The key to a great observation is knowing what to look for and how to organize the intel for it to make sense. Throughout my career, I've stumbled upon several different tools and frameworks. Some are useful. None of them do the actual observing for you.
They organize what I've already sensed and heard, they don't generate it. The real insight comes from sitting in the room, asking the uncomfortable question, and noticing what people avoid saying. The frameworks just help me sort it so it makes sense to others too. So before we get into the toolbox, know that it's the smaller half of the story.
The entry-point tools
A few frameworks earn a permanent spot in my process, mostly because skipping them has bitten me before.
Supply chain analysis. Porter's 5 forces covers the players closest to you. Mapping the full supply chain, start to finish, catches the dependencies further out. The AI unicorns leaning on a handful of suppliers are a good reminder that this applies far beyond physical goods.
Customer profiles. I skip personas, they say too little about actual behavior. Customer Jobs to be Done and Value Proposition Canvas do more work, because they start with needs and pain instead of demographics. Paired with a tiered ICP (I use three), so you're not trying to be everything to everyone.
PESTLE. My favorite lens on the world outside the business. The value isn't in listing macro factors, it's in tying each one directly to your business and your customers.
McKinsey 7S. Useful for breaking an organization into its hard and soft parts. Where it falls short is governance, it doesn't ask whether the owners are clear on their ambitions, or whether management actually has the mandate to act on the strategy. I've written more on that gap here.
That's the toolbox. Any of them can be picked up in an afternoon. What actually separates one observation from another happens below.
What I actually spend my time on
Not everything I observe comes with a name or a template. Most of it is discipline, repeated the same way every time, regardless of the client or the industry.
Numbers
I don't jump straight to conclusions when I open a P&L. I start with the raw data itself, no filters, no interpretation. Then the history behind it: how we got here, what changed and when. Only once I understand the trajectory do I look for deviations, the numbers that don't behave the way the rest of the pattern would predict. And once I find one, I challenge the assumption behind it, because a deviation is rarely explained by the first answer you're given.
The first time I ran this discipline all the way through, it took six months. A complete Business Impact Analysis, on a product the organization hadn't attempted in 50 years. Thorough almost to a fault, and ten years later I still hear it's the document people pull out before launching anything new. Flattering, and also a little concerning, because I'd seen the right answer by the end of month one. The extra five months didn't change the recommendation. They just made me feel safer making it. I know better now: the discipline is for finding the deviation, not for building a monument to it.
Behaviors
Behaviors are best observed live. I favor sitting in on meetings where I can stay in the background, alongside interviews and conversations where I can watch the dynamic between people, not just what's said. From there, a layer deeper: what was promised versus what actually happened. How decisions get made in the room. What gets followed up on, and what quietly drifts. What gets noticed, and what doesn't.
Decisions
I look at how open the dialogue actually was. How clearly the decision itself was formulated. Whether it has an owner, and whether there's a clear next step attached to it. If a decision has stalled, I trace back to find exactly where it got stuck, and why.
A final word of caution
Getting stuck in analysis paralysis is a real danger, finding one more piece of information, adding it, even though it barely moves the recommendation. Many failing companies get stuck exactly there, eager to find one more piece of the puzzle before getting to work.
My recommendation is simple. Cover the ground above, then ask yourself: "What information would force me to re-evaluate my observation?" And go look for that, and nothing more.