For the last decade, I’ve given my thoughts on trends for the year ahead. But trends are merely what could happen. This year, I’ve flipped the focus and look at some of the biggest questions coming into 2024. They don’t really have easy answers but tell us a lot more than a prediction that may or may not come true or is a bit anodyne.
Onwards
What do we need to answer in 2024?
I read a lot of trend decks for work. Some are useful. Some are exceptionally well researched. Others clearly exist to position the SaaS tool (and it’s almost always SaaS tools) as the answer.
When I sat down to write my predictions for this year, there were a lot of don’t knows and maybes. Not the kind of content that makes for compelling reading.
So rather than try to shoehorn in a trend, let’s look at this from a different way. The below are questions I don’t necessarily have the answer to, but we should probably be asking this year.
What happens when you can't trust anything?
Dystopian visions of the future are usually a strong premise for science fiction, and they're often underpinned by the concept of Trust. You can draw a pretty straight line from Orwell's 1984 through to the likes of The Running Man and The Hunger Games. Even Disney's most recent outing, Wish, plays with these themes.
But while The X Files' "Trust No One" mantra made great supernatural entertainment in the nineties, as we approach 2024, it appears to be the default. Reality literacy, if that's a thing, is low.
The concept of trust only really works if a large majority buys into a Hobbesian contract: everyone agrees to play by the same rules and ensure bad actors aren’t rewarded. But when bad actors push at the rules and discover blowback is minimal then there's no disincentive.
This may seem alarmist. History is full of panics such as Y2K that have amounted to nothing. And concerns about trust may too pass. But the collapse of trust is more abstract than a specific panic like Y2K and involves several shifts that touch on multiple elements from tech to politics, via everything inbetween.
Possibly the best - or worst - place to start is the decline of traditional media. Journalism - especially the kind that holds power to account - is never easy to monetise. It's why, pre-internet, classifieds were a source of revenue for many media organisations. And no matter how much disruptors such as Buzzfeed and Vice come in to challenge the status quo, making money is still a problem. That's why traditional local media has been gutted.
At the same time, bias is more profitable than neutrality (not that journalism is ever really neutral) while those organisations who strive for neutrality are called out for any hint of bias (often from those with a position of bias). The net result is a fragmentation of news consumption in a way that can be partisan or inaccurate or both.
Amplify this through social media, where platforms are at best overwhelmed by the scale of disinformation and at worst have a reluctance to tackle bad actors. Throw AI generated content into the mix and it requires the average person to question almost everything they see or read. That’s exhausting, when everything from wood carvings to election campaigns can be faked.
In the last week, we’ve had explicit deepfake videos of Taylor Swift go viral on X. Disinformation is already a game of whack-a-mole. Generative AI makes it even harder (for more on this I recommend Casey Newton at Platformer).
With major elections coming in 2024, plus several potential tinderboxes around the world, from Russia to the Middle East to Taiwan, humanity’s critical thinking could be stretched to its limits.
What is AI’s definition of useful?
AI will almost disrupt multiple industries and to change the way we work. But for all its promise, it’s still more cool than useful, especially when looked at through the prism of the industry I work in (media and marketing).
Yes, the natural language it uses is impressive and convincing, and the images that can be created through prompts are stunning, but are they useful?
Chat GTP, Dall-E, and Midjourney are seriously impressive pieces of technology but a lot of the work is a quick, glorified intern who sometimes makes mistakes and whose output is only as good as the person who gives the instructions.
Midjourney can replace stock imagery, but you still have to pay for it, and searching iStock may be quicker than entering in the right prompt in some occasions. Chat GTP can write good performance marketing headlines that get a better reaction from audiences, but editing and fact checking longer prose can take longer than a subject matter expert writing a piece with clearly labelled sources.
Sports Illustrated is an example of what happens when content is created for content’s sake without any due care and attention. It may generate traffic and advertising dollars but damages the brand in the process. Every tech business now feels compelled to add AI, but are the AI functions useful?
At the moment, we’re still at somewhat of a blank canvas with AI. It’s not to say the technology isn’t useful, but given the potential behind AI, it seems somewhat of a waste if the best we can come up with is replacing average human copywriters and photographers with average machine generated copy and images.
So what’s likely to happen in 2024? A safe prediction is that AI embeds itself naturally in a lot of functional work to the point where it becomes second nature. Research GTPs could save marketers a lot of time getting started.
Mark Ritson and Tom Roach, two marketing minds not noted for hyperbole with new technology, both are enthused about the capabilities. Ritson believes AI enabled ‘synthetic data” could be a game changer, while Roach gives plenty of practical examples as to how his agency, Jellyfish, is using generative AI.
