This week: Meta’s AI gets cleverer, loyalty programs admit they trained customers to chase discounts, marketers are told to put rules around AI before it runs the meeting, A/B testing reminds us evidence still beats opinions, and analytics finally stops forcing a choice between depth and scale. Big picture: the tools are getting smarter, but common sense is still doing most of the heavy lifting.
Meta’s AI Gets Smarter (Still Not Doing Your Job)
Meta has outlined a round of AI-driven system changes across Facebook, Instagram and Threads, aimed at improving content discovery and ad performance. The update claims stronger engagement with original content and better matching between users and what appears in their feeds. On the advertising side, Meta is also refining attribution models to show incremental conversions more clearly. Video continues to be pushed hard, with time spent watching rising in key markets.
For marketers, this is less about new toys and more about what the machine is learning to reward. If the system is trained to surface what people engage with, then bland or lazy creative is going to sink faster than ever. There is also a reporting wrinkle here: when attribution models change, results can look better or worse without anything else actually changing. That is not performance, that is maths with a haircut.
The bigger issue is control. Meta’s AI will happily optimise toward whatever signal you feed it. If that signal is weak or misaligned with business goals, you will get very efficient nonsense. This is where strategy still matters, even if the platform pretends it does not.
This connects directly to our Digital Marketing Strategy and Paid Media work at Belmore Digital. If your reporting or optimisation settings have not been reviewed recently, now would be a good time. If you want a second opinion on how your campaigns are being measured, talk to me about our strategy and paid media services.
Loyalty Didn’t Die — It Was Replaced With Coupons
There is a growing realisation that customer loyalty has not vanished, but has been diluted by years of points schemes and short-term offers. Many programs now reward transactions rather than commitment, which trains customers to shop around instead of stick around. The result is a generation of “loyal” customers who are loyal only to the best deal this week. That is not loyalty, it is price sensitivity with branding.
This happened because activity was easier to measure than attachment. Sign-ups look good in reports, but they do not tell you if someone would choose you without a nudge. Over time, businesses confused engagement with devotion, and discounts became the main relationship tool. It works until someone cheaper turns up.
A better question is what behaviour you actually want to encourage. Do you want repeat visits, referrals, or tolerance for price rises? Those are very different goals, and they need different incentives. If the program design does not match the behaviour you want, it is just admin dressed up as strategy.
This ties closely to our CRM services at Belmore Digital. If your loyalty program mostly exists to look busy, it probably needs a rethink. Talk to me about designing customer programs that reward the right behaviour instead of just issuing points.
Set the Rules Before AI Runs the Meeting
There is a polite warning doing the rounds: if you do not define how AI should behave, it will decide for itself. The distinction between automation and AI is important here. Automation follows instructions. AI looks for patterns and makes suggestions based on what it sees. Without clear rules, those suggestions quickly turn into decisions.
The risk is not that AI makes mistakes. The risk is that it optimises for the wrong thing very efficiently. If the goal is clicks, it will get clicks, even if those clicks are useless. If the goal is speed, it may sacrifice quality. Machines are obedient, not wise.
The fix is dull but effective. Write down what success actually means before switching anything on. Decide where automation can act freely and where humans must step in. That boundary will matter more as systems become more capable and more confident.
This sits squarely within our Virtual CMO work at Belmore Digital. If AI tools are creeping into your marketing stack, they should be working to your rules, not making them up. Talk to me about putting structure around how these tools are used.
A/B Testing: Still Better Than Arguing in Meetings
A recent piece on A/B testing serves as a reminder that controlled experiments are still one of the few ways to prove cause and effect in marketing. By testing two versions of a page or message, teams can see what actually changes behaviour instead of guessing. The benefits are familiar: improved conversion rates, clearer decisions and fewer pet theories being treated as facts.
What often gets skipped is the discipline part. Good testing starts with a hypothesis and ends with learning, not celebration. Most tests will fail or produce small gains, which is the point. You are removing bad ideas one by one, not waiting for a miracle.
Random testing is just noise with spreadsheets. Useful testing links directly to business outcomes such as qualified leads, completed purchases or faster decisions. When that link exists, optimisation becomes part of strategy instead of a side hobby.
This connects directly to our Conversion Rate Optimisation and Analytics services at Belmore Digital. If your site has never been tested in a structured way, there is almost certainly money being left behind. Talk to me about building a testing plan that ties to real business goals.
Insight and Scale Finally Stop Fighting
There is a shift underway in how analytics is done. For years, marketers had to choose between small, detailed research and large, shallow datasets. AI is now making it possible to analyse open-ended feedback and behaviour data together, at scale. In theory, this brings what people say and what they do into the same conversation.
This is useful because it removes one of marketing’s oldest excuses. You no longer have to pick between understanding customers and measuring them. Reviews, comments and survey answers can be processed alongside performance data to show both patterns and reasons. That is far more practical than staring at charts and guessing what they mean.
The catch is interpretation. Automated insight does not understand brand nuance or commercial context. Someone still has to decide what matters and what can be ignored. The machine can summarise, but it cannot prioritise for your business.
This aligns with our Data, Analytics and Customer Insight work at Belmore Digital. If your reporting still lives in separate silos, there is an opportunity to bring behaviour and meaning together. Talk to me about how to make your data more useful without losing the human view.
Annnd That’s a Wrap
AI is getting louder, platforms are getting cleverer, and metrics are getting shinier. None of that fixes weak strategy, lazy creative or fuzzy goals. If this week proves anything, it is that good marketing still comes down to clear rules, proper testing and knowing what behaviour you actually want to change.
Recent Belmore Digital Articles:
You may have missed these articles and be champing at the bit to read them :-)
