User Engagement Strategies

Our collective “algorithm literacy” is all over the place right now, and I blame it on the chasm between the post analytics made available to us, and the actual mechanisms that impact how content performs. The reality is that the data we have access to is just a sliver of what’s actually going on inside the feed. Unsurprisingly, in the absence of a more comprehensive overview, we tend to over-index on the direct user engagements, like: impressions reactions reposts comments But under-index on the indirect user engagements like: dwell time scroll depth completion rate session duration My main beef (in all honesty, the better word is concern, but this is a social media platform after all 🤷♀️) is that LinkedIn posts with methodologically flawed A/B tests keep dominating the feed, and they inspire a kind of oversimplification of very complex systems. What’s important to understand is that post performance is impacted by a dense consolidation of factors, probably more than we can comprehend, and many of them hinge on time — session length and reaction time being major ones. So if you’re wondering why long comments, and carousels, and good hooks, and pictures make content perform better ➡️ it’s because they increase time spent on platform. And if you haven’t caught on to why basic reactions, emoji comments, and walls of text barely move the needle ➡️ it’s because they *don’t* increase time spent on platform. Here’s your reminder that algorithms are not easily "hacked", you don't have control over user behaviour, and your best strategy is to make good content that people actually want to devote their time and attention to.

Are you still polling your backend every few seconds just to get updates? That’s not just inefficient — it’s a terrible user experience. In my latest video, I walk you through how to implement real-time communication in .NET using SignalR. We’ll build an order tracking system that sends live updates from the backend to the frontend. ✅ Configure SignalR in .NET ✅ Build and secure your Hub ✅ Target specific users using User() ✅ Secure it with JWT ✅ Update the UI instantly via WebSocket Oh, and I’ll also show you how to make your SignalR client methods strongly typed — cleaner code, fewer bugs. Watch the full tutorial here: https://lnkd.in/e6hH5gak If you’re building modern web apps with ASP .NET Core - this is a must.

Want to boost your charity's LinkedIn presence? Here's how to maximise engagement and create meaningful connections on the platform 👇 Post consistently Maintaining a regular posting schedule is crucial for LinkedIn success. Aim for 3-4 posts per week, focusing on peak engagement times (typically Tuesday to Thursday, 9am-2pm). Use LinkedIn Analytics to understand when your specific audience is most active and adjust accordingly. Share impact stories LinkedIn audiences respond particularly well to real stories of impact. Showcase human rights wins, volunteer achievements, and organisational milestones. Include compelling statistics and data to demonstrate the tangible difference your charity makes. Remember to always obtain proper permissions and maintain dignity in storytelling (more on this soon). Interact with other charities Expand your network by engaging with other organizations in your field. It will allow your charity more visibility and access to new and aligned audiences. Leverage advocacy Encourage your colleagues to share and engage with your content. Their networks can significantly amplify your reach. Communicate with internal staff on how to share posts effectively and provide key messaging points. This organic approach often generates more authentic engagement than paid promotion! Mix up your content formats LinkedIn rewards diverse content types. Alternate between text posts, images, videos, and documents. Native LinkedIn videos typically perform better than external links. Use carousel posts for impact reports and infographics - they're great for explaining complex issues in digestible formats. Engage authentically with your community Don't just broadcast - participate in conversations. Respond to comments promptly, ask questions in your posts, and actively engage with other organizations in your sector. Join relevant LinkedIn groups and contribute meaningfully to discussions. This helps build a genuine community around your cause. Optimize your company page Ensure your page is complete with an engaging about section, regular updates, and clear calls-to-action. Use keywords that your supporters might search for. Keep your banner image and profile picture aligned with your brand guidelines. Track and adapt Use LinkedIn's built-in analytics to monitor what works if you don't have access to a social monitoring tool. Pay attention to post engagement rates, follower growth, and click-through rates. Test different approaches and refine your strategy based on data, not assumptions. Charity social media managers - what other LinkedIn strategies have worked well for you?

