Portland Water Bureau AI Algorithm Offers 40% Discount to Wealthy CEO in Pilot
The Portland Water Bureau's machine learning algorithm, developed by SERVUS to identify utility customers eligible for discounts, reportedly selected Columbia Sportswear CEO Timothy Boyle—one of the city's wealthiest residents and largest water consumers—to receive a 40% bill discount in a pilot program intended to help low-income customers.
What happened
The Portland Water Bureau contracted with SERVUS in February to use machine learning to distribute discounts to customers who cannot pay bills in full, under a $350,000 contract. The contractor ran pilot programs using randomization software to test whether its algorithm could accurately identify customers' ability to pay.
In October, Columbia Sportswear CEO Timothy Boyle received a letter from the Water Bureau offering him a 40% discount on his water bill as part of the first 90-day pilot test. Boyle earned more than $1.6 million in compensation the prior year, held Columbia shares worth more than $1.7 billion, and was the ninth-largest residential water consumer in the city in 2021, using more than 770,000 gallons annually.
Reported impact
Boyle was offered a utility discount intended for customers who fall behind on bills and cannot pay in full, despite being among the city's wealthiest residents and highest water consumers.
Index analysis
The Water Bureau's $10 million customer assistance program has been means-tested since 1995 but has never reached its annual goal of helping 10,000 customers, assisting at most 8,500 per year. The bureau intended the algorithm to identify eligible customers who had not sought discounts due to lack of awareness or discomfort sharing income information.
Governance controls
The City Council approved the $350,000 SERVUS contract in February, and the machine learning pilot covered randomized selection of customers for one 90-day billing period, with no guarantee of selection in a second round or long-term discounts.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.