I think you have to start by defining what 'credible' means for your specific project. If you're looking at literary analysis or tracking publishing trends, you can't just rely on a single type of source. For historical context or reception, nothing beats primary material like author manuscripts, letters, and contemporary reviews. University press publications and peer-reviewed journals in literature or media studies are your bedrock for scholarly arguments.
But a lot of modern novel research, especially on genre fiction or digital publishing, lives outside traditional academia. Industry reports from places like WordsRated or the Author Earnings datasets, while not peer-reviewed, offer hard numbers on sales and readership patterns that you just can't get elsewhere. I cross-reference those with data from platform blogs like Kindle Direct Publishing's or Radish's insights. The trick is transparency—checking where they got their data and whether they have a commercial bias.
For fan communities and impact studies, archived forum threads or subreddit analyses can be surprisingly robust qualitative sources, showing how interpretations form in real time. It's a patchwork method, really, where credibility comes from triangulating between the ivory tower, the industry's spreadsheets, and the readers' own words.