This isn't terribly hard. The problem, as anyone who has done lots of web scraping knows, is finding the patterns to consistently and accurately extract data from pages.
To illustrate, if I am comparing names of a product...
A simple system will get text inside of this div and compare to the text in the span on another page to make sure the products are right.
A robust system will shingle and compute the Jaccard similarity of the names, use an algorithm to find likely manufacturing numbers that correlate to the product on the page and compare those between pages. The end result would be a metric/probability that you've identified the same product. Based on a threshold you then decide whether it is or isn't the same product.
That may seem overkill, but that's often what it takes to get a web scraper to reliably extract data. Because if you make buy/sell decisions from those prices just one bad extraction can cost you big $$$.
I usually just pray that there is a public API available.