A.CRE
, ,

Episode 9 of Multipliers: The Traffic Engine Nobody’s Building

Most CRE operators spend a fixed percentage of every reservation paying a third-party platform to exist. They know it is happening. They accept it as the cost of doing business. What very few of them have asked is: what would it take to build a direct pipeline that makes those fees optional rather than mandatory?

Michael Belasco asked that question before Olympic RV Park opened. Less than a year later, the park is beating its projections, and a significant part of why comes down to a traffic strategy that most operators in the outdoor hospitality space are not running. It is not magic. It is geotargeting, retargeting, AI-generated content, an attribution platform Michael built himself, and the discipline to connect those pieces into something that actually works together. The result is a moat built on NOI that compounds as long as the strategy runs.

The second half of this episode covers something equally practical: how to actually learn AI in a way that sticks. Spencer’s answer is his teenage daughter, who is now outpacing him at Roblox game development using Claude, because she started with something she genuinely cared about and let the learning follow. The parallel for CRE professionals is direct and actionable.

In this episode of the Multipliers podcast, Spencer Burton and Sam Carlson break down what Michael has built at Olympic, why it works, and why the AI arbitrage it represents is available to any CRE operator willing to tinker, before the window closes.

Episode 9 of Multipliers: The Traffic Engine Nobody’s Building

Michael Belasco is absent this week. Spencer and Sam are on their own, which gives them license to talk about him freely, and they take it. The episode opens with Sam wearing an Olympic RV Park shirt, which becomes the entry point for a detailed breakdown of what Michael has built there and why it is working. Michael is approaching his one-year mark and beating 2026 projections, which Sam describes as genuinely surprising, not because he doubted Michael, but because the model he put together was conservative and reality has outpaced it. Spencer is flying to Portland after this recording for a ULI council session where he will spend ninety minutes teaching a room of real estate professionals to build a live database and connect AI to it. The episode connects back to Episode 8, where the group reflected on the non-obvious multipliers that compound over a career. This one is more tactical: a live case study in what AI-powered real estate operations actually looks like when someone does the work.

Why This Episode, Why Now

The timing of this episode is not accidental. The outdoor hospitality and transient real estate space (RV parks, self-storage, short-term rentals, market-rate multifamily) is structurally dependent on third-party distribution platforms in the same way that hotels depend on Expedia or vacation rentals depend on Airbnb. Those platforms extract a cut of every reservation, typically somewhere between five and twenty percent depending on the service, and most operators treat that drag as fixed. It is not fixed. It is just how most people have chosen to compete.

What Michael has done at Olympic is carve out a meaningful share of his reservations through channels he controls: organic search traffic, direct-to-website ad campaigns, and a retargeting infrastructure that keeps his park visible across the customer journey in a way that most operators are not doing. Sam, whose professional background is in digital marketing and traffic strategy, describes what Michael has built as genuinely unusual for the asset class. Not because the tactics are secret (they are not), but because most operators default to the same playbook and never explore what is possible with a bit of effort and the right tools.

The connection to AI is direct. Michael’s attribution platform, the piece that tells him what is actually working and in what sequence, was built through tinkering. His AI-generated event calendar, which drives organic traffic from people who are not yet thinking about RV parks, was built the same way. These are not enterprise software solutions. They are things a motivated person built by trying, iterating, and paying attention to what the data said.

Spencer’s conversation with a UNC student, where a hiring manager asking how young people can bring AI into an organization, and the young person not knowing how to answer, sets up the second half. The answer Spencer gives is the same one he would give to any CRE professional sitting on the sidelines: tinker. Pick something you care about, use AI to do it, and let the learning follow the doing. That is exactly what his daughter is doing with Roblox. It is also exactly what Michael did with Olympic.

Episode Highlights

Here are the themes that stood out.

1. What Michael Built at Olympic: Why It’s Working

The outdoor hospitality business model is structurally similar to hotels in one important way: distribution is dominated by third-party platforms that control a significant share of where guests discover and book properties. For RV parks, these are not Expedia-level brands; they are niche platforms that RV enthusiasts use to find parks, and they they operate on the same economics. You play their game, you pay their cut, and your NOI is reduced accordingly.

