## Metadata - Author: [[collin-prather|Collin Prather]] - Full Title:: The Fallacy of Data-Driven Strategy - Category:: #🗞️Articles - Document Tags:: [[Data culture|Data culture]], - URL:: https://locallyoptimistic.com/post/the-fallacy-of-data-driven-strategy/ - Read date:: [[2025-03-23]] ## Highlights > Richard Rumelt argues that a good strategy begins with an insight which reframes your situation and surfaces new sources of strength. ([View Highlight](https://read.readwise.io/read/01jpr4vdya4vvkx8v06yb3r6c0)) > Rumelt, however, represents an insight as the sort of realization that, by definition, is not the result of some predictable process and cannot be deduced from an existing base of knowledge. ([View Highlight](https://read.readwise.io/read/01jpr56f6xrbdghvj3wz9by9pj)) > **Identifying a true business insight is less about solving for** ***x*** **and more about proving new theorems from scratch.** ([View Highlight](https://read.readwise.io/read/01jpr56rvxxmgjttmj3wr8gbcx)) > An analyst can quickly identify which web page has the best conversion rate, but there is no query that can determine whether overhauling the pricing model will produce more revenue in 3 years time. This is why Tristan Handy [writes](https://roundup.getdbt.com/p/down-with-experimentation-maximalism) about how committing to rigid experimentation can be a poor approach for companies that haven’t yet found their product-market-fit. Experiments may help find a *local* optimum, while what they really need is a “*global* improvement that’s worth optimizing in the first place.” ([View Highlight](https://read.readwise.io/read/01jpr573veafmre4z6vpymt10p)) > Andrew Gelman and Guido Imbens helpfully [point out](https://www.nber.org/papers/w19614) that most statistical frameworks enable us to estimate the effects of causes, while many real-world situations instead call for us to find the causes of effects. ([View Highlight](https://read.readwise.io/read/01jpr59e3p882qtknqqwvyvfe7))