--- title: "Unite's mission and handbook" description: "Unite's thesis, story and mission, in the handbook the founder publishes in the open, one chapter at a time." canonical: https://unitebr.com/en/mission last_updated: 2026-10-16 --- # Unite's mission and handbook The handbook 1. Why Unite exists 2. How we got here 3. Who we build for 4. How we decide what to build 5. How we make money 6. How the team works 7. What we value 8. Where we're going We publish one chapter at a time, in the open. # Why Unite exists Weverton Guedes · Founder and CEO · São Paulo, August 2026 ## Our mission Build the infrastructure for creative science. In practice, building this infrastructure means making every creative decision contribute more to business results. More than a trillion dollars goes into advertising every year, and much of the return on that money depends on choosing what to produce and what to put live. Unite was founded to bring data intelligence, science and method to that moment, increasing the impact of every decision about budget, production and brand exposure. ## Why this is the mission Advertising uses far less science than it could. We have the Ehrenberg-Bass empirical generalizations, decades of IPA Databank analysis, norm databases from Kantar, System1 and Ipsos, plus the massive tests run by Google and Meta. Not to mention copy testing, which has been around since the 1920s. The problem is that this knowledge arrives in the wrong format at the wrong time. A category average doesn't help you decide which ad should go live tomorrow. By the time the report shows up, weeks later, the team has already shipped a pile of other things, the vendor keeps the knowledge and all the brand gets is a final slide deck. Even when the evidence exists, it almost never makes it into the daily routine. There's no time to look for it, no access to the history and no incentive to revisit a decision while it can still be changed. General rules help day to day, but they don't help you choose between the options the team actually has in hand. For that, you have to analyze your own ads and use the results of each campaign to improve the next one. Without that loop, the team follows best practices and still doesn't know what really works for the brand. A best practice saved in a PDF on Drive loses to a deliverable on a tight deadline. Formal testing is expensive and takes too long, so it almost always goes to the big brand film and not to performance ads, which get most of the direct attribution for results. Incentives get in the way too. The market pays for delivery and rewards what's new. Nobody gets promoted for avoiding a mistake the company keeps making. And nothing compounds. The brand runs research, learns, forgets and, some time later, launches something similar to what it had already ruled out. Medicine went through something similar. The literature already existed when evidence-based medicine took off in the '90s. What changed was using that evidence at the point of care: the doctor still makes the call, but now with real information. That's what I mean by creative science. Making a better decision at the moment of the ad, learning from the result and applying it to the next round. ## Why now? Advantage+ (Meta Ads) and Performance Max (Google Ads) have automated a large part of media buying. Generative models have slashed the cost of making variations, and today you can produce five hundred ads in an afternoon (quality is still a challenge, but that will get better). The bottleneck has moved from the ability to produce to the choice. Someone has to decide which ads deserve to exist. For many brands, media has also gotten more expensive. More advertisers fighting for the same attention, auctions pushing prices up and new costs showing up on the bill, taxes included. Getting the creative wrong costs a lot more today than it used to. Teams are slow to notice. The creative wears out, frequency climbs, CTR drops and CPA shoots up. By the time the signal shows up on the dashboard, two weeks of budget have already gone to an ad the audience is sick of seeing. With volume this high, that mistake becomes routine. ## Where the money gets lost Every week the team picks what goes live and commits budget, production and brand exposure on very little evidence. The bad ad shows up on the monthly bill. The better alternative, which nobody produced, doesn't show up anywhere. Before launch, a weak choice burns production hours and already puts budget at risk. Producing more variations increases the number of attempts (but doesn't change the hit rate of each one: even if volume in theory helps you find winning creatives, it has to start from good hypotheses). After launch, the fatigue read has to arrive in time to swap the ad in the same round. Analyzing it two weeks later solves nothing. Between rounds, hours go to exporting, renaming and cross-referencing spreadsheets. It's operational work tying up people who should be thinking critically about insights, culture, the brand and creative hypotheses. Business results don't depend on creative alone, and ad manager data doesn't prove what caused them. Even so, choosing better, reacting sooner and avoiding repetition changes how much money gets lost along the way. Part of the gain we want to deliver for brands comes from recovering the huge amount of money being lost today. The value of the infrastructure shows up when all of these choices get better together. ## The loop between decision, production and media Validate comes in before the spend. It helps the team compare ads, assess risk and decide what deserves investment (and it isn't limited to digital: today it already pre-tests OOH too). Optimize tracks what ran on channels like Meta, TikTok and Google, links each ad's attributes to performance, monitors fatigue and turns results into learning for the next decision. Create takes that learning back into production, at the pace and volume the operation needs. The central surface records the hypotheses, the brand knowledge, the