I disagree with your input about adding affiliate links, because I think that they would end up working contrary to the goal of this project. Sellers constantly reuse listings to sell updated versions of their products, with variable results. This is a huge issue with Amazon listings.
OP lets you add an affiliate link to order Graeter's ice cream. There's 20 years of 5-star aggregate reviews and a link to go buy it, so you do. But in year 21, Graeter's changed their all-sugar formula to corn syrup, and now you falsely believe that the aggregated five-star rating is supposed to apply to the corn syrup formula.
If one is designing for a lasting product review, the information needs to be snapshotted at a certain point in time- ie "In 2020, this part number from this company was constructed well." Modern listings may reflect the 2020+ product, rather than the 2020 construction that got the good reviews.
> If one is designing for a lasting product review, the information needs to be snapshotted at a certain point in time- ie "In 2020, this part number from this company was constructed well." Modern listings may reflect the 2020+ product, rather than the 2020 construction that got the good reviews.
In theory, that’s what the dates next to Amazon/App Store reviews are for. The problem with making them immutable though is: say you write a 5 star review, but the next day, the device craps out (inside the warranty). You contact the manufacturer who informs you that they can’t help you because it was caused by “user error” or whatever (basically, you can’t return/replace it). So you’d like to amend your review to mention that it crapped out, but you can’t. There also is the point of reviewing the product, not the company, but if a product and company sucks, why not warn the other consumers?
> In theory, that’s what the dates next to Amazon/App Store reviews are for.
There's no way to chart the average review score over time, though, is there? If people saw several years of 4.5-averaged scores and then a sudden drop to 3-average, that would tip them off, but looking at dates next to individual reviews doesn't help because any single review by itself varies wildly with the knowledge and opinion of the reviewer.
As for immutable reviews- that's a good point, although it seems like you could get around that by leaving reviews editable through at least the listed warranty period of the item. Or add a separate set of reviews that only become active after a certain period of time after you left your first review, but now we're getting into anal-retentive nerd territory that no average user is incentivized to venture into. ("Why, of course I'd like to log back on to a review site I used once six years ago and talk about how my washing machine is functioning!")
Edit: Actually, the type of person who would seek out and help populate a "reliability rating site" might be the type of person who would respond with details to a six-year-later email that asks for commentary on a product, especially if you can enter the review straight from an email reply.
Steam also offers a graph of positive/negative reviews over time. But the more prominent recent reviews vs all time reviews usually conveys all the information needed
OP lets you add an affiliate link to order Graeter's ice cream. There's 20 years of 5-star aggregate reviews and a link to go buy it, so you do. But in year 21, Graeter's changed their all-sugar formula to corn syrup, and now you falsely believe that the aggregated five-star rating is supposed to apply to the corn syrup formula.
If one is designing for a lasting product review, the information needs to be snapshotted at a certain point in time- ie "In 2020, this part number from this company was constructed well." Modern listings may reflect the 2020+ product, rather than the 2020 construction that got the good reviews.