LinkedIn shut down the Employee Advocacy tool. Why LinkedIn couldn't make it work? Here are the three myths that killed traditional employee advocacy programs: Myth 1: "Our employees will naturally share company content" Reality check: They won't. Without incentives, even your most loyal employees scroll past your branded posts. Why? Because sharing corporate content doesn't benefit them personally. Myth 2: "Internal advocacy is enough to reach our buyers" Wrong. Your employees' networks overlap heavily with each other. You're basically shouting into an echo chamber of people who already know about your company. Myth 3: "Employee advocacy is free marketing" Nothing's free. Traditional programs require massive internal resources: content creation, training, platform management, and constant nudging. The hidden costs are brutal. Meanwhile, companies following this broken playbook are missing the real opportunity. The smartest brands are treating their employees like creators—and paying them like creators too. Add your employees to creator campaigns, but incentivize them properly. A few hundred dollars per post transforms reluctant participants into enthusiastic advocates. Better yet? Expand beyond your internal team. That Microsoft engineer who's not your employee but understands your buyers' challenges? They're probably more credible than your own marketing team. This is exactly why Synnc was built—to connect brands with credible corporate creators who can speak authentically about industry challenges, whether they're internal or external. What's been your experience with employee advocacy programs?

X just published its algorithm. Here's what actually powers your feed, you'll be surprised 😯 Musk posted the full code, organic posts, ads, all of it, on GitHub. The engagement weights are now public: 🐦 Reply that the author responds to: +75 🐦 Reply to a tweet: +13.5 🐦 Profile click + like/reply: +12.0 🐦 Retweet: +1.0 🐦 Like: +0.5 🐦 Video watched halfway: +0.005 A like is worth almost nothing. A conversation you actively participate in is worth everything. One reply chain where you engage back outperforms hundreds of likes. The negative signals are just as revealing. As some experts estimated: ✖️ Reports carry a -369x penalty. ✖️ Blocks and mutes hit -74x. And those penalties don't expire on the tweet, they follow your account. What this means for creators: stop optimizing for likes. They're the lowest-value action on the platform. Ask questions. Spark conversations. Reply to your own posts. The algorithm rewards depth, not applause. What this means for audiences: the research behind these weights is harder to ignore now that the code is public. Studies consistently show that anger and out-group hostility drive the most replies. The formula that rewards replies is, structurally, a formula that rewards divisiveness. That's not an opinion — it's the math. Which brings me to the bigger point. Every major platform, YouTube, Instagram, TikTok, LinkedIn, runs on algorithms that determine what billions of people see every day. None of them are required to show you how. X is the exception, not the rule, and only partially because of regulatory pressure in France and the EU. Platforms should be required to publish their ranking logic regularly. Transparently, in plain language, with meaningful accountability. 🤔 What do you think, should algorithm transparency be regulated, or does that create more problems than it solves?

There’s a difference between customer advocacy and collecting logos. Too many early stage SaaS companies still think the game is: ☑ Get a big name logo ☑ Put it on your homepage ☑ Job done (Dare I even say one of our competitors' site has quite a few logos that aren't their customers, some are even ours 🤷 ) ...and I think buyers are wising up to this... Same goes for influencers. If they’re not using the product or if they’re not mentioning you without being paid to, you’re not building brand. You’re renting inauthentic attention. At trumpet 🎺 we’ve started leaning heavily into real advocacy. The kind that isn’t bought, briefed or branded. What it looks like: → Users looping in colleagues offering great case studies → Our Slack channels getting DMs oh how users are getting results and telling us excitedly → Prospects mentioning us before we’ve even reached out → Champions posting about us unprompted on LinkedIn → Customers recording walkthroughs for their wider team → Referrals landing in our inbox with a one-line intro That’s the gold. It doesn’t just drive pipeline. It compounds. If you want to actually use advocacy beyond a throwaway customer quote on your site, here’s what we’ve found works: 🔁 Give them a reason to share. Make the product so damn useful (or delightful) they want to talk about it. Most advocacy is a reflection of product quality, not just marketing effort. 🧠 Involve them early. Co-create features, roadmap check-ins, share sneak previews. People advocate for what they help shape. 📦 Re-Package the advocacy. Turn casual quotes into killer social proof. That Slack comment? It can become a slide. That LinkedIn post? It can become an ad. But only if it’s genuine. 🙋 Celebrate the humans, not the logos. Spotlight your champions. Not the company they work for, but the person who took a bet on you. They’re the ones who’ll take you into their next company. We're not perfect at this yet - but it’s the main kind of “marketing” that doesn't need large budgets and that everyone can do. Logos are easy. Advocacy is earned.

