Data-driven PR still needs a human reality check

In June, Lululemon organised a yoga festival on the Great Wall of China, which was attended by more than 2,000 guests. The brand had enlisting a Chinese celebrity and designed the PR event to celebrate Chinese culture.

But within 24 hours of the first images going live, the campaign fell through. Several users called out the ceremonial drum on stage as a Japanese taiko instrument leading to boycott calls. Lululemon had to issue a public apology, removed all content from its platforms and stated that it would adopt “a more rigorous attitude” in future.

No sentiment analysis had predicted the outrage and no algorithm had flagged it. The crisis was the result of a cultural blindspot that machines were not trained to see.

The illusion of measurement

This is but one example of the challenges that the PR and communications industry faces today. With the advent of tech permeating every level, it has convinced itself that data is the same as understanding.

As clients increasingly demand dashboards, sentiment scores and impression counts, agencies, in a bid to to demonstrate ROI, resort to the language of data analytics. The result is that vanity metrics have replaced genuine insight.

When it comes to media monitoring and sentiment analysis, reports say that in controlled laboratory conditions, advanced models report accuracy rates above 90 per cent.

But in the real world that figure drops to between 75 per cent and 88 per cent, depending on the task. And that is because personal biases like sarcasm, irony, emotions and cultural subtext are overlooked by the machines.

There is also a growing shift in how users interact with the platforms themselves. A recent survey by Brave Bison and MTM found that 94 per cent of regular social media users have taken steps to manage the content they are shown.

Instagram and Threads have responded with user-controlled algorithm features such as the Threads ‘Dear Algo’ function allowing users to curate their own feeds.

As audiences increasingly shape their own information environments, the idea that a brand can rely on algorithmic reading of ‘the conversation’ is becoming less defensible.

The MENA dimension

The stakes are higher when we bring this to our regional context. The MENA region is not one cultural, social or economic monolith. The 400 plus million people, while united by a shared language and culture, still have inherent differences in outlook shaped by socio-economic, historic and cultural realities. When marketers embark on campaigns in the region, we know that one size doesn’t fit all.

So when it comes to campaign execution and measuring the impact, having a nuanced understanding of the market becomes even more relevant. That is where homegrown agencies make a difference; they come with a stronger cultural context of the region that no machine can instantly showcase.

In fact, the Arabic language itself presents a specific technical challenge that is well-documented: the vast diversity of dialects across the Arab world make automated sentiment analysis significantly less reliable than it is for English.

An algorithm trained predominantly on English-language data will not understand or catch the difference when scanning Arabic media output. This is where homegrown agencies make a difference.

A PR and communications professional who grew up in the region, who understands its history and culture, is well placed to identify the nuances and map a real sentiment analysis.

From vanity metrics to genuine influence in PR

None of this is an argument or case against data. Sentiment tools are no doubt useful for processing volume, identifying broad trends and potential crises or challenges that demand closer attention. The problem is not the tools; it is the mindset and misplaced confidence that the tools are sufficient for PR.

The Lululemon incident becomes relevant because it was not caused by any dashboard failure. The brand had access to the same planning, monitoring and listening platforms as everyone else. But probably what it lacked was someone in the room with the cultural knowledge to look at that Japanese drum and understand what it would mean to a Chinese audience.

Boasting about four billion impressions can tell a client how far a message travelled. But it says nothing about whether it landed, what it meant to the people who received it or what it will cost to repair if it was misread.

The brands that will build lasting influence in this region are those that treat data as the starting point of analysis, not its conclusion. The winners are those who invest in the cultural fluency, senior judgement and regional expertise that no algorithm can replicate.