losses and the wins, and that's how the team separates what it actually predicted from the explanation people tend to make up in the final report. ## What makes this loop possible ### Instrumentation The ad manager gives you an ID and a metric. Hook, product, pacing, persuasion mechanism, narrative and so on are left out (exactly what someone needs to decide what to produce). Without these attributes, the team knows how much the ad spent but can't turn that into a creative decision. Instrumentation links ad, attribute, hypothesis, delivery and result. And it records why the ad was chosen before it goes live. This matters because account data is observational. The platform pushes more budget to whatever is already performing, until it saturates (which is happening faster and faster, both because of how the algorithms work and because of culture and consumer behavior). The analysis has to separate what was observed from what was tested. It also takes the right analysis methodology. If two ads were delivered to different audiences, the difference in results between them isn't down to the creative. No model fixes a badly built comparison. ### Brand memory Today, information about the creative work is scattered across folders, emails and decks. The ad sits in one place, the result in another, and the reason behind the choice almost always gets lost. With this memory, the next brief starts from what happened, without depending on someone remembering, exporting everything again or rebuilding the story. A lot of a brand's knowledge today lives in people's heads, and when someone leaves, they take that learning with them. The history has to belong to the brand, not to whoever passed through it. That's how a company avoids rework and doesn't pay twice for the same thing. ### Creative systems Creative intelligence is only worth something when it reaches execution. What the brand has learned has to make it into the brief, guide production, go through approval and show up in the ad that goes live. When that path depends on copying information from one tool to another, the learning is obviously going to die. Unite Flows speeds this up, automating what's repetitive and keeping context across creation, approval and media. With that, the team produces faster and can apply what the brand has learned, at the volume the operation demands. ### The creative intelligence engine The memory is exclusive to each brand. Separately, our engine identifies patterns in broader data that brands have authorized us to use. A brand rarely lands many hits in a year. With a hit rate between 5% and 8%, you can't form a reliable pattern by looking only at one account's history. The engine also helps at the start. Customers come in with what already works in the market and build their own history on top of it. A pattern only helps when context is taken into account. Category, region, language, platform and seasonality, plus countless other factors, will change the answer, which on top of everything isn't static: it changes over time. ## The human role The tool organizes the evidence and makes the history searchable. Whoever is accountable for the choice still makes the decision. Taste, cultural timing and the courage to try new things don't come out of a math formula. In regulated industries, the choice also carries legal and brand responsibility. Automation doesn't take that away from anyone. ## On throwing everything into a generic model Plenty of gurus say all you need is to connect a model to your Drive, your spreadsheets and your ad manager to get intelligence and results. They forget that the model only uses what you give it, and the decisions that really make a difference depend on much more than that. Not even the most capable model can recover the normalized modeling and the curated, proprietary context it takes to deliver high quality if nobody makes them available. It also doesn't correct for media bias, audience shifts or creative fatigue, and a one-off answer in ChatGPT or Claude doesn't become institutional memory. The learning has to reach the brief without depending on someone asking the same question again. Without that foundation, the model finds incomplete files and scattered answers, and no history ready to guide the decisions that matter. ## On Meta closing its own loop Meta wants to close generation, audience, delivery and measurement inside its own platform. Advantage+ is already the default campaign setup. Meta can distribute and even produce at scale, but what goes out into the world is the brand's call. The brand is the only one with the full context that decision requires. ### Why measurement has to be independent Large advertisers run incrementality tests because they don't rely only on what the platform itself reports. Meta measures what it attributes to itself and doesn't compare its creative with another channel. That read has to come from outside. What Meta learns improves only its own system. The brand can't look into that archive, and the party holding the data is the same one running the auction. The critical data on how people respond to the brand's creatives running on Meta's inventory will never be made available to the brand. Again, there's no incentive to share it. For a small single-channel retailer, closing the loop inside the platform may be enough. For a large brand, history and testing have to cut across channels; they need a platform-agnostic layer. ## What extends the impact ### 01 Creative quality in MMM MMM splits budget across channels using aggregated data and a long history. In other words, it handles portfolio and budget management, and creative decisions are left out. Creative quality still enters these attribution models as a residual. So I'm betting on the synergy between our creative infrastructure and the MMM players, to improve this kind of read and bring our work closer to business results. For creative to become a useful variable in MMM, you need a tagged archive, valid comparisons and time series. Down the line, we want to deliver