If your message needs to convince someone to talk to you, it’s already too late. Outreach only works when recognition comes before reach. Most people jump straight in DMs with copy-paste intros like  "Hey, we help businesses grow 10X," or  “Let’s connect, I’ve got something that’ll help your business.” But every time you message someone, they don't just read your pitch. They read you. Your headline. Your content. Your comments. If none of that builds credibility, your DM doesn’t stand a chance. At our agency, we’ve sent over 10,000 DMs in the last few months across industries, and here’s the 5-step strategy that actually works: 1. Brand warm-up  Before sending a single message, we make sure the profile has already built trust. That means a clear headline, proof-driven bio, and at least 3–4 posts that make prospects curious before contact. 2. Audience Mapping We segment precisely like “B2B founders doing $1–5M” or “coaches scaling beyond 30K/month.” Relevance beats volume every single time. 3. Strategic engagement  Before reaching out, we spend a few days engaging on their posts with comments, insights, and small interactions. When they finally get a DM, they already recognize the name. 4. Strategic Engagement  We don’t pitch. We start conversations. Every message connects to something they’ve said, done, or shared. That’s what makes it human. 5. Follow up and nurture  If they don’t reply, we don’t chase. We stay visible, show up in comments, and keep adding value. Have you ever replied to a DM just because the person already felt familiar? #LinkedInStrategy #B2BMarketing #PersonalBranding #SalesEngagement #RelationshipMarketing

One of the most common topics I am asked about is the LinkedIn algorithm. LinkedIn doesn’t openly share how the algorithm works in its entirety but when you’re working across multiple client accounts, having regular conversations with peers, and staying close to platform trends, you start to spot what’s working (and what’s not). Here’s what I’m seeing right now if visibility and reach are part of your strategy. ➡️Prioritising Relevance Over Virality LinkedIn has an emphasise on relevant, professional content rather than chasing mass virality. In fact, LinkedIn’s editor-in-chief Dan Roth has stated the platform “is not designed for virality,” but instead for sharing useful knowledge and insights”. ➡️Your Network Connections Drive Visibility LinkedIn favors content from your own network meaning your connections and followers are far more likely to see your posts. The takeaway: the makeup and engagement of your network now directly impact your reach. ➡️Expertise Signals Expand Your Reach The feed algorithm doesn’t just assess what you posted it also looks at who is posting and whether you’re have domain expertise. If you consistently share content in your niche and have the professional background to back it up, the algorithm will distribute your posts more broadly. Conversely, if you post on a subject completely outside your known field, expect limited reach. ➡️ Commenting Beyond The Surface We know that the algorithm gives more weight to thoughtful, substantive comments than to quick reactions. But did you know that LinkedIn is evaluating who is commenting? If your post on marketing draws many comments from marketing professionals (i.e. people relevant to the topic), that’s a strong positive signal - if it's the same people without professional relevance it has the opposite effect. ➡️Quality of Engagement Over Quantity It’s not about getting lots of engagement; it’s about the right engagement. LinkedIn explicitly encourages creators to focus on reaching a targeted, relevant audience. The algorithm measures engagement quality being who is engaging. For example, a dozen comments from respected peers in your industry will boost your post more than a hundred random likes from outside your field. ➡️Low-Quality Content Gets Demoted LinkedIn’s algorithm actively filters out content deemed low-quality or spam, so certain tactics will hurt your reach. Similarly, tagging a bunch of people who aren’t relevant to the post or overstuffing your post with hashtags are red flags. Posts with poor formatting or error-ridden text can also be classified as “low quality.” 🔍 Key Takeaways The algorithm is always evolving but the core principle and advice remains the same: create content that’s useful, relevant, and credible to your audience. 💬 Curious to see how your content stacks up against these points? Or have you noticed shifts in your own reach lately? Let me know in the comments. #LinkedIn #Marketing #Content

