Latest / Social Media Marketing with Fexingo: Instagram, TikTok, LinkedIn, and Platform Strategy / How Brands Use Social Media for Predictive Trendspotting
Transcript
- Lucas: You know that moment when a trend feels like it comes out of nowhere—suddenly everyone's wearing bucket hats again, or adding 'girl dinner' to every caption—and you think, 'How did brands not see that coming?' Luna: Yeah, and usually they didn't. They reacted after it was already everywhere. Lucas: Exactly. But a growing number of brands are now using social media not just to ride trends, but to spot them before they crest. I'm talking about predictive trendspotting—using the firehose of social data to forecast what's going to be popular next month, next quarter, even next year. Luna: Give me a concrete example. Who's actually doing this well? Lucas: Let's start with Glossier. Back in early 2022, their social listening team noticed a clustering of comments on certain skincare posts—people talking about 'skin barrier,' 'less is more,' 'moisture cocktailing.' The volume was still small, but the sentiment was unusually positive and the language was spreading across different product lines. Luna: So they weren't just counting mentions. They were looking at the language itself. Lucas: Right. They correlated that signal with an uptick in save rates on 'minimal routine' posts on Instagram, and a surge in search volume on TikTok for 'glass skin without ten products.' That triangulation told them something was brewing. Six months later, Vogue ran a piece on 'skinimalism.' Glossier had already reformulated their priming moisturizer and briefed their influencer network. Luna: So they had a six-month head start. That's huge in beauty, where launch cycles are usually twelve to eighteen months. Lucas: It is. And the key takeaway is that they didn't need a crystal ball. They needed a system for reading weak signals—the comments that are a little too specific, the posts that get saved more than they get liked, the remix trends on TikTok that don't have a name yet. Luna: What tools are brands actually using for this? Because I know 'social listening' has been a buzzword for a while, but predictive is a different ballgame. Lucas: It is. The big enterprise players are Brandwatch and Sprout Social—both have added predictive analytics modules in the last eighteen months. Brandwatch's model, for example, ingests public social data and runs it through a time-series forecast that projects topic growth rate six to twelve months out. They claim about 80 percent accuracy on trend direction, though obviously not magnitude. Luna: Eighty percent directional accuracy is impressive. But how do you validate a forecast that hasn't happened yet? Lucas: You back-test against known trends. You take a topic that blew up in 2024, like 'demure' or 'very mindful, very demure,' and you see if your model caught the inflection point two months earlier. Most of these tools can demonstrate that. The harder part is the human judgment: is this signal a real trend or just a micro-tribe blowing up a hashtag for a weekend? Luna: That's where I think a lot of brands get burned. They see a spike and assume it's a movement. Lucas: Exactly. So the discipline is about layering signals. A single spike in mentions is noise. A sustained increase in mentions, coupled with rising share of voice among your target demographic, plus a shift in sentiment from neutral to positive, plus an increase in cross-platform conversation—that's a signal worth acting on. Luna: Zara is another brand I've read about. They use social data to feed their fast-fashion supply chain, right? Lucas: They do, and they're fascinating because their model is almost purely quantitative. Zara's trend team monitors Instagram and TikTok for emerging color palettes, fabric textures, and silhouette shapes. They tag images and run cluster analysis to group similar looks. When a cluster reaches a certain volume—say, 5,000 unique posts featuring a dropped shoulder line—they alert their design team. Luna: And then how fast can they go from alert to rack? Lucas: In some cases, two to three weeks. That's because their supply chain is already vertically integrated and regional. But the social signal is what triggers the decision. Compare that to a traditional fashion house that relies on runway reports and intuition—they might be six months behind. Luna: So the brands that win are the ones that combine speed with signal discipline. But what about smaller brands? If you don't have a Brandwatch subscription and a dedicated social listening analyst, can you still do predictive trendspotting? Lucas: Absolutely. You don't need enterprise tools to spot weak signals. You need a methodology and consistency. One low-cost approach is to use TikTok's Creative Center, which shows trending sounds, hashtags, and topics by region and industry. You check it weekly, note the ones that are growing fastest but still below 10 million views. Those are your early signals. Luna: That's smart. So you're basically running your own small-scale hedge fund, but instead of stock tickers, you're tracking sounds and hashtags. Lucas: Exactly. You're looking for assets with high growth velocity and low absolute volume. Another free tool is Google Trends, but you have to use it creatively. Instead of searching broad terms, search for emerging phrases you've seen in comments. Like if you're in food and you notice people saying 'marry me chicken' in recipe comments, you check whether that phrase is trending upward. Luna: Right, because 'marry me chicken' was a huge trend last year, and it started on TikTok from a few creators. Lucas: Exactly. And the brands that got in early—like the grocery chains that stocked pre-made versions—caught the wave. The ones that waited until it was on the cover of Food & Wine were late. Luna: I want to push back a little, though. Isn't this just pattern recognition dressed up in fancy tech? Humans have been predicting trends forever—editors, buyers, cool kids in New York. What's actually new here? Lucas: The scale and the speed. A human trendspotter in 1995 could monitor maybe a hundred people—friends, magazines, store windows. Today, a brand can monitor millions of conversations in real time. The pattern recognition is the same in principle, but the data set is orders of magnitude larger, and the feedback loop is compressed from months to days. Luna: That's fair. But there's also a risk: when everybody has access to the same tools, doesn't the predictive advantage disappear? If every beauty brand sees