EN
Back to the archive

The encyclopedia · R&D & Science · Strategic decision · 2019–2022

ABI's synthetic-control platform turned noisy retail tests into clear wins

Anheuser-Busch InBev built a generalized synthetic control engine, TestOps, to detect small-but-real sales wins in a large, noisy retail market.

Anheuser-Busch InBev

the move

Anheuser-Busch InBev runs retail experiments on promotions and portfolio changes, but the effects are often only a one to two percent shift in sales — real, yet easy to miss amid the noise of a huge and volatile market.

The company and MIT built TestOps on a generalized synthetic control, a statistical method that creates a fitted comparison market so a small genuine improvement stands out from the noise, across many tests at once.

The platform increased the share of tests that yielded a significant discovery, and it now runs physical experiments impacting roughly 135 million USD of revenue each month.

why it works

  • Small real sales effects are buried by market noise
  • A synthetic control gives each test a fitted, untouched comparison
  • The method scales across thousands of physical retail experiments
  • Trusting a small-but-real win lets the firm act on it and roll it out
the payoffCompare against a synthetic control groupinspired

what transfers

To prove a small effect on a large, noisy business, build a fitted control instead of trusting a raw before/after comparison; the method, not the volume of data, is what shows the win.

what came after

TestOps won the 2022 INFORMS Daniel H. Wagner Prize, and ABI is rolling it out globally. The platform routinely identifies innovations delivering a 1 to 2 percent increase in sales volume.

references

spotted an error? The archive wants to know.

same kind of clever