Mental Math and Estimation is about fluency and numbers sense. Fluency is when students have had enough practice to achieve automaticity, or know basic facts like multiplication without having to apply strategies. More important than memorization is the understanding of the concept and the ability to apply efficient strategies.
Number talks are the best way to promote fluency, because if we only increase the practice, student will never be nudged to using more efficient strategies. There's an old saying: If a student is counting on their fingers past the grade levels where that is appropriate, and all we do is increase drill and practice, the student will get faster.....at counting on their fingers.
It is important for students to represent various strategies. A strategy is the method, a model is how we draw that. See the model of the number line at right. The strategy is counting back in chunks, the model is the open number line.
Consider using small assessments, like entrance/exit slips, that require students to show one or two specific strategies. Use posters (anchor charts) in the classroom to help students talk about and name the strategies they are using.
"Mad minutes" or other bell work, especially used as spaced practice, can be effective if they are not timed. While some students enjoy the challenge of timed tasks, others develop math anxiety that can erode their self efficacy and have a negative impact on their attitude toward math anxiety. Students that really enjoy being timed can compete against their own time, but there should not be classwide timing or time limits. See Growth Mindset?Rich and Safe Mathematical Environment.
Estimation is two-fold. In One sense, it is the ability to look at a collection of objects and have a sense of how many there are, how long something is, etc.
There are some fun web sites that support estimation:
Estimation is also using number sense to make a judgement about the feasibility of a solution. This is a core principal of mathematics, and we need to be pausing continuously to take a step back from procedural problems and ask, "is this answer reasonable?". Using a strategy of predicting an answer "too low" and "too high" before doing a calculation allows entry points for many learners while promoting estimation of an expected solution.