It's easy to get an image of (say) Haiti right _after_ the earthquake - you know the disaster just happened. But it's really nice to have an image of Haiti right _before_ the earthquake so you know where the damage happened, so you want baseline imagery of every city in the world to better direct post-disaster relief efforts. The more up-to-date, the better.
It's good for identifying areas where illegal logging operations are happening _right now_ - Count the trees in at-risk areas on a daily basis to track illegal logging groups.
Or image every port on the planet to identify international shipping trends and monitor loading / unloading efficiencies.
Or image every soybean plantation in the world to check crop health. [insert crop type above for the ag industry you're interested in]
Or image every major glacier every day and correlate it with temperature data to gather more scientific data about climate change and sea level predictions.
All these are simultaneously achievable with a flock of >100 3U cubesats taking ~4m resolution imagery. They are not possible with a half dozen ~1M resolution conventional satellites... Maybe you could do one of the above things.
There's a lot of value there - the more conventional satellites add to it, for sure, but they're not going to totally replace it anytime soon.
What they do get you is a cheap payload that maybe is "good enough" for certain applications. The issue is Planet is realizing they are not "good enough" for mapping applications.
altitudewavelength/aperture = pixel size
Planet's cubesats are at ~500km and visible light is ~500nm, and cubesats are at most 10cm per unit, so resolution is just:
5e55e-7/1e-1= 2.5 (all in meters).
There are some games you can play with super high contrast to improve on that, but it's pretty close to a hard limit. So to improve, you need a larger cubesat platform, like perhaps a 12-unit cubesat for up to 20cm aperture (1.25m resolution) or even a 27-u cubesat for 30cm aperture, but at that point, you're talking about a custom cubesat launcher (although the 12u one isn't too bad), and you might as well just develop a different platform that isn't constrained so much by the cubesat standard.
Which makes the Terra Bella merger make sense.
The optimum size of a cheap, disposable (6-12 month lifespan until deorbit due to the extremely low altitude) mass constellation of tiny Earth observation satellites is probably a little bit larger than the cubesat platform. And if you're launching them often enough (and Planet is, since they make up a huge proportion of the total number of cubesats ever launched... I think the majority?), you can afford to develop your own standard.
ISS's ability to launch cubesats is something that's fairly easy to upgrade and has been upgraded in the past. And other launchers for larger satellites from ISS have been used. And, of course, the same thing applies to secondaries on other rockets or even dedicated launches.
(4kg/satellite)*(50 satellites) is only 200kg. You could increase that to 10kg per satellite for a larger platform and you're still at just half a ton, which is pretty small for a secondary payload still, with the potential for sub-meter resolution, so I definitely see an upgrade beyond the 3U cubesat in Planet's future.
[1] There's a factor of 1.22 in there for circular apertures for the usual definition of resolving power (i.e. overlapping of the minimum of the Airy's disc of one point source with the maximum of the other... this being a case where you're trying to resolve two bright sources from each other). But again, this is for bright sources and there's assumptions about contrast ratio when discussing resolving power, so generally speaking, I just use "1" instead of "1.22", since "1" is a little more conservative and relevant to more than just two bright sources next to each other with a dark background.