Are you asking which tools to use or what to measure? Strategy to me is what to measure, although there is some technical strategy on how to measure and display the data so it isn't overwhelming and useless.<p>First, measure most everything at the day, week, month and yearly level, then do week over week, month over month and year over year comparisons. Usually this means we store the most granular level and then rollup from there.<p>What to measure depends on what you are doing, but IMO these are some general things you should always measure: signups, cancels, time on site, time from signup to first use, A/B tests of course on features, pricing etc. This is in addition to the normal reliability, server metrics etc. I recently setup our sites to start tracking peoples first visit until signup, of course this isn't foolproof as it relies on a cookie, but I think it will help us over time.<p>The key to the analytics/metrics is you want to be able to cross section them and be able to look at when you add a new feature, or remove something etc.<p>Frankly, besides storing normal metrics, we store every API call, the time it takes, who made it, which machines were involved etc. This lets us basically look at everything that happens. From that I can get signups, cancellations, engagement statistics and a whole host of other things, this solves most of the business questions and almost all operational concerns like machines being whacked.<p>Not saying my ideas are 100% right or even correct in all circumstances, but thought I'd at least share.