The frame is three things: one number rules at each stage of the business, businesses fall into a handful of models each with its own metrics, and there are five stages you go through in order. Around that sit dozens of case studies from 2012, which is the part that held up worst.
What I took from it
Drawing the line before measuring. Deciding in advance what number would count as good, and writing it down. Without that line, any result can be explained afterwards, and it always gets explained in the flattering direction. It is the cheapest discipline I know and the hardest to keep.
Second: a good metric is a ratio, not a total. A total only ever goes up, which is why it reassures without informing. A ratio can be compared with last month's and with somebody else's.
Third: one number rules per season and the rest are context. Working alone this has an immediate consequence, because the typical failure is not measuring too little but measuring ten things and deciding with none of them. Since I started applying it, the monthly review opens with one figure chosen in advance and the others exist only to explain it.
What I left out
Nearly all the case studies, which aged badly, and the vocabulary of the era, which now sounds like a startup conference. There is also an odd hole for a book about metrics: professional services are missing from the business models it catalogues, and that is precisely where deciding what to measure is hardest, because there are no users and no conversion rate, there are conversations and capacity. That work you have to do yourself.