The Front Lines
SubscribeContent Studio
Omar Tawakol.
CEO · Rembrand
Guest
Omar Tawakol
CEO
Company:
Rembrand
Location:
Los Altos, California, United States
Loading episode...
Listen onApple PodcastsSpotify

The $1M Revenue Trap: What Rembrand's Omar Tawakol Learned About False Product-Market Fit

Omar Tawakol has sold multiple companies. He knows what traction looks like. So when Rembrand had repeat purchases from brands across three countries, multiple campaigns, and over a million dollars in revenue, he didn't celebrate — he started asking harder questions.

"Every single signal of product market fit was there," he said on a recent episode of BUILDERS. "And I thought I had product market fit for the way we were doing it. And I was still wrong."

What he found when he dug in is one of the more useful — and underreported — failure modes in early-stage B2B: revenue that looks like PMF but is actually founder-dependent. And it's almost impossible to see from inside the deal.

The Insight That Started Everything

Omar has been in digital advertising since its earliest days. He's sat in the rooms where hundreds of billions in ad spend gets allocated. And for years, he kept noticing the same thing — the executives making those decisions were going home and paying to avoid their own product.

"You are the guys who are in the room who allocate hundreds of billions of spend on advertising and then you go home and not watch ads," he said.

That contradiction became Rembrand's founding logic. Every advertising format that succeeded did so by embedding the brand inside the content. Search moved from banners to sponsored listings. Facebook solved its monetization problem by making ads native to the social feed. TikTok followed the same pattern. Video now accounts for 55% of global content consumption time, and it had no native equivalent. Rembrand was built to be that solution: AI that inserts brand products — a Pepsi can, an Audible billboard, a Cadillac that wasn't in the original shot — directly into existing video scenes. The quality bar is unambiguous: "If you ever realize that we did it, virtually we have failed."

Why the Category Name Determined the Buyer

Getting the product to work was only half the problem. Knowing what to call it cost them more time than Omar expected.

Rembrand spent years as a "virtual product placement" company. Technically accurate. Operationally, it was routing them to the wrong buyer and the wrong budget. Product placement is bespoke and negotiated — no standardization on supply, demand, measurement, or purchase mechanics. Media operates at scale on entirely different infrastructure: standardized specs, programmatic pipelines, Nielsen and Kantar measurement, video budgets with defined annual line items.

"You have to have standardization on supply, standardization on demand, standardization on measurement, standardization on how you buy," Omar said. "None of that happened with product placement. So we realized it was the wrong analogy."

Moving to "in-content advertising" wasn't a semantic shift. It changed who picked up the phone, which internal team owned the budget, and how the buy was measured and renewed. The name is a routing mechanism, not a description.

What the $1M in Renewals Was Actually Measuring

Back to the revenue that wasn't PMF.

Omar's original signal for fit was simple: repeat purchase rate. If customers kept coming back, the product was working. "Two years ago I thought it was very simple. How much repeat purchase do I have signaling that they're happy? And I thought that is all I need to look at."

The problem: he hadn't accounted for the founder variable. "I was wrong because I didn't calculate that I'm not a first time founder and that my co founders are also not first time founders. So you get a little bit of allowance from your clients in those situations."

Experienced founders carry relationship-based credit that their product hasn't earned independently. Clients were buying Omar's track record, not a repeatable motion. The actual PMF signal he needed was different in kind, not just degree: "What I really needed was very clear metrics on their side on how they plan to make these buys even when we're not in the room. Like what line item are they planning for to fit this budget for every year." The question isn't whether they renew. It's whether they've built you into their annual planning without a founder-driven conversation to get there.

Repeatability Over the Impressive Demo

Running alongside the PMF discovery was a separate reckoning on product strategy. Rembrand's early executions leaned into what the technology could do — highly customized, AI-enabled placements that demonstrated range and capability. They won deals. None of it scaled.

The correction came from watching operators in Asia, Africa, and the Middle East who were successfully running in-content advertising at volume. The difference was deliberate: they weren't doing the impressive work. "They started to scale this stuff precisely because they weren't doing the sexy things, they were doing the repeatable things." Standardized placements, consistent specs, high volume. The bespoke execution wins the first deal and impresses in the pitch. Repeatability builds the category.

Why Rembrand Stopped Building Its Own AI Infrastructure

The third major inflection was technical — and it started with an architectural constraint that predated the current foundation model era.

Early on, Rembrand had to build its core AI capabilities from scratch. The specific problem — inserting objects into high-quality video scenes with moving cameras and moving people, with physically accurate occlusion, color grading, and lighting — required a three-dimensional spatial understanding of the scene that diffusion models weren't designed to provide. "None of the diffusion models could do the things we needed because we needed physical understanding of the space you're in closer to what you're seeing now with what they call world models." So they built the underlying algorithms themselves.

