Dan Saffer, who wrote the definitive book on the subject, defines a microinteraction as having four parts: a trigger (what initiates it), rules (what happens), feedback (what communicates what is happening), and loops and modes (how it behaves over time). The trigger can be user-initiated (pressing a button) or system-initiated (a new message arriving). The feedback is almost always the design challenge: how does the system communicate what it did, without interrupting the flow of the larger task?
The most effective microinteractions are the ones users never consciously notice. The subtle spring animation when a toggle switches, the brief scale pulse when a like is tapped, the smooth colour transition when a form field validates — these do not register as designed moments. They register as the product "feeling right." Their absence, however, is felt: interfaces without microinteractions feel inert and unresponsive, even when they are technically functional.
Microinteractions are expensive to design well and easy to design badly. An overlong animation on a frequently-performed action (like a send button) becomes a tax on efficiency — the user has to wait for the product to finish its celebration before they can continue working. The discipline is calibrating the feedback to the significance of the action: small actions deserve subtle feedback; meaningful actions deserve meaningful feedback.