Wind observations are important on the smaller scales and in the tropics for forecasting the weather. Several sources of high-resolution winds exist that may be exploited in deterministic NWP analyses. Scatterometer winds do verify well with local buoy wind measurements and show spatial wind spectra extended well to the small scales. Similarly, high-resolution radiosonde data exists, depicting the wind shear in the vertical in detail. The beneficial impacts in NWP of scatterometer and Doppler Wind Lidar wind profiles will be illustrated. However, it turns out that NWP analyses that verify well, like that of ECMWF, do not describe the tail of the horizontal wind spectrum well and rather appear in approximate balance with a rather dissipative NWP model on scales of a few hundred kilometers. The observed high-resolution wind information is thus apparently not well used in NWP analyses. Similarly, the vertical shear of the horizontal wind is generally strongly underestimated in global NWP models, as verified by high-resolution radiosonde measurements, while such shear poses one of the basic atmospheric instability mechanisms (e.g., in the Eady growth rate). Performing small-scale analysis appears in conflict with medium-range weather forecasting. How can a better wind observing system help to improve small-scale analyses and weather forecast skill?