Oreo Dunked in the Dark. Marketing Has Been Chasing That Moment Ever Since

On February 3, 2013, during a 34-minute power outage at Super Bowl XLVII, Oreo posted an image of a single cookie with the line "You can still dunk in the dark."

The post earned about 15,000 retweets and 20,000 Facebook likes within a day on purely organic traffic, with no media spend and reached an audience comparable to that of a broadcast spot that cost around $4 million.

Marketers have cited it ever since, though the focus has often been on the wrong part, in my opinion. The part they missed is the same discipline that now separates brands that use agentic AI well from those that still struggle to do anything beyond basic marketing automation.

The flawed premise came from the assumption that this type of reach had to be planned and produced in advance. Traditional marketing teams planned against a calendar, made creative assets weeks or even months in advance and bought media against forecasts on the same (or longer) timelines.

The blackout tweet showed that an unplanned reaction, placed in the right moment, could outperform that model, breaking a longstanding approach and the assumptions that went along with it. In the ensuing years, brands also quickly learned that personalization, even in small ways, also had the potential to move the needle. So, an impossible-for-2013 paradigm was set: real-time, contextually-aware personalized brand moments became the holy grail for brands to achieve.

The 13 years since have been an effort to rebuild what Oreo actually had that night, though the effort started by misreading its most innovative feature.

Let’s look at the journey that marketers and martech have taken to build this capability at a scale that can truly deliver both the spontaneity and customization needed, and how we got from a Big Game tweet to 1:1 personalized marketing.

Power out? No problem. pic.twitter.com/dnQ7pOgC

— OREO Cookie (@Oreo) February 4, 2013

Oreo's Blackout Tweet Was Never Actually Spontaneous

As improvised as it may have felt in the moment, the tweet in question was the output of a well-prepared system. Oreo and its agency, 360i, ran a command center during the game, with representatives from the brand's other agencies present and an agreed-upon approval chain. Roughly 15 people were in the room. The brand had spent the prior summer running Daily Twist, 100 consecutive days of topical posts, which trained the team to move quickly inside a defined process.

When the lights went out, the team waited. After confirming that nothing serious had happened before posting (because a power outage at a stadium could have signaled something much worse), Lisa Mann, who ran the Oreo business at the time, later said the tweet took two years to make.

After the reaction and earned reach the single post received, the rest of the industry copied the creative approach behind it, yet skipped the operating model that made it possible. Preparation, delegated authority and the judgment to wait until it was deemed safe to post made the original post possible. While this may have felt like “real-time” marketing, it was far from it.

How Long Did the Oreo 'Dunk in the Dark' Tweet Actually Take to Prepare?

Despite posting within 34 minutes of the blackout, Oreo's team had spent the prior summer running 100 consecutive days of topical posts to train the same approval process — Lisa Mann later said the tweet took two years to make.

Why Newsroom-Style War Rooms Couldn't Scale Real-Time Marketing

With some of the personalization approaches and multi-channel listening and first-party data approaches marketers only dreamed of still years away, trying to capture lightning in a bottle (again and again), brands built newsrooms through 2013 and 2014 on the theory that simply putting creatives in a room would manufacture the next viral moment. This did not go as hoped.

Ahead of the following Super Bowl, Ad Age published an article that says it all: "Go Home, Real-Time Marketing. You're Drunk," prompted in part by a run of garbled live-tweets from J.C. Penney that were not able to capture that Oreo magic. Forced relevance produced some rather unfortunate outcomes, such as DiGiorno using a domestic-violence hashtag to sell pizza, or Kenneth Cole tying a product promotion to unrest in the Middle East. Both of these drew immediate backlash and public apologies.

Despite these failures, brands persisted in prizing speed without ensuring the necessary judgment and governance were in place as well. Being able to spot a trend has often taken precedence over deciding if that trend has a meaningful connection to the brand and its audiences.

Which Brands' Real-Time Marketing Attempts Backfired After the Oreo Tweet?

J.C. Penney, DiGiorno and Kenneth Cole all drew backlash and public apologies after forcing relevance into live moments without the judgment or governance behind Oreo's original post.

Related Article: 10 Worst Marketing Misfires and Lessons Learned

From Shared Super Bowl Moments to Millions of Private Micro-Moments

There is another critically important component here as well. While Oreo’s ability to react in real time was impressive, it also depended on being centered on a very large, very public and shared event that happens only once a year. Conversations around real-time also began to acknowledge that every customer had their own mini Super Bowls, goldmines for brands that could capitalize on them.

The reframe came from Google in 2015 with the micro-moments concept. The argument was that the decisions worth influencing were the intent-rich moments happening on individual phones — the I-want-to-know and I-want-to-buy queries — rather than the shared broadcast event. Those moments number in the millions and occur continuously.

This changed the problem. A stadium blackout happens a few times a decade. A customer reaching a purchase decision happens constantly and privately. Reacting to that volume required the capability to move out of the room and into the infrastructure, because no staffed newsroom could see or serve moments at that scale. Those private moments drove far more revenue in aggregate than any televised event.