As foundation model training costs scaled into the billions, that strategy stopped making sense. Rembrand wasn't spending billions on training, and the revenue side wasn't yet generating billions to justify it. The pivot: stop competing at the infrastructure layer, and build around proprietary data instead. Ten years of unique video and placement data became the asset — used to fine-tune open source models that other companies had spent billions developing. "We have 10 years of data that's unique and we take that data, we do the fine tuning and the retraining of those models and now we have a defensible moat that allows us not to have to spend billions on the training."

The business model reinforces the strategy. Rembrand's B2B use case is profitable in multiple markets — which means the revenue funds ongoing model training. Omar draws the contrast directly: "If you look at Sora, OpenAI, they had to close down Sora because it was burning a million dollar a day. And they didn't have the business to fund it." A data moat plus a profitable B2B motion is a different kind of defensibility than foundation model scale.

Rembrand is now live daily — TV shows, films, distributed across multiple countries, with some of the largest publishers in the world. The category is beginning to function like one.

The full conversation with Omar Tawakol is available now on BUILDERS.

Six takeaways from this conversation.

Actionable for Sales & Marketing Tech Builders founders

  1. Category naming is a buyer routing decision, not a branding exercise
    Omar spent years calling Rembrand "virtual product placement" before realizing the label was sending him to the wrong room. Product placement is a bespoke, negotiated, content-owner-driven transaction — no standardization on supply, demand, measurement, or purchase mechanics. In-content advertising plugs into existing media buying infrastructure: video budgets, Nielsen/Kantar measurement, programmatic pipelines. The name change wasn't semantic — it changed who picked up the phone and which budget got unlocked. Founders building new categories should define the name by where it routes the buyer's mental model, not by what the technology does.
  2. Repeat revenue can mask a founder-dependent business
    Rembrand had multi-country repeat purchases across multiple campaigns, over $1M, every signal pointing to product-market fit. Omar concluded he was wrong. The reason: experienced founders get relationship-based allowance from early clients that first-time founders don't. Customers were buying Omar and his co-founders, not a repeatable product motion. True fit, in his definition, means buyers have a named budget line item, clear measurement criteria, and a plan to allocate spend to that line item annually — without Rembrand in the room to shepherd the deal. The pressure test isn't renewal rate. It's whether the deal happens when you're not there.
  3. Innovation budgets are a trap dressed up as traction
    When a buyer slots you into an innovation or experimentation budget, you are outside the line items that renew automatically. Omar's approach: accept those deals selectively, but immediately run the measurement studies needed to qualify for the buyer's repeatable video or CTV budget — and fund those studies yourself if necessary. The goal isn't the current deal. It's getting onto the annual planning spreadsheet. Until your product has a line item in the buyer's budget template, you don't have a scalable revenue motion, you have a series of one-off conversations.
  4. Scalability beats the impressive demo — every time
    In Rembrand's early years, AI-enabled bespoke placements — think animated product integrations, highly customized scene work — won deals and impressed clients. None of it scaled. The shift came from watching operators in Asia and Africa who were doing the boring, standardized, high-volume version of the same thing precisely because they weren't chasing the impressive execution. Omar's rule: if you can't describe the exact repeatable motion, you have a proof of concept, not a business. Bright lights get you the first deal. Standardization gets you the category.
  5. In applied AI, the moat is data and workflow — not the model
    Rembrand originally built proprietary AI infrastructure from scratch to solve the video insertion problem. As foundation model spending scaled to billions, the math stopped working. The pivot: take open source models, and fine-tune them on ten years of proprietary video and placement data that no competitor can replicate. New models keep shipping — Rembrand takes them, runs them through the data moat, and ships a defensible output without billion-dollar training costs. The business model matters too: a profitable B2B use case funds ongoing training. Sora burned a million dollars a day with no B2B revenue to support it. That's the contrast Omar is building against.
  6. Don't optimize for acquisition — optimize for value
    At BlueKai, Omar pulled an executive aside when he noticed the exec's eyes lighting up at the mention of a potential acquirer. His instruction: stop. Building a company toward an exit changes the decisions you make daily in ways that quietly erode the value you're trying to sell. Omar's framework is simple — stay focused on building real value, don't manufacture attractiveness for a hypothetical buyer, and if a serious acquirer shows up, you haven't done anything to make yourself less valuable. The exits come when value is undeniable, not when you've been managing optics.

More episodes

Johannes Hoech
CMO · Sift
Nis Frome
Marketing Consultant
Kamron Kunce
Rod Hess