Probably you might agree visualizing value of a feature vs. work spent would look something likes this…
In this circumstance, it makes perfect sense to release the feature in a windows where value can be maximized:
But there is one thing missing… If we intend to scale the feature over a large population the expected return is likely an exponential curve where at a certain break point, the feature have gathered enough momentum to achieve viral spread in the crowd. If we add this curve on top of the others;
We now see that releasing the feature according to our normal cadence will short circuit our economic payoff and minimize our return.
With this in mind, I advocate for not treating all your features the same. Many will make use from being part of a regular release cycle, but if we are interested in maximizing our economic return, not all.