X just open sourced their algorithm. No hand-tuned features. No manual weights. Just a Grok transformer learning from your behavior. Every time you open X, the system needs to answer: out of hundreds of millions of posts, which ones should appear in your feed? The old approach involved manually designing features and carefully tuning how much each feature mattered. But the new system uses a Grok-based transformer looks at your engagement history (what you've liked, replied to, shared) and learns what's relevant to you. The pipeline has two stages: 1️⃣ 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 Narrows millions of posts down to ~1000 candidates from two sources: 𝗧𝗵𝘂𝗻𝗱𝗲𝗿: An in-memory store that does sub-millisecond lookups of recent posts from accounts you follow. 𝗣𝗵𝗼𝗲𝗻𝗶𝘅 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹: This is more interesting. It uses a 𝘁𝘄𝗼-𝘁𝗼𝘄𝗲𝗿 𝗺𝗼𝗱𝗲𝗹 to find relevant posts from accounts you don't follow. - User tower: Encodes your features and engagement history into a normalized embedding - Candidate tower: Encodes all posts in the corpus into embeddings - Similarity search: Retrieves top candidates via dot product similarity 2️⃣ 𝗥𝗮𝗻𝗸𝗶𝗻𝗴 The Grok-based transformer takes your engagement history and the candidate posts, then predicts probabilities for ~15 different actions: liking, replying, reposting, clicking, but also negative actions like blocking or muting. The model outputs probabilities for all these actions, which get combined into a final score. Positive actions get positive weights and negative actions get negative weights, pushing down content you'd likely dislike. What you're seeing in your feed right now is the product of this system processing your engagement history through a transformer architecture - originally designed for language modeling, now repurposed to predict what content keeps you scrolling 🤔 GitHub repo: https://lnkd.in/d2vvF5ee

Exciting research from Snap Inc.'s engineering team! Just came across their paper on Universal User Modeling (UUM) that's revolutionizing how they handle cross-domain user representations. The team at Snap has developed a framework that learns general-purpose user representations by leveraging behaviors across multiple in-app surfaces simultaneously. Rather than building separate user models for each surface (Content, Ads, Lens, etc.) and combining them post-hoc, UUM directly captures collaborative filtering signals across domains. Their approach formulates this as a cross-domain sequential recommendation problem, processing user interaction sequences of up to 5,000 events and using sliding windows of 800-length subsequences to balance computational efficiency with capturing long-range dependencies. The architecture leverages transformer-based self-attention mechanisms to model these sequences, with a clever design that projects feature vectors from different domains into a shared latent space before applying multi-head attention layers. The results are impressive! After successful A/B testing, UUM has been deployed in production with significant gains: - 2.78% increase in Long-form Video Open Rate - 19.2% increase in Long-form Video View Time - 1.76% increase in Lens play time - 0.87% increase in Notification Open Rate They're also exploring advanced modeling techniques like domain-specific encoders and self-attention with information bottlenecks to address the challenges of imbalanced cross-domain data. This work demonstrates how sophisticated user modeling can drive substantial engagement improvements across multiple recommendation surfaces within a large-scale social platform.