Choosing a political platform is an area of political judo. The candidate who has the most effect on it is the presidential candidate. Because the presidential candidate has to garner enough votes to win a majority of the delegates, the presidential candidate nominally has enough votes to control the platform. But in practice, other groups have influence over the delegates and may be able to convince them to put in planks that are stronger or weaker than the candidate would have chosen.
It might not seem to matter, but candidates can be held responsible for planks in the platform. If candidates do not support a plank, it hurts them twice. First, the party base may not appreciate the lack of support for the party position and withhold their active support. Second, swing voters may not believe them. They may believe that the candidate is implicitly supporting the platform by running as the party. Or that support for the party is support for the party leadership, which does support the platform.
You could see this happen in 2010 and 2014, when candidates who at least nominally fit their districts retired (e.g. Heath Shuler) or failed to win reelection (e.g. Walt Minnick) in the face of candidates who were in the right party. This is why Nancy Pelosi's popularity (or lack thereof) matters.
Beyond all that, it is harder for candidates to oppose their president in things that are on the platform. This makes it easier for presidents to rally support for things that have already been tested in their presidential campaign. For example, Democrats found sixty votes in the Senate for healthcare reform in 2010, even though several Senators lost their jobs as a result.
Party platforms are themselves marketing documents. They signal what the party finds important. There are still people who believe that the Republicans' 1994 platform, the "Contract with America", was a big part of why Republicans won that year.