the creative variable and integrate with the companies that already measure channel, budget and incrementality. ### 02 Agents after the foundation An agent can only propose a creative round when ad, delivery, result and reason are linked. Before that, it generates generic variations. Today Validate comes in after production. The next step is to reach the brief, before that money is spent. ### 03 The whole brand and other media We also want to look at distinctive assets, consistency and wear-out over the years, across campaigns and channels. We started with digital because that's where there's data for every single ad. Today Optimize reads Meta, TikTok and Google, Validate pre-tests OOH and DOOH too, and Create produces for any media. Other media like retail media and CTV are also on our radar. The same discipline will come with that expansion. ## Brazil first, then global We started in Brazil because large advertisers here produce at volume, work with agencies and have brand and performance teams judging creative in different ways, with no method, and with all the challenges I've already described. Since budgets here are smaller, the product has to squeeze the most gain out of every front we work on. If we deliver lift at scale here, we'll know for sure we're ready for bigger markets. The history belongs to the brand in any country, and we're already set up for other markets. ## The missing profession Between media, data and creative there's a job nobody fully owns: bringing those worlds together (the clash between media/BI and creative is a classic). The Creative Strategist essentially fills that space. They own the hypothesis, the brief, the test and the learning. The role already exists in other markets, but it barely exists in Brazil, especially at large brands (it's still very much a DTC and DNVB thing). We're betting big on this specialization as the missing link to get creative intelligence executed in practice, from inside brands' operations. We want to help establish this profession. ### Self-interest A market with trained strategists is a better buyer of what we build (I'd rather admit that than pretend we have an educational calling). We're starting with the Academy: free bootcamps with certification, taught by heavyweight instructors who work in the industry. First cohort in São Paulo, then Latin America. This will extend Unite's impact, because someone has to learn how to run the method. The core is still better creative decisions. ## On calling this science I chose the word science because it forces you to show your basis, your uncertainty and your limits. We won't present correlation as causation or sell prediction as certainty. If the data doesn't support the answer, we'll say so. When we get it wrong, we'll correct it in public and explain what changed. You can hold me to that. ## What Unite is not Media buying stays with the platforms. They do it better than we could. Reporting is part of the process (as a by-product; it isn't what we sell). Traditional research runs at a different pace. We work at the level of each ad, in the operation's real time. Creative teams and agencies stay at the center, including on the calls that depend on taste. We bring evidence and clear the grunt work out of the way. AI for generating ads isn't the product. Volume without the brand's history and knowledge is pointless; it only scales the problem. ## The plan 1. Improve the decision before the spend. Validate has to help the team compare ads and change the choice while there's still time. 2. Learn while media is running. Optimize has to spot patterns and fatigue while there's still time to avoid losses and amplify gains. 3. Take the learning into production. Create has to turn evidence into a creative round with the efficiency, speed and quality brands expect. 4. Make the loop run inside the operation. We need consistency across organizing, briefing, production and approval, automating what's repetitive and keeping context between stages. 5. Instrument the entire loop. Hypothesis, ad, attribute, delivery, result and reason have to stay linked. 6. Turn that history into brand memory and scale the aggregate pattern engine we've already built. 7. Connect creative quality to MMM, incrementality and other business evidence. 8. Train the people who will run this discipline, with scope, seniority levels and a shared yardstick for brand, performance, data and creative. ### To close: what could change the game in ten years Today consumers search, compare and buy. Over the next ten years, much of that is likely to move to AI agents, and agents don't watch ads. Half of consumers already say they'd let an agent swap their favorite brands for a cheaper one (in most categories, price is always the number one reason). In other words, half the loyalty a brand spent thirty years building goes back to square one in a prompt. And what's left to compete on? Being legible to machines, a job that doesn't exist on any org chart today. Media is moving in both directions at once: cheaper per impression, because media providers create new placements faster than the world creates people, and more expensive per result (for the same reason). Ads can now run almost anywhere, even on an assistant or a fridge, while consumer attention gets more diluted and more contested. Delivery is likely to change too. Instead of a finished file, brands may start sending the rules for what any machine can assemble on their behalf, with auditable provenance where the category is regulated. Then comes the countermovement: when everything looks artificial, what's raw and handmade becomes valuable again precisely because it's harder to produce. None of these things alone changes the game, but together with many other shifts we can already see coming, they will have a big impact on our industry (at least once a decade, something new completely changes how the advertising industry works). A lot (or everything) can change, and that's why this text is public. When it stops describing reality, I'll update it.