Reflections from P&G: How to Build Insights

The term “Insight” gets thrown about a lot. But many, including myself, use the term wrongly. After my stint at P&G, I’d walked away with a better understanding of what an Insight is. This post shares my learnings, and what we can do to build better insights.

Why Insights Matter

Insights are important for a variety of reasons:

• Insights allow us to really understand the consumer, customer or whoever we’re serving.

• Insights allow us to work on the right solutions and make the right decisions.

What is not an Insight

I’ll begin by sharing what aren’t Insights.

• Observations: Direct observations aren't insights. Rather, they are clues that could lead to an insight.

  • E.g. A stomachache itself is not an insight. It’s a symptom of a deeper issue, which could potentially be the insight.

• Data: Data communicate observations. They aren’t insights.

  • E.g. Higher revenues for Product A vs. Product B simply means consumers are buying more A than B. Insights are why consumers buy A over B.

• Intuition: Intuition or ‘gut feel’ is subjective. Again, it could potentially lead to an insight.

  • E.g. We receive more revenue from Gen Zs when we advertise more on Instagram than TV, because Gen Zs likely use Instagram more than TV. This sounds like an insight, but is a subjective (albeit logical) opinion.

• Logical Connections/Statements: Deductive logic statements aren't insights. I used to mistake them for insights. But I realise that while all insights must be logical, not all logical statements must be an insight.

So what exactly is an insight?

What is an Insight

An insight is an ‘AHA!’ moment that must be built. Insights are formed by piecing together observations, data and intuition across different perspectives. In practice, insights act as a logical response to an issue, which in turn helps build a solid recommendation.

To build an insight, a deep understanding of the issue, target audience and (data) context is critical, in order to piece together the right dots.

This sounds abstract. I personally found this hard to grasp, until after reflecting on my previous experiences. So below, I’ll share what I mean from my own examples and reflection. Like many, I’d previously misunderstood what insights were.

Insights must be synthesised; they are not obvious

I’ll begin with an example to illustrate what an insight is. I use this when tutoring IB students (17-18 year olds) so forgive me for its simplicity and lack of data. For a business-related example, hop to the next point/example.

Example 1: Lunch Options

Observation: You go for lunch with your colleague. The menu has 4 items.

1. Beef Fried Rice ($10)

2. Tom Yam Soup ($7)

3. Claypot Rice ($12)

4. Mutton Curry ($9)

You know that your colleague is usually thrifty and spends minimally on food. He also has an important meeting the next morning. But for lunch, he orders the Beef Fried Rice for $10. Why?

There could be various reasons:

• He likes beef over soup and mutton.

• He’s craving fried rice.

• He wants to treat himself before the meeting.

But in this case, he shares that he doesn’t want to risk eating spicy food right before the important meeting.

As a result, the insight is that your colleague generally doesn’t react well to spicy food.

This illustrates what insights are:

• Insights are not immediately obvious.

• Insights are usually deeper than what is explicitly said/conveyed.

• Insights change based on specific contexts.

Insights are ‘AHA!’ moments that must be synthesised and pieced together.

Insights must be validated by data

Insights are more credible and impactful when validated by data. This is obvious.

What’s not obvious, however, is how to correctly use data. Knowing how to correctly use data is even more important than just using data to back an insight up.

Previously, I’ve often received feedback that I appeared to be force-fitting data to support my point. I strongly suspect this stems from the ‘PEEL’ method we are taught as students, where we write the topic/main sentence, followed by evidence to support the topic sentence.

Now, I’ve learnt that there’s a fine line between using data to validate my insight, instead of using data to support my insight.

This example illustrates the difference.

Example 2: Declining Revenue

Problem: Brand A revenue is falling, while competitors’ revenue is increasing. Why?

Before

In this example, I used data (ad spend) to support my observation and insight. While this finding is logical, it isn’t an insight. Nor is it a correct use of data to back the insight up. Why?

• It rehashes the observation.

• It’s based on (logical) intuition and gut feel.

• Other factors beyond ad spend could have caused revenue decline.

• Decisions can’t be made based on this. (So… do we just increase ad spend to drive revenue? Wouldn’t this incur higher costs?)

This shows how logical statements are not always insights!

After

The recommendation would then be to increase TV + Social impressions by doing A, B, C.

*I’ll write a separate post on how to create recommendations. 

Now, I’ve used data to validate my hypotheses, which logically synthesise into an actionable insight. Decisions can be made from here.

Insights must be built in context

Data is important to validate hypotheses and build insights. But I realise that the context of data is even more critical than just using data itself. Why?

• Context enables understanding of real consumer/customer/user behaviour

• Context-driven insights result in sharper and more relevant recommendations for the right setting

To verify context, question the (1) degree/strength, (2) scale/extent and (3) timeframe of data.

This example illustrates what I mean.

Example 3: Social Impressions and Market Share

Observation: In Jan’22 - Dec’22, Social Media Impressions correlates strongly with Brand A’s Market Share.

(1) Degree/Strength

• What is the degree of correlation? (r = ?) Is it strong?

(2) Scale/Extent

• Is this observation/behaviour a one-off or recurring observation?

• Is this applicable just to Brand A, 1 group of Brand A’s competitors or across all Brands in the category, industry or consumer segment? (1 data point, 1 cluster of data points or across the sample population?)

(3) Timeframe

• Is this applicable/observed at just 1 moment in time? Or is this observation evident across multiple moments in time?

These questions naturally lead to deeper ‘Whys’ to build the insight. But that’s beyond the scope of this section on Context.

How to Build Insights: Actionable Takeaways

So, how could you and I build better insights? By consistently practising these few steps.

• Observe deeply.

• Keep asking why, why, why.

  • Recognise context - different contexts lead to different behaviours.

• Validate with data and research.

  • Develop tool proficiency to validate this faster