to the original data removes a major source of latency and fragmentation, not to mention significantly reduces compliance risk. It reduces the number of copies that must be governed and gives marketers a better chance of activating audiences that reflect what they know about their customers now—not what they knew when someone exported the file. The industry has spent years moving customer data into more places, so the next phase should be about making it useful in fewer.
Activation Is the New Competitive Advantage
For much of the past decade, marketers competed on their ability to collect customer data. The brands with the largest customer files, most sophisticated CDPs, cleanest identity graphs, and most elaborate cloud environments appeared to have an obvious advantage. I believe this phase of marketing is coming to a close.
Let's face it, most large organizations now possess more customer data than they can ever effectively use. They have transaction histories, loyalty activity, website behavior, service interactions, product preferences, consent records, and predictive scores scattered across an impressive collection of systems. The problem is no longer whether the information exists, but whether what the company knows can influence a media decision before that knowledge becomes irrelevant.
This makes data activation—not accumulation—the next competitive battleground. A brand that can identify a valuable customer but needs three weeks and seven approvals to reach that person does not have a meaningful data advantage. Neither does a company that can calculate churn risk with remarkable precision but continues serving acquisition ads to customers who canceled yesterday. Intelligence trapped inside an enterprise system may be analytically interesting, but it is commercially inert.
Organizations that get this right will build a much tighter loop between customer behavior and media action. This means a purchase will update suppression, a change in CLV will affect bidding strategy, and a change in consent status will propagate across destinations. Campaign exposure must flow back into the company’s data environment where it can be connected to revenue, retention, returns, and other outcomes the advertising platform cannot see on its own.
Importantly, this does not mean every decision must happen instantaneously. “Real time” has become one of those marketing phrases applied indiscriminately to everything, including processes that would be perfectly fine happening tomorrow or at least within a few hours. The appropriate speed depends on the use case. Cart abandonment may require action within minutes, but churn prevention can operate daily. Customer-value models may only need to refresh weekly. The point is not to make every signal move at maximum velocity. It is to ensure that each signal moves fast enough to remain useful.
Getting there will require marketers to think differently about the technology they have already purchased. A CDP is not valuable simply because it creates a unified profile. That capability matters only when it unlocks better decisions downstream. Nor is a clean room useful because two parties can put data inside it without frightening their respective legal departments. A cloud warehouse is not a critical piece of architecture because it contains an immaculate table with 400 customer attributes.
These systems create value when the intelligence inside them changes what the company does and how it does it. This means evaluating first-party data investments according to the decisions they improve. Did suppression reduce wasted impressions? Did value-based targeting attract more profitable customers? Did better signals improve the models and decisions governing the campaign? If the answer is no, the organization may have built an impressive data estate without constructing a road out of it.
There is also a larger strategic issue at stake. Brands that cannot activate their own intelligence remain dependent on whatever audiences, optimization events, and measurement frameworks the major platforms choose to provide. Those platforms may be extremely capable, but they are black boxes and their incentives are not always identical to the advertiser’s. They can optimize toward the outcomes they observe, using the data available inside their walls, but they cannot independently know which customers become profitable, which transactions are returned, or which conversions the business considers genuinely valuable.
First-party data gives the advertiser a way to bring its own definition of success into the media system.
As AI assumes more responsibility for execution, if anything this definition will become even more important. The agent optimizing a campaign needs to know what the company actually values, not which button the platform has labeled “conversion.” Brands that supply current, governed, business-specific signals will be able to direct these systems with considerably greater precision. Everyone else will be deploying increasingly sophisticated technology to chase increasingly generic outcomes.
Marketers have already spent billions of dollars collecting, cleaning, organizing, enriching, and protecting first-party data. The vault is built and the treasure exists. Now somebody has to figure out how to use it. The winners will not necessarily be the brands with the most data. They will be the ones that can turn what they know into action—safely, continuously, and fast enough to matter.
------------------------------------------------
Rio is an executive with 20+ years at the intersection of strategy consulting, AdTech, data, and media. He's a trusted advisor on customer experience, digital strategy, and marketing transformation. He's a partner at Credera, Omnicom's consulting arm. He's also a podcast host, writer, and public speaker focused on the future of advertising and AI-driven infrastructure.
Comments