Teikametrics has launched an hourly bidding tool for advertisements running on Amazon, and uses a proprietary metric to track performance.

Three years in the making, the hourly bidding algorithm is powered by an econometrics and machine-learning data model that for the first time, according to the company, allows Amazon advertisers to adjust bids automatically, adapting to changes in the marketplace.

The tool — available in Teikametrics’ platform Flywheel, which the company calls a retail optimization platform — combines data from advertising, transactions and cost of goods sold, using algorithms to optimize the campaigns, said Teikametrics CEO Alasdair McLean-Foreman, a Harvard-educated economist and likely one of the first sellers on Amazon in 2003.

“The concept of hourly bidding relates to the idea of being responsive and dynamic,” he said. “We adjust keywords as needed, but if we need more data to make the adjustment we wait longer. It predicts the optimal bid at much higