2 ms·
The key wasn't so much the optimizations it was the fact they started collecting data in real time and was able to do proper forecasting and measurements. A big
by dmix 10y ago
The key wasn't so much the optimizations it was the fact they started collecting data in real time and was able to do proper forecasting and measurements. A big problem with BigCo's like this is they are slow to change.
This shows how important data collection is to improving performance in any system (beyond "If you can't measure it, you can't manage it..."). Rather than after-the-fact analysis, the improvements could be rolled out faster as experiment results are backed up with live data. Making it less risky for management.
Plus they could easily continue doing multiple ongoing experiments in different aisles, without a big investment in 'waterfall' style optimizations.