Designing Smarter Social Media Strategies with Machine Learning

Public speaking? Completed it ✅ (and I absolutely loved it)

A couple of weeks ago I got to try something I’ve always wanted to do. Public speaking. And guess what… I absolutely loved it.

What did I speak about you ask?

In my talk, ‘Designing smarter strategies: Machine learning for impactful social media strategy’, I spoke about the intersection between creativity and data, and how every single post we publish causes ripples. These ripples carry insights that, when mined and labelled properly, and as our dataset grows, reveal fascinating clues into how our audiences really think, feel, and behave.

To explain this I combined established ideas from psychology, marketing, behavioural science, and machine learning, reinforced by my own work at The MTM Agency, and shared a top-level exploration of how raw social media data, often overlooked in favour of surface-level dashboards, can and should be mined and labelled across three core dimensions: content, text, and time. Once that data is sufficiently structured, and as the dataset grows, pairing it with machine learning techniques allows us to uncover the underlying behavioural patterns that explain which variables are actually driving results.

I outlined how this kind of long-term, strategic approach allows us to change our level of magnification and raise the level of our work by moving beyond traditional, sporadic day-to-day output (or firefighting, as I call it) and away from an overreliance on short-term performance indicators like vanity metrics. Another reason for the need to understand these patterns is that in doing so it helps supports a reasonable case for repositioning social from being seen as the “fluffy” discipline within marketing to a practical, testable science.

That said, this more scientific approach should never replace what makes social media special, and that is the creativity within it. What I suggest and anticipate is that social will increasingly become a space that blends creativity and data to compound long-term success. Because when you understand not just what your audience wants, but how they behave and respond, it becomes a fact that you can plan more intelligently and act with far greater precision.

I truly believe that it is an approach of this kind that will be the next step for success on social media. Furthermore, it is most likely that the path to the best solution will be pioneered by agencies or by any business producing a high volume of content with a large enough dataset to build a more holistic understanding of social consumption and engagement. From there, client or business-specific algorithms can be designed that don’t just react to what got likes yesterday, but calculate what should work next. Not in isolation, but as part of a sequence. A series of posts with the right mix of variables, structured to produce the kind of ripples that lead to an objectively meaningful increase in success metrics, not just engagement.

Visually I see it a bit like how Deep Blue played chess or how AlphaGo approached the game of Go. Not by predicting the future perfectly, but by analysing everything that has come before, weighing up the available options and most recent inputs, and computing the next best move based on all reasonably measurable variables and against operational constraints. I see this as virtually the same process for algorithms that sequentially react and organise the next best post based on the most recent engagement. In this way through every newly labelled post and every new interaction they will ultimately feed into a system that becomes increasingly smarter and more accurate over time.

As I mentioned previously, this system should never replace creatives or social media managers. It should be used only to support and enhance their success. Ultimately it is the partnership between the two that will help us not only make sense of the madness that is social media but also better understand what creatives need to include to maximise results. Almost like giving them a box to work within and then the freedom to colour beautifully inside it. Ultimately, these steps, supported by the right operational workflows, help reduce the threshold for error. We get more shots on target, we miss less often, and over time this builds exponential learning that aligns more closely with wider business goals.

As you can imagine, the talk was dense. There was so much I had to remove due to time constraints. Like the classic gorilla video covering selective attention, and some of the neurological theories I wanted to use to frame why chasing trends can actually damage brand positioning. Especially if people aren’t looking for you anyway, and especially when they are in a liminal state while consuming content.

Towards the end, I could only briefly touch on how this process could be extrapolated into the design of intuitive omnichannel ecosystems. Systems that become increasingly aware of customer needs and behaviours. Systems that should help brands position themselves more ethically in todays AI gold rush, where transparency, authenticity, intelligence, and autonomy all matter equally.

Overall, I was genuinely overwhelmed by the positive feedback I received both online and in person. I massively appreciate The MTM Agency for giving me the opportunity and everyone who supported my first (and hopefully not last) public talk.

Who knows, maybe I’m onto something, maybe I’m a mad scientist, or maybe I just need more data… Either way, I’m excited to keep exploring how these specialisms intersect to help us become more intelligent, more creative, and a little less reliant on luck and intuition.