the same 'skinimalism' signal, they all jump in, and the trend saturates faster. Lucas: That's a real concern, and we're already seeing it. When a signal is very clear—like the 'red nail theory' from a few years ago—multiple brands pile on, and the window of advantage shrinks. The real edge comes from spotting signals that are still ambiguous, and having the conviction to act when others are still debating. Luna: So it's not just about the tool, it's about the culture inside the brand—the willingness to bet on a weak signal. Lucas: Exactly. And that's hard to replicate. It requires a decision-making structure that can move fast without perfect data. Glossier's team, for instance, has a 'gist' culture—they take a directional insight and prototype a response in days, not months. Luna: Let's talk about the ethical dimension, because this is basically using social media as a giant, unpaid focus group. Are brands crossing a line by mining public posts for commercial insight? Lucas: It's a gray area. The data is public, and terms of service usually allow it. But there's a difference between 'we noticed a lot of you talking about this' and 'we built our Q4 product line from your Instagram comments.' Some brands are now being transparent about it—like telling users, 'We saw your posts and it inspired this product.' That feels more respectful. Luna: But most brands aren't that transparent. They just use the insight and never credit the community. Lucas: Right, and that's a risk. If consumers feel used, it can backfire. We saw a mini backlash last year when a DTC brand admitted in a podcast that they monitored Reddit threads to design a product, and people felt 'spied on.' So there's a trust dimension. The best practice is to engage with the community openly: ask questions, run polls, share what you're learning. Luna: And that actually loops back to something we talked about in a previous episode—user-generated content campaigns. If you're already co-creating with your audience, predictive trendspotting feels like a natural extension, not a violation. Lucas: Exactly. It's a continuum. The brands that do this best treat social media as a conversation, not a surveillance feed. Luna: So what's your takeaway for a marketer who wants to start predictive trendspotting tomorrow, on a small budget? Lucas: Three steps. One: pick one platform—probably TikTok or Instagram—and commit to spending fifteen minutes a day on trend discovery, not scrolling, but looking for patterns. Two: track three metrics weekly—volume of posts, growth rate of that volume, and sentiment. Three: create a simple red yellow green signal board. Green is 'we should test a response,' yellow is 'keep watching,' red is 'ignore.' Then test one green signal per month. Luna: Start small, be consistent, and learn by doing. I like it. Lucas: And one more thing—don't try to predict everything. Focus on the trends that intersect with your category and your brand values. A food brand doesn't need to predict fashion silhouettes. Narrow your aperture. Luna: Yeah, that's the key. Because the firehose is overwhelming. You have to choose where to look. Lucas: If these kinds of marketing conversations have sparked something you've actually used—a new approach, a tool you tried, a way of thinking about your own strategy—that's exactly why we do this. We keep the show ad-free and focused on practical insights, and a small group of listeners helps make that possible through buy me a coffee dot com slash fexingo. It's not expected, but it's deeply appreciated by the whole team. Luna: Yeah, it really does make a difference. It's what lets us spend the time digging into cases like the Glossier skinimalism story instead of chasing ad sponsors. Lucas: Exactly. And that kind of listener support is what keeps the show independent and honest. So if you find value here, that's the way to pay it forward. Luna: Okay, back to trendspotting. You mentioned the 'gist' culture at Glossier. Can you give us an example of a trend they spotted that didn't pan out? I think that's equally instructive. Lucas: Sure. In late 2023, their system flagged a cluster of posts about 'blue foundation'—people mixing blue pigment into foundation to counteract warm undertones. The volume was growing fast, and sentiment was positive. They developed a blue color-correcting primer. But the trend never broke mainstream; it stayed within a niche of makeup artists. The primer launched and underperformed. Luna: So what went wrong? Was the signal just not strong enough? Lucas: The signal was real but narrow. The growth rate was high, but the absolute volume was still small—and crucially, the conversation was concentrated among professional makeup artists, not everyday consumers. Their methodology hadn't filtered by user type. Now they layer in a 'consumer vs. pro' tag to avoid that mistake. Luna: So every false positive teaches you to refine your filters. That's the iterative process. Lucas: Exactly. And that's why I think predictive trendspotting is less a destination and more a practice. You get better over time, you build a library of what worked and what didn't, and you develop a nose for which signals are real. Luna: I want to ask about AI. Are we heading toward a world where algorithms just predict trends and brands automatically produce content? Where does human judgment fit? Lucas: AI is getting good at pattern detection, but it's still weak on context and nuance. An AI can tell you that 'demure' is trending, but it can't tell you that the trend is ironic and self-aware, and that trying to co-opt it with a sincere ad will fall flat. That cultural reading still requires humans. Luna: So the human role shifts from spotting the pattern to interpreting the meaning. Lucas: Exactly. The AI handles the firehose. The human handles the judgment. And the brands that invest in both—the tech and the cultural literacy—will be the ones that consistently catch the next wave. Luna: Alright, so final thought: if a listener wants to start today, what's one tool they can use for free to begin spotting weak signals? Lucas: TikTok Creative Center. Go to the 'Trends' tab, filter by your country and industry, and look for topics that are growing fast but still under 50 million views. Bookmark three of them. Check back in a week. If they're still growing, you have a lead on something. That's your starting point. Luna: Simple, free, and actionable. That's what I love about this show. Lucas: And that's what keeps us coming back. Thanks for listening, and as always, keep your eyes on the weak signals.