Spencer explained the financial logic clearly: if third-party platforms are taking five to twenty percent of every reservation, and NOI is one of the two variables that drives property value, then any strategy that reduces that commission load has a direct and compounding effect on value. Michael identified this early and decided to build a parallel distribution channel he controls.

Sam’s read on what Michael has actually built is worth taking seriously, because Sam thinks about traffic and conversion for a living. The combination of organic search strategy and targeted paid campaigns driving directly to the park’s own booking platform means Michael is acquiring guests at a cost lower than what he would pay in third-party commissions for the same reservation. Run that math across a full season and the impact on NOI is meaningful. Run it across a portfolio of parks, which is where Michael is heading, and it becomes a genuine competitive moat.

Spencer’s broader point is one that CRE professionals across asset types should hear: most operators underestimate how much value is available through direct distribution, because the path of least resistance is to do what everyone else does. Self-storage operators, senior housing operators, market-rate multifamily operators: all of them have some version of the same dynamic. The third-party platforms are not going away, and you still need to play that game. But building a direct pipeline alongside it changes the economics in your favor.

2. Geofencing vs. Geotargeting: One Works and One Doesn’t

Sam used the Olympic example to walk through a distinction that comes up constantly in CRE marketing conversations and is almost always misunderstood: the difference between geofencing and geotargeting.

Geofencing is the version that sounds impressive in a sales pitch. The idea is that your phone gets tagged when you enter a physical location, and ads start appearing on whatever apps you are using. Sam was direct about the reality: geofencing ads convert terribly. They show up in low-attention environments like games and background apps, and the connection between seeing an ad while playing Angry Birds and booking an RV site is just not there. The technology works. The conversion does not.

Geotargeting is different. You draw a radius on a map, say fifty miles around Port Angeles, Washington, and you serve ads to the right people on social media platforms within that area, where they are actually paying attention. Someone scrolling Instagram and seeing a well-placed ad for an RV park near a destination they are thinking about visiting is a very different interaction than an interrupt in a mobile game. The attention quality is higher, the intent signals are clearer, and the conversion rate reflects that.

What makes Michael’s version of this particularly effective is what happens after the first touch. Social media platforms have display networks that follow users to other sites after they have engaged with an ad. A potential guest might see the park on Instagram, then see it again on a news site, then encounter it once more while browsing something unrelated. By the seventh or eighth touch, they are ready to book, and because Michael has built the retargeting infrastructure correctly, that journey is not accidental. It is engineered.

3. Attribution: The Holy Grail of Digital Marketing in CRE

Knowing how to drive traffic is only half the problem. The harder half, the piece that most CRE operators never solve, is attribution: understanding which of your marketing activities is actually driving reservations, and how they interact with each other to move a guest through the decision process.

Sam called it the holy grail of digital marketing, and the reason it is hard is that it is never just one thing. A guest who books might have found the park through an organic search, been retargeted through the display network, and then converted on the seventh ad impression after their spouse mentioned wanting to go camping. Attributing that booking to any single channel misses the reality of how the customer journey actually works. Most CRE operators, faced with this complexity, either give up on measurement entirely or over-attribute to the last click and miss everything that contributed upstream.

What Michael has built, through tinkering rather than purchasing an enterprise solution, is an attribution platform that gives him enough clarity to actually run the strategy. Sam described the moment when the attribution picture became clear as a turning point: once you know what is working and in what sequence, the whole system becomes actionable. You know where to put the next dollar. You know which channel is contributing to which conversion. You can optimize with something like precision rather than guesswork.

The vending machine analogy Sam used is apt: if you can reliably put one dollar in and get three or four out, you want to put in as many dollars as possible. The bottleneck for most operators is not access to the channels. It is not knowing which dollars are working. Michael solved that. And it took tinkering, not an expensive platform.

