There are already a couple of products in market like this, goodui.org being the main one. However, personally, I don't believe that this kind of meta-analysis of experiments from different sites has too much value, for a couple of reasons. I'm only really outlining this stuff for you because it might be food for thought if you want to create products for the experimentation community: Whilst the theory of meta-analysis is technically sound, it is also fraught with issues relating to the complexity of web experiences: Tests must be subjectively categorized and then analysed as 'patterns' when there is zero evidence that that thing was the cause of the result. For example, you might think a test won because you removed the voucher code box, but it equally could have won because of the change in page layout. No two tests are in any way the same in any controlled way. Meta analysis mainly comes from sciences where experiments can be repeated under (relatively) identical controlled conditions. This is impossible and A/B testing cannot really be compared to that kind of controlled experimentation. By comparing two different experiments you instantly remove any control you had around audiences etc because they will not be the same. The greatest benefit to be had from experimentation is by integrating it with business and brand strategy. A good business strategy is aimed at carving out a unique and differentiated direction of travel for that brand. The opportunity with experimentation is to iteratively develop and innovate the business through learning and adaptation. This means that experimentation should aim at being meaningful within the context of the brand strategy. Meta-analysis of experiments from different sites is, in many ways, the opposite of this because it seeks to make the ‘lowest common denominator’ changes which would be beneficial to any business anywhere. Conversion uplift and revenue might be achieved but what has been furthered in terms of strategy and differentiation? This is not to say that it isn’t valuable if you simply want to do ‘CRO’ but my argument is that nobody should merely do CRO.