Elsewhere, cropping and replacing elements in photography is already embedded into a lot of smartphones as well as Adobe Suite, and performance marketing copy is a very obvious place where machine can triumph human.
The second development around AI is much more internal focused and goes beyond mildly useful create outputs. Most businesses have very specific problems or needs or questions around large data sets where AI may be able to advance the business, or suggest solutions that aren’t obvious.
For example, an automotive business that wants to predict when people are ready to buy a new car and what type of car they’re likely to be in the market for may find answers in a very large but messy data set. Public health bodies are notoriously bad at tying data together for better patient outcomes.
But here the problems are very specific to the business or public body, and the data a LLM would need is highly confidential. ChatGTP could do some of the job well, but there’s always a question of how private will the data be, and how well does a large blank canvas fit a specific problem set.
So expect some of the more interesting applications of AI to come from businesses or bodies who probably won’t share these with the world until much later.
Then there’s the AI powered tech and applications that we don’t know because we can’t comprehend or imagine the output. In the early 2000s, most of us were used to the idea of what a mobile phone did (make calls, send texts, play Snake), and we could probably envisage our phone doing some of the jobs that are standard to most smartphone apps, but that didn’t mean we knew what an iPhone would look like or that we needed smartphones.
This is probably where AI is at right now. The interesting applications of AI probably are another 6-12 months down the line as people start asking questions and experimenting with building on what exists today into something we won’t have thought of. Think of Open AI and other LLMs as a platform to build on, in much the same way as programming languages. What’s put on top of the code does more than the code itself.
That sounds vague, but if I’d thought of what “it” might be, I’d be pitching for investment right now.
What are the implications of the evolution of the next generation of social media?
What classifies as social media today? Two of the biggest channels - YouTube and TikTok - are essentially broadcast mediums with social elements. It’s entertainment first and foremost with a higher barrier to entry but potentially higher reward if your objective is to build a profile.
Text-based social networks still have their place - a low barrier to entry, but easy for users to shout into the void. Threads is an interesting experiment on what happens when you tried to build a Twitter replacement in real-time but it still doesn’t have the cultural cache that Twitter, for all its faults, commands, even today. Bluesky is an interesting experiment on what happens when you try to build the perfect decentralised Twitter replacement in painstakingly slow real time. Mastodon is still too much work for the average person.
Meta’s platforms, meanwhile, are still what they’ve always been: the equivalent of a country’s largest bank. Sure, you could move elsewhere and you probably have a couple of other bank accounts you use for specific purposes. But it’s too much hassle as that’s where your money is. But they’re not exactly always enjoyable life partners.
The tech industry may be excited about decentralised networks, and they may well be the future, but for the average person who just wants to chat to their friends, connect with people who they may want to know in the future, or discuss mutual interests with strangers, what does this offer them that a less complicated alternative? If TikTok and Discord are social networks, what does that make Fortnite, Roblox, or connected soccer games from EA?
The net result for advertisers is a very fragmented media landscape, which both helps and hinders Pinterest has always had excellent planning signals - the platform’s very good at knowing if you’re about to book a holiday, buy a house or have a baby - but it doesn’t necessarily offer the scale when it comes to sophisticated mass market penetration.
Some marketers may be very keen to target active Subreddits but does this give better ROI than a broad buy across Meta? And what does the death of cookies and personalised targeting mean? Probably not as much as doomsayers think.
What this will probably boil down to for marketing is twofold: it’s never been more important to have a clear brief that a media planner or agency can really understand and make the best choices on a limited budget. And that 1+1=3 for sophisticated mass market penetration. TV won’t build a brand alone, but can TikTok?
Those with the best econometric models will probably have the best answers to this question.
What happens when people stop spending?
Nearly two years ago, I wrote that the cost of living crisis was the biggest trend to watch in marketing. At risk of repetition, the cost of living crisis is still one of the biggest trends in marketing going into 2024.
When even UK money saving expert Martin Lewis says he’s out of ideas and the government needs to step in, and that, in general, most people in the big economies seem to be pretty gloomy about their finances, we’re going to be in for a bumpy ride, and no amount of AI can persuade people to buy if they don’t have money - although it may be able to help set pricing at a level that benefits both brand and consumers.
The last ‘big’ downturn was the GFC in the late 2000s, when many senior marketers were still relatively junior. A lot of today’s marketing leaders won’t have had to face shrinking markets or mass trading down caused by recessions.
Those who occupy the middle ground of average brands who don’t have enough equity to have a wider price elasticity are likely to be worst hit. The product isn’t distinctive or different enough to prevent trading down, but it can’t justifies price rises that bring it on par with perceived superior competitors.