4. The Event Calendar Play: Organic Traffic Plus Retargeting

One of the more interesting specific tactics from this episode is the event calendar. The Olympic site maintains a calendar of local events: festivals, outdoor activities, things happening in the area around the park. Those pages generate organic search traffic from people who are not necessarily thinking about RV parks but are interested in what is happening near Port Angeles. Spencer described many of these event listings as automatically generated through the AI infrastructure Michael has built, which means the content machine runs without requiring manual effort for every post.

Here is why this matters for retargeting: once someone visits any page on the site, even an event calendar page for a strawberry festival, they can be retargeted with park-focused ads. They signaled interest in the area by showing up. The park can now follow them through the display network with messaging relevant to their apparent travel interest. Most of them are not going to convert immediately. Some portion will, over time, especially as the retargeting sequence builds familiarity.

Spencer and Sam were careful to note that none of this is particularly novel in the world of digital marketing. These are standard tactics. What makes them unusual in the CRE context is that almost nobody in the asset class is running them. The gap between what is possible and what most operators do is not a technology gap. It is an attention gap. Most operators are not thinking about their website as a traffic and retargeting asset; they are thinking about it as a listing, and changing that frame changes what is possible.

The amenities content plays the same role. Pages about nearby rivers, hiking trails, and outdoor activities draw in visitors who are actively planning trips in the area. Each one of those visitors can be retargeted. None of this requires a marketing department. It requires someone who understands the logic and has built the infrastructure to execute it.

5. Tinker With Something You Love

Spencer was heading into this episode determined not to talk about AI. He ended up spending most of the second half on it, because a conversation he had the day before with a UNC student made the connection unavoidable.

The student described a pattern she was seeing across job interviews: hiring managers at CRE firms asking candidates how they would bring AI into the organization. Most of the organizations asking that question, she said, had already admitted they did not know either. They were looking for young people to help them figure it out. Her problem was that she was in the same position. She did not know how to answer the question.

Spencer’s answer was the same one he gives consistently: tinker. But he was honest that tinkering feels like a chore when it is detached from something you actually care about. His daughter is the counter-example. She loves Roblox, a game-building platform where all the games are created by community members. She had always been good at the design side but could not write the scripts to make her worlds functional. Spencer spent a weekend with her connecting Claude to the Roblox coding environment via MCP and handling the scripting while she handled the design. By the third weekend, she was doing the scripting herself. By the time they recorded this episode, she had built a complete gun mechanics system with recoil logic, corner-lean animations, and distance-based accuracy modeling, and she had outgrown her dad entirely.

The reason it worked, Spencer argued, is that she never thought of it as learning AI. She thought of it as building her game. The AI was the tool. The game was the point. That is the advice he gave the student: find the thing you are already trying to do and use AI to do it. The proficiency will follow. At some point you have to connect the skills to a professional context, but the fastest path to fluency is genuine motivation, not obligation.

6. The Two-AI Workflow: Prompt Engineer and Builder

Spencer’s practical AI workflow deserves its own section because it is one of the clearest descriptions of how to use multiple AI tools effectively without wasting money or producing suboptimal output.

The core insight is that different AI systems are tuned for different jobs, and mixing those jobs inside a single context window produces worse results than keeping them separate. Spencer runs two parallel threads when building something. The first is a prompt engineering thread, typically Claude in chat mode, whose only job is to think through what he wants to build, ask clarifying questions, and produce a clean, well-structured set of instructions. The second is a building thread, either Replit or Claude Code, whose only job is to execute those instructions and produce the output.

The reason this matters is context window pollution. If you brainstorm, iterate, and build all in the same conversation, you are forcing the coding model to carry the weight of all the exploratory thinking that preceded the actual task. That creates confusion and degrades output. By keeping the brainstorming in one thread and the building in another, you give the builder a clean, precise brief and get a much better result.

Sam uses a similar approach with Claude and Manus, making the same basic separation between thinking and building. He also noted the cost dimension: Claude in chat mode is cheaper per token than most agentic builders, so using it as the prompt engineer and only spinning up the more expensive builder once the brief is fully worked out saves money and improves quality at the same time. Spencer’s ULI Nashville session, where he planned to build a live database with a room full of real estate professionals in ninety minutes, would use Replit as the builder specifically because Replit handles deployment automatically, which lets an audience see a live, hosted result at the end of the session without a separate deployment step.

