Valid’s cover photo
Valid

Valid

Software Development

San Francisco, CA 1,802 followers

The infinitely scalable ad agency.

About us

Valid is a tech-enabled marketing agency that pairs expert marketers with proprietary, best-in-class tools to achieve extraordinary results. Our technology amplifies human capabilities to deliver marketing outcomes that were previously impossible.

Website
https://valid.co/
Industry
Software Development
Company size
11-50 employees
Headquarters
San Francisco, CA
Type
Privately Held

Locations

Employees at Valid

Updates

  • Valid reposted this

    We built two AI influencer personas: Mia and Marcus. Mia hit 54K Instagram followers in 3 weeks. Marcus just crossed 100K followers. And when brands “hire” them to promote their content, they get the same or better results that they would with human influencers. Mia and Marcus have real audiences that we built the old-fashioned way: by posting great content that resonates with their target audience. Marcus targets a tech audience that skews male. Mia specifically targets users interested in women’s health and wellness. We figure out what resonates with the same playbook we use for paid media. We find organic content that's already performing well in the target category, then reverse-engineer what made it work: script, format, hook, pacing, etc. That’s how we’re able to grow followers by tens of thousands every week. And we take the same approach to our AI influencers’ sponsored posts. We're trying to spawn an army of these guys. In the future, I imagine treating our army of AI creators as testable variables, the same way we treat any other creative asset. We’ll be able to run sponsored posts across multiple AI creators, measure which one connects with the brand, then keep going back to the one that performs. It's the same logic as going back to a human creator whose creative does really well. We’re just doing it with AI. Have you experimented with AI influencers yet?

    • No alternative text description for this image
    • No alternative text description for this image
  • Valid reposted this

    View profile for Rahul Bhardwaj

    Valid2K followers

    Every ad in this reel was made by Valid in the last 90 days. Here's what we've learned about making creative that actually performs. It's equal parts art and science. A few things that really matter: 1) Create for a very niche audience. The viewer needs to feel like "that's totally me." Good performance ads don't have broad appeal. 2)The ad should spark curiosity. You don't need to do the full sell in the 15-second ad. You need to get the person to click on the product. Let the landing page do its job. 3)A good hook stops the scroll and makes you watch to the end. 4)When you have a winning concept, iterate. 5)Don't let ad concepts that feel good die if they don't work right away. If your gut says it'll work, often you're just 10% away from a winner. You'll notice none of this has to do with whether it's AI or not. AI is an implementation detail. To make good ads with AI, the process is the same as making good ads the old way. We still judge effectiveness on ROAS. Not on the fact that it's AI.

  • Valid reposted this

    View profile for Jake Baumann

    Valid2K followers

    AI ads are finally here. And they don't suck. For 6 months, we've been quietly building Valid AI Creatives and we've already worked with brands like Replit, Tolan, and HoneyHomes. The problem with AI ads isn't the technology. It's that 99% of AI creative feels soulless. Uncanny. Skippable. So we built something different: AI production, human thinking. Here's how it works: 1. We connect to your ad account 2. We study what's actually working 3. We find concepts worth producing 4. We make the ads and launch them And the early results are wild: • 17x higher win-rate than other agencies in one account • 71% drop in CPA in another • Earned #1 spot in multiple accounts spending over $1M/month • Media buyers allocate 53% of their spend to our creatives on average That last one matters most. Media buyers don't throw money at bad creative. They follow performance. If you want creative that stops the scroll and actually converts, we're ready.

  • Valid reposted this

    Turns out… AI ads don’t have to feel like AI ads. AI speed. Human taste = Faster. Better. Scalable. The early result from our beta customers Valid are wild: • 17x higher win-rate than other agencies in one account • 71% drop in CPA in another • Earned #1 spot in multiple accounts spending over $1M/month • Media buyers allocate 53% of their spend to our creatives on average If you want creative that converts, reach out.

  • Valid reposted this

    Today I’m excited to announce Valid AI Creatives. Already trusted by brands like Replit, Tolan, and HoneyHomes. For the past 6-months, we've been quietly building a system that produces AI-generated ads engineered to perform on Meta/TikTok. We don't make the uncanny, soulless AI-generated ads you're used to. We’re making creative that stops the scroll. AI ads that make you feel something. Built on real strategy, real hooks, real storytelling. AI-production, human thinking. Here’s exactly how it works: 1. We connect to your ad account 2. We study what's actually working 3. We find concepts worth producing 4. We make the ads and launch them The early result from our beta customers are wild: • 17x higher win-rate than other agencies in one account • 71% drop in CPA in another • Earned #1 spot in multiple accounts spending over $1M/month • Media buyers allocate 53% of their spend to our creatives on average That last one matters most. Media buyers don't throw money at bad creative. They follow performance. If you want creative that stops the scroll and actually converts, we're ready. I’m so excited to bring this out of stealth. Now back to work.