What Year Did Google Introduce the Micro-Moments Concept?

Google reframed real-time marketing in 2015 around intent-rich micro-moments — individual, continuous decision points — rather than shared broadcast events like the Super Bowl.

How CDPs and Next-Best-Action Engines Replaced the War Room

The decade that followed was largely operational. Customer Data Platforms unified the behavioral signals. Marketing automation and triggered journeys replaced scheduled batch sends. Real-time decisioning and next-best-action engines scored the appropriate response for an individual at the point of interaction, drawing on unified data, business rules and models rather than a fixed campaign plan. Segments gave way to individual-level decisions made in the moment a behavior occurred.

While platforms matured in their capabilities and in their ability to take signals that allowed them to react, teams still struggled to keep up, whether due to content, data or prioritization issues. Adobe's 2025 research found that 75% of practitioners still consider real-time personalization a major challenge, and maintaining coherence across an enterprise is hard enough without building the infrastructure for real-time decisioning.

What Percentage of Practitioners Call Real-Time Personalization a Major Challenge?

Adobe's 2025 research found 75% of practitioners still consider real-time personalization a major challenge, even with CDPs and next-best-action engines in place.

What Agentic AI Actually Changes About Real-Time Decisions

Which gets us to today. Automated systems monitor for signals, select an action based on a set of policy parameters and execute or escalate the action when the situation breaches a predetermined threshold. This is less about production than operations. Sense something, make a judgment call within policy bounds, take action, or escalate for review.

Say you have a retailer running this sort of thing in 2026. Your loyalty member’s travel itinerary changes, creating a buying opportunity. Local weather influences the need for certain stock. A shopper drops a shopping cart mid-session. The system notices each of these things, makes a judgment call bounded by approved policy parameters, and acts or escalates the decision. That escalation point is the same judgment call 360i made at the Superdome, except it’s encoded as a permission limit rather than a person deciding to escalate and wait.

What Share of Brands Does Gartner Project Will Use Agentic AI for 1:1 Interactions by 2028?

Gartner projects 60% of brands will use agentic AI for one-to-one customer interactions by 2028, even as only 39% currently have a shared customer data platform to support that rollout.

Why Brands Still Struggle to Deliver on Real-Time Personalization

Of course, it’s one thing for brands to understand the needs and build scalable real-time personalization into their strategies, and quite another to implement it. The adoption data shows the distance between capability and practice.

2026 research from Adobe found that 62% of companies plan to use agentic AI for conversational customer engagement within 18 months, while only 39% have a shared customer data platform capable of supporting that rollout;t, yet Gartner (rather optimistically in my opinion) projects that 60% of brands will use agentic AI for one-to-one customer interactions by 2028. Yet despite all the planning and projections, a 2026 Braze report finds that only 53% of consumers say brands accurately predict their needs.

Brands have already bought the tools to react at scale. Still, the data and technology connectivity that enable them to produce relevant reactions are what are holding companies back from delivering better real-time experiences.

I would categorize some of the near-term challenges to prioritize as the following:

• Confusing personalized content for valuable information, which can lead to the “creepiness” factor rather than having customers feel seen and heard
• Confusing cultural context awareness for brand relevance, or simply being “hip” to what’s happening doesn’t equate to brand equity.
• Overindexing on real-time when forgetting about consumer messaging fatigue

And of course, there is the sheer complexity of tying everything together end-to-end, both internally and externally, from a systems and data perspective, which leaves many organizations still struggling to get even parts of this in place.

What Percentage of Consumers Say Brands Accurately Predict Their Needs?

A 2026 Braze report found only 53% of consumers believe brands accurately predict their needs, despite widespread investment in personalization tools.

Scaling Oreo's Governance Model to Every Customer, Every Channel

Oreo's result came from a team that was ready to act in a very specific context, a pre-orchestrated and delegated chain of approvals, and had decided in advance to wait for the right moment to act. Coincidentally, those are orchestration and governance functions, practiced before the industry used those terms and prioritized in-the-moment marketing.

Agentic systems raise the requirements for fulfilling this in a truly data-driven, one-to-one way rather than replacing it, because the number of moments a brand can act on is now effectively unlimited, and a large volume of ungoverned automated actions carries more risk than a single bad tweet did.

Although few (if any) brands have operationalized this at a truly omnichannel, 100% customer scale, the infrastructure is there to react meaningfully in real time, and the martech platforms and components exist to create real-time moments and personalized experiences for any customer, anywhere.

As many readers might agree, that’s easier said than done, and we’ll still be working out the kinks for a while; however, this is not easy even for the most sophisticated teams. Yet the thought process that set up a Super Bowl war room and an operational infrastructure that accepted near-real-time content input laid the foundation for much of what brands are still striving for today.

Learn how you can join our contributor community.