7. The AI Arbitrage Window Is Open, For Now

Sam closed the episode with a point that pulls everything together. What Michael has done at Olympic, using AI to build a direct traffic engine, an attribution platform, and a content machine that runs autonomously, is not the standard playbook for RV park operations. It is the playbook of someone who saw an arbitrage opportunity and moved on it before the rest of the market caught up.

That arbitrage exists because most CRE operators are still doing things the way they have always been done. They are on the third-party platforms because everyone is on the third-party platforms. They are not building direct traffic because it requires effort and knowledge that most people in the asset class have not developed yet. The gap between what AI makes possible and what most operators are actually doing is wide enough right now to create a real competitive advantage for the people who close it.

The honest caveat is that this window will not stay open indefinitely. As more operators figure out what Michael has figured out, the tactics become table stakes rather than differentiators. The people who build this capability now will have a head start in experience, data, and infrastructure that is hard to replicate quickly. The people who wait will find the playbook more crowded and the returns more compressed.

Sam’s framing was direct: Michael is probably among the first real estate operators to capitalize on this arbitrage at this level. That is meaningful. The tools available through communities like AI.Edge exist precisely to help CRE professionals close the gap: understanding what is possible, learning the tools, and building the operational capabilities that turn AI from a curiosity into a competitive moat.

The Bigger Idea

The two halves of this episode are really one argument. Michael’s traffic engine at Olympic is what you get when someone with domain knowledge in real estate, curiosity about digital marketing, and willingness to tinker figures out what AI makes possible in their specific context. Spencer’s daughter’s Roblox game is what you get when someone with genuine motivation and access to the right tools figures out what AI makes possible in their specific context. The asset class is different. The mechanism is identical.

The conventional assumption in CRE is that digital marketing is something you hire out: you engage a firm, you pay the retainer, you play the platforms the way everyone plays them. What Michael demonstrated is that an operator who understands the underlying logic of direct traffic, attribution, and retargeting can build something more effective than what most hired agencies produce, at a fraction of the cost, with tools that are now genuinely accessible to anyone willing to learn them. That is a meaningful shift.

The attribution piece is the one that gets underappreciated. Sam has been saying for years that attribution is the holy grail of digital marketing, and he means it: without knowing what is working, you are spending blind. The reason most CRE operators have never solved it is that it is genuinely hard to do manually. AI makes it tractable. Michael tinkered his way to a solution that gave him the clarity to actually run the strategy. That clarity is now a competitive advantage that compounds with every season of data.

Spencer’s student question is the version of this problem that most young professionals are facing: how do I answer the hiring manager who wants to know how I will bring AI into their organization? The honest answer, and the one Spencer gave, is that you cannot answer it convincingly until you have actually done it. The path is not to study AI abstractly. It is to use it for something specific, build something real, and let the answer emerge from the experience. His daughter did not study game development. She built a game.

The AI arbitrage in commercial real estate is real and it is available right now: in outdoor hospitality, in self-storage, in market-rate multifamily, in any asset class where distribution is currently controlled by third parties and where direct traffic would improve NOI. The operators who figure this out first will have something that is genuinely hard to replicate quickly. CRE Agents was built for exactly this kind of work, bringing AI capability into real estate operations in a way that produces real outputs. The window is open. The question is whether you move while it is.

Frequently Asked Questions about Episode 9 of Multipliers: The Traffic Engine Nobody’s Building

Michael built a direct traffic engine that drives reservations through channels he controls: organic search, targeted paid campaigns, and a retargeting infrastructure, rather than relying solely on third-party booking platforms that take a commission on every reservation. Sam, whose background is in digital marketing, described it as an unusually sophisticated strategy for the outdoor hospitality asset class. The result is that Michael is acquiring guests at a cost lower than what he would pay in third-party commissions, which improves his NOI and compounds over time as the strategy matures.