  • Valid reposted this

    We mispronounced "hazelnuts" on purpose. Here's the full breakdown of why. Watch the Pixar-style animated ad for Coco attached below and you'll catch it. The word "hazelnuts" is off. The pronunciation is wrong. We built it that way from the start. A mispronounced word creates a small cognitive snag. The brain registers something slightly unexpected and holds onto it longer than a clean, polished line. That cognitive snag produces comments. People point it out. Some ask if it's intentional. Some just quote it back. All of that activity registers as engagement. More engagement tells the algorithm the ad earns attention, which lowers the cost to distribute it. We built one deliberately wrong word into the script to move a media-buying metric. That’s the performance data side of things. The other layer to this is the Pixar-style animation, which offers a whole new creative format for storytelling. Traditionally, timelines and budgets kept animation out of the paid feed. Animation is the next creative lever in paid acquisition. Valid Studios is our answer to that. We're in the process of building it now, more to come. But the goal is Super Bowl-level quality, AI-native workflows, and a production timeline measured in days, not months. If you're running paid creative right now, what's the most counterintuitive production decision you've made that moved a performance metric?

  • Valid reposted this

    In ad tech, most off-the-shelf tools get you 80% of the way there. The last 20% is where you find the performance gains. The problem: SaaS tools are built for the average use case, a one-to-many model that works well for software that achieves the goal but doesn't need to be right for your specific situation. Workday and Rippling are good examples. We use them because they’re the best option out there for what they do, but they’re cumbersome because they’re not tailored for your unique way of working. But in marketing, that 20% gap has repercussions beyond annoyance. It hinders your ability to optimize performance. And what you're optimizing for changes constantly. A tool built for the average marketer can't keep up with that. That’s where a marriage between Claude Code and MCP servers comes in. You can get 100% of the way there because the tool is completely personalized to your particular application. We built our own MCP server that allows you to query results directly from Meta and TikTok, plus mobile measurement partners like Adjust and Rapswire, and then we connected it to Claude Code. With the two together, we can set up the exact reporting, polls, and analysis that matter to our clients. The performance data is joined with the creative data in the shape we want it, so Claude Code has all the data it needs to do whatever we ask it to do (and we ask it to do a lot). Then every time what we want changes, or what our clients’ want change, we just tweak our setup. We can spin up a new dashboard or new automation in seconds. Our MCP server is hands-down the team’s favorite tool because it makes it so easy to dial in that last 20%. If we’re experimenting with a new optimization strategy and we need new insights or different types of alerts, we just…ask our friend Claude. To get something like this to work, you need a really strong data pipeline (we use Clickhouse) and a tool that knows how to read that pipeline and assemble the data correctly. That's the architecture. And it’s what the next wave of software looks like. If you’re going to DIY this: Don’t try to connect an account to an MCP server without first reviewing the platform’s terms of service. And if you’re working with an agency connecting to your accounts via MCP, ask them for the documentation to show you they’re doing it the right way.

    • No alternative text description for this image
  • Valid reposted this

    Can AI UGC work? It can. I've seen it. The real question is, "Does your story work?" Because AI UGC just scales whatever you put into it - good or bad. A lot of companies experimenting with AI UGC aren’t seeing the results they expected. In almost every case, the creative was technically fine. But the story was missing. I wish I had a secret formula to share. Good creative doesn’t follow a formula - it shifts too much from brand to brand and ad to ad. But two things show up in ads that perform: → A hook that stops the scroll and gives you a reason to keep watching (not just pattern interruption) → A message delivered in a storytelling structure because that's how our brains retain information AI UGC can execute both of those things. But it can't invent them for you. The teams seeing real results with AI UGC are the ones who've already figured out their story. They're using AI to scale something that already works. What's been the hardest part of getting AI UGC to perform for your brand?

    • No alternative text description for this image
  • Valid reposted this

    Apple's App Store review queue is bursting at the seams right now. Vibe coding is a big reason why. It’s become super easy for anyone to create an app. The sheer volume is enough to cause delays. But the other issue is tools like Replit and Vibecode let users change the functionality of an app (and even build apps within apps) without going through App Store review. Apple's Review Guidelines have always prohibited this. The difference now is vibe coding makes it easy to do at scale, and Apple has to look at every app more closely. Here’s the takeaway for mobile app teams: App Store review is now the step that slows development to a crawl. And review delays are going to get worse before they get better. Build that into your launch timelines now. I agree with Apple on their underlying principle. App Store review is a good thing for the end user. It keeps the scammy stuff out. Even if 99% of vibe-coded apps are totally fine, Apple can't afford the 1% that aren't. Are you already seeing longer review times on your apps? What's your current buffer in your release schedule?

    • No alternative text description for this image

Similar pages

Browse jobs

Funding

Valid 2 total rounds

Last Round

Seed

US$ 5.5M

See more info on crunchbase