I once heard a talk where the marketer suggested that from time to time, people working in the industry should go through their house, make note of all the brands, and then ask themselves in brutal honesty how many they’d miss if the brand disappeared tomorrow.
That strikes me as excellent advice, with a follow on that if your brand is in that very disposable category, ask yourself what can be done to bring it a little bit closer to surviving the cull.
How will the loss of cookies combined with privacy changes affect marketers?
There are two schools of thought when it comes to the impending cookie and privacy changes. The first is that it will destroy digital marketing and that marketers should spend the coming months doing all the can to minimise the impact (usually with a SaaS tool or consultancy being hawked by the person with the hot take in mind).
The second, less prevailing view - at least if my LinkedIn is anything to go by - is that it removes a lot of distractions for marketers and that the industry can remove its obsession with personalised, targeted advertising and get back to driving penetration and salience.
Interestingly, marketing scientist Rikard Wiberg ran a recent experiment where spend within Meta was moved to be optimised for reach instead of conversions. Unsurprisingly, conversions dropped but their MMM revealed that incremental sales increased when campaigns were bought by reach.
Wiberg’s experiment illustrates the issues with hyper targeted media. Last touch attribution for conversions is easy to understand, easy to attribute and comforting in terms of immediate return on investment. But how many conversions would have clicked and converted anyway regardless of being served the advert? And how many potential sales are lost by overly narrow targeting?
This isn’t to say targeted digital advertising isn’t effective. Wiberg also notes that targeting tends to work better the smaller the budget.
Retargeting or “high-intent” targeting can be useful in nudging those who need a final push to get over the line (although it’s arguable if it’s cheap conversions as many customers will have been exposed to the brand elsewhere), while there will always be your impulse shoppers - the digital equivalent of walking into a shop you’d not intended to visit after spotting a pair of shoes you like in the window.
The question here is less about targeting and data as it is about multi-touch attribution and brand size.
Digital targeting through third party data can be murky, but it also will have helped level a little bit of the playing field for smaller brands or strangely named ecommerce sites. Creating a good shop window in a high footfall part of town for the impulse or close-to-buying customer became a lot more cost effective.
And, as ever, Byron Sharp’s rule of big brands having an advantage in growth simply because they’re big holds true here. The more things change the more they stay the same.
What does it mean when everything looks the same?
The Guardian recently published a piece looking at the common aesthetic of coffee shops: neutral colours, Instagram-friendly dishes, a hint of industrial decor. It works on one level. Like an independent, decentralised McDonalds, you can be reasonably sure what you’re going to get. Is it a problem that this industry defaults to the mean?
The same question can be asked of creativity, from music to writing, and especially marketing. It’s long been a challenge on social media that the majority of brands stick to the same template of mimicking what’s popular. Somebody should keep a running tally of adverts that use the tagline “[x] your way” or variations on “you do you”.
SEO specialists have been around long enough to create copy that all somehow sounds the same (look at your competitor’s copy for some of your key ranking terms and try to spot the difference).
AI has the potential to enhance creativity, but the majority of images generated or copy still sounds or looks the same, which makes sense as LLMs are essentially the next best guess and very few people are genuinely good at writing creative prompts.
Does this matter? In some respects, largely not. A person searching for information on air fryers or jewellery insurance is hardly likely to care if the result in #3 reads the same as the top ranking page in Google.
A lot of low level use of AI is to fill content rather than showcase creativity. Organic social media is hardly an effective channel for a lot of brands anyway, so trying to game the algorithm at the expense of brand identity is often neither here nor there.
Regression to the creative mean is, if anything, probably going to get more pronounced in 2024. Why risk failure when the playbook shows you how to get rewarded for mediocrity? If people don’t care about brands, by and large, they’re unlikely to care about the sea of sameness given how easy it is to tune out.
But this also means the rewards for standing out and simply being a bit different are greater. As is the possibility of failure. Really, it comes down to how risk averse are board and marketers. Nobody will get sacked for saving a designer’s wage by using AI but somebody may lose their job for a bold idea that doesn’t work.
Again, does it matter? The answer probably depends how you judge success and interpret the data.
No links to other stuff in this issue.
I’ve taken ages to write this particular newsletter, for a variety of reasons, and it’s quite long, so let’s just end with a song.
Playing us out this week is Bill Ryder-Jones’ This Can’t Go On. It’s a bit early to say his new LP Iechyd Da is the album of the year, but it’s s strong contender from the former member of The Coral. It’s full of wistful, dreamlike, melodic folk and psychedelic tinged songs, often touching on Ryder-Jones’ mental health struggles. If you’re a fan of The Coral, Super Furry Animals, and 60s artists like Scott MacKenzie, you’ll probably enjoy this.