Geofencing tags a device when it enters a physical location and serves ads on background apps, a strategy that sounds compelling but converts poorly because the attention quality is low. Geotargeting uses social media platforms to serve ads to specific audiences within a defined geographic radius, where people are actively engaged and the intent signals are much stronger. Sam was direct: geofencing is the word that sounds impressive in sales pitches and underperforms in practice. Geotargeting, combined with display network retargeting, is the strategy that actually moves the needle.

Retargeting allows you to continue showing ads to someone after they have visited your website, following them across other sites and apps through display networks. Michael uses it to stay visible to potential guests who showed interest by visiting any page on the Olympic site, including event calendar pages, but did not book immediately. Over a sequence of seven or eight touchpoints, a meaningful portion of those visitors eventually convert. The key is having a reason for people to visit the site in the first place, which is where the organic content strategy, including event calendars, amenity pages, and local area guides, feeds the retargeting funnel.

Without attribution, knowing which marketing activities drove which reservations and how they worked together, and you are spending without knowing what is working. Sam called it the holy grail of digital marketing because it is genuinely hard to solve: a guest who books might have come through organic search, been retargeted three times, and converted after a social media ad. Attributing that booking to just one channel misses the whole picture. Michael built an attribution platform through tinkering that gave him enough clarity to actually optimize the strategy, and that clarity is what allows him to run the vending machine: put a dollar in, get three or four out.

Spencer made the argument explicitly: self-storage, senior housing, market-rate multifamily, and any other asset class where third-party platforms control a share of distribution faces the same economics. If you can reduce the percentage of reservations or leases coming through third-party channels, and acquire those tenants or guests more cheaply through direct channels, you improve NOI and build a moat that compounds. The tactics differ by asset class, but the underlying logic is the same. Most operators are not doing this because they have not prioritized it, not because it is impossible.

Spencer was responding to a UNC student who did not know how to answer hiring managers asking how she would bring AI into their organization. His advice: do not try to learn AI abstractly. Find something you already care about and use AI to do it. His daughter loves building in Roblox and has become genuinely proficient with Claude by using it to build her game, not because she was trying to learn AI, but because AI was the tool she needed to do the thing she wanted to do. The proficiency follows the motivation. At some point you need to connect the skills to a professional context, but the fastest path to fluency is doing something real.

Spencer runs two parallel threads when building something: a prompt engineering thread and a building thread. The prompt engineer, typically Claude in chat mode, does the exploratory thinking, asks clarifying questions, and produces a clean set of instructions. The builder, either Replit or Claude Code, takes those instructions and executes them. The reason for keeping them separate is context window quality: mixing brainstorming and building in the same thread forces the coding model to carry all the exploratory noise, which degrades output. Using Claude for prompt engineering is also cheaper per token than most agentic builders, so the workflow improves both quality and cost.

Claude Code builds locally: whatever it creates lives on your computer and requires additional steps to deploy and share. Replit builds in the cloud and includes deployment automatically, so hitting publish makes the result immediately accessible to anyone. Spencer uses Replit when he needs to deploy quickly or show a live result to an audience, as he planned for the ULI Nashville session. He uses Claude Code for CRE Agents deployments where the added flexibility and control of a local build is worth the extra setup. For most people new to building with AI, Replit is the more accessible starting point.

Sam described it clearly: there is a gap right now between what AI makes possible for CRE operators and what most operators are actually doing. The operators who close that gap first by building direct traffic engines, attribution platforms, AI-generated content strategies, and operational automation, will have a competitive advantage that compounds over time and is hard to replicate quickly. Michael is probably among the first RV park operators to have done this at this level. The window will not stay open indefinitely as these practices spread, but the advantage for early movers is real.

Pick something specific you are already trying to do and use AI to do it. Do not try to learn AI in the abstract, as that approach produces the student who can describe the technology but cannot answer the interview question about how to actually use it. Michael learned by building his traffic engine. Spencer’s daughter learned by building her game. The proficiency comes from doing something real, not from studying the tools. Resources like AI.Edge and platforms like CRE Agents exist to help CRE professionals move from curiosity to capability, but the move still has to start somewhere specific.

by