Perhaps https://justfacts.votesmart.org/ is what you’re looking for?
I think this is a quite hard problem, perhaps much harder that is seems initially. Besides the fact that different voters have different opinions and priorities about policy (or character, for that matter), it is for the most part not the case that serious candidates lay out their policy prescriptions in detailed/possible ways. That’s because doing so tends to make oneself less electable. The way to get into office is to keep things vague enough that any particular voter could convince themselves that the candidate supports, or at least might be persuaded later to support, whatever the voter is looking for. And if you keep things vague then you can present an impossible platform, such as promising funding for particular programs without specifying what unpopular policies (taxes or budget cuts) would produce that funding. Furthermore electorates change their mind and a successful politician needs to be able to follow the crowd without being accused of flip-flopping. Finally, candidates also want to be able to tailor their messaging to different constituencies or donor groups.
The bottom line is that candidates and political parties (at least, serious ones) aren’t going to subject themselves to a GiveWell-like process, at least not without a really significant and legible benefit (votes or $$) available on the other end.
When it comes to local or regional elections, the problem is even harder because the issues vary from place to place and office to office, and information about candidates is not widely disseminated, often available only in the local paper or candidate forum.
Is it possible that organizations are privately deliberative, using a scout mindset to identify the best approach, but then publicly confident, projecting a soldier mindset to focus action, attention, funding, volunteers, etc., on the chosen model?
Social and political action are quite different from international aid funding, in that they are naturally adversarial. In such an environment it might not be best to project your uncertainties or vulnerabilities.