- "Is it necessary to decompose a multivector into a multiplication of two other multivectors?"
No, and as you suggest, decomposition of a multivector in terms of basis vectors, is a straightforward way to compute the reverse.
Take for example
$$ A = a + b \mathbf{e}_1 + c \mathbf{e}_2 \mathbf{e}_3 \mathbf{e}_4,$$
where $ a, b, c$ are scalars, and $\mathbf{e}_k$ are all mutually orthogonal. Reversion can be performed term by term, resulting in a sign flip for the trivector term.
$$ \tilde{A} = a + b \mathbf{e}_1 - c \mathbf{e}_2 \mathbf{e}_3 \mathbf{e}_4.$$
- For your bivector example $ A = \mathbf{n} \mathbf{m} $,
first note that such a factorization may not
generally be possible. For example,
an $\mathbb{R}^{4}$ bivector such as $ A = \mathbf{e}_{12} + \mathbf{e}_{34} $ cannot be decomposed into a product of two vectors (this is an example of a bivector that
is not a 2-blade.)
If the bivector has such a factorization, that decomposition amounts to a choice of basis.
For example:
$$A = \mathbf{e}_{12} + \mathbf{e}_{23} + \mathbf{e}_{31},$$
may be factored as
$$A = \left( { \mathbf{e}_1 + \mathbf{e}_2 - 2 \mathbf{e}_3 } \right) \frac{ \mathbf{e}_2 - \mathbf{e}_1 }{2},$$
or
$$A = \frac{ \mathbf{e}_3 - \mathbf{e}_2 }{2} \left( { 2 \mathbf{e}_1 - \mathbf{e}_2 - \mathbf{e}_3 } \right).$$
We may compute the reverse from the original representation
$$ \tilde{A} = \mathbf{e}_{21} + \mathbf{e}_{32} + \mathbf{e}_{13},$$
or using the first factorization
$$\begin{aligned}\tilde{A} &= \frac{ \mathbf{e}_2 - \mathbf{e}_1 }{2}\left( { \mathbf{e}_1 + \mathbf{e}_2 - 2 \mathbf{e}_3 } \right) \\ &=\frac{1}{{2}} \left( { \mathbf{e}_{21} - \mathbf{e}_{11} + \mathbf{e}_{22} - \mathbf{e}_{12} - 2 \mathbf{e}_{23} + 2 \mathbf{e}_{13} } \right) \\ &= \mathbf{e}_{21} + \mathbf{e}_{32} + \mathbf{e}_{13},\end{aligned}$$
or using the second factorization
$$\begin{aligned} \tilde{A} &= \left( { 2 \mathbf{e}_1 - \mathbf{e}_2 - \mathbf{e}_3 } \right)\frac{ \mathbf{e}_3 - \mathbf{e}_2 }{2} \\ &= \frac{1}{{2}} \left( { 2 \mathbf{e}_{13} - 2 \mathbf{e}_{12} - \mathbf{e}_{23} + \mathbf{e}_{22} - \mathbf{e}_{33} + \mathbf{e}_{32}} \right) \\ &= \mathbf{e}_{21} + \mathbf{e}_{32} + \mathbf{e}_{13}.\end{aligned}$$
One should get the same answer regardless (and we do.)
Also observe that such a 2-blade factorization depends on the orthogonality of the factors. If that were not the case, then there would be a scalar term in the product, and the result would not be a 2-blade. To see that, consider the following product of vectors
$$A = \left( { \mathbf{e}_1 e^{i \theta} } \right) \mathbf{e}_1 = \cos\theta - i \sin\theta$$
where $\{\mathbf{e}_{1}, \mathbf{e}_{2}\} $ is assumed to be an orthonormal Euclidean basis, and $ i = \mathbf{e}_{1} \mathbf{e}_{2} $ is the pseudoscalar for the plane that $ A $ lies in (i.e. $ A \wedge i = 0 $). For this to be a bivector, we require $ \theta \in \pi/2 + n \pi $. That is, the two factors must be orthogonal.
If one wanted to prove that the reverse of any bivector $ A = \mathbf{n} \mathbf{m} $ is independent of the factorization (if such a factorization is possible), the task is essentially to show that the reverse is independent of any change of basis.
$\widetilde{A}= \tilde{m}\tilde{n}$
– Peeter Joot Aug 10 '21 at 13:15