"Microsoft Was There" Is Not Intelligence
A large enterprise is not one company. It is a dozen divisions with separate budgets, separate strategies, separate competitors and separate reasons to be in a room. Each has its own view of the market and does not necessarily know what the others are doing.
So when a competitive intelligence system reports that a large company was present at an event, it has told you almost nothing. Present how? Which part? Selling what, to whom?
Why this matters more than it sounds
Consider a hyperscaler at three events in a quarter: a manufacturing trade fair, a healthcare congress, a developer conference. Those are three different teams pursuing three different markets, and reading them as one company's "event strategy" produces a description of something that does not exist.
Now consider the reverse, which is the case that actually costs money. You compete with one division of a large company. You want to know whether they are moving into your category. But their aggregate participation is dominated by unrelated divisions, so the signal you care about is buried in noise generated by teams you will never encounter.
An aggregate number across a large enterprise is not a summary. It is a average of unrelated things.
What the data usually gives us
Participation records typically name the company, not the division. An exhibitor list says the parent brand. So the division has to be inferred from context — the event's category, the co-participants, the products shown, the role taken — rather than read off a field.
That inference is doable and it is not free. It requires judgement, it will sometimes be wrong, and at scale it costs real money to compute. Which means honesty about two things.
First: where we cannot attribute a division, we say so rather than quietly assigning one. A confident wrong attribution is worse than an acknowledged gap, because it corrupts every comparison built on it.
Second: deep division resolution is not free, so it is offered as premium work rather than promised universally. Claiming to do it for every company at every event would be a claim we could not honour at the price.
What we do universally is build for it: the data model carries division attribution with a confidence level from the start, so that resolving a company properly later is a matter of running the work rather than rebuilding the schema.
What it looks like when it works
You see a large competitor's presence split by the part of the business that showed up. Their industrial division has been at the same four events for six years. Their cloud division entered your category eighteen months ago and has appeared at three events since, twice with a speaking slot.
The second fact is the one you needed, and it is invisible in the aggregate. Combined, they would have produced a slightly elevated overall number that meant nothing in particular.
The general principle
Any measure computed at the wrong level of aggregation produces a number that is precise and useless. Large companies are the most common place this happens, because they are the ones most likely to be doing several unrelated things at once — and the ones whose behaviour you most want to read.
Calculate your Share of Presence at whr.ai.