Search assumes you already know what you are looking for. In deal making, that is the part you do not know.
Every product review eventually asks for it. Where is the search? People want to type. A box is familiar, it demos in three seconds, and it makes a product feel like it respects your intelligence. Leaving it out looks like an oversight, or worse, like arrogance.
So it is worth explaining, because the reasoning is the product.
Picture the query you would actually type: fintech investor London Series A. You get back what matches those words. You do not get the fund that describes itself as financial technology. You do not get the partner who has stopped writing at Series A this year and has not updated anything to say so. You do not get the operator who has never been called an investor by anybody and would write your first cheque tomorrow.
What you get is a tidy list of what you already knew to ask for, and no indication of the size of what you missed. That last part is the real cost. A search that returns nine results feels complete. Nothing on the page is willing to tell you it should have been forty.
A query returns what matches your words. Matching returns who is ready, whether or not you would have thought to type them.
Underneath, the difference is not clever ranking. It is that the engine does not work in words at all. It works from a fixed vocabulary: 66 business concepts across nine dimensions. Fintech and financial technology are one concept rather than two strings that happen to mean the same thing, so the question of which phrase you typed never arises.
It also works from what people want rather than only from what they are. A search for investor returns people with the word in their title. Matching returns people who are deploying, because raising and deploying are a pair, and being a pair is a fact about intent rather than about vocabulary.
Now the honest part, because a post about a design decision that does not name what it cost is marketing.
You give up the feeling of control. With a box, you steer: you type, you refine, you feel the system responding to you. With matching, you are shown, and being shown can feel like being managed, especially in the first week, and especially if what you are shown is short. That is a real loss and we are not going to pretend it is secretly a gain.
What we would say is what the control was actually made of. The steering a search box gives you is steering within your own blind spots. You can only type what you already thought of, and the thing you most need in a deal is usually the thing you did not think of, or thought of too late.
There is a version of this that would have been easy and dishonest: ship the box, put matching behind it, and let people feel like they are driving. It performs well in a demo. It also trains people to trust a list whose completeness nobody can vouch for, which is the same failure we spend a whole other post on.
The mornings are the compromise we did make. At most eight things, with the reason each one surfaced attached, and a cap rather than a target, because twenty things to know is not a brief, it is a backlog. Eight things you can argue with is a better interface than a box that will always return something.
None of this is an argument against retrieval in general. Finding a document you know exists is a genuine problem with a genuine answer, and that answer is a box. Finding the person who is ready this week is not retrieval, and dressing it up as retrieval is how you end up with a product that is pleasant to use and quietly useless.
We would rather show you eight things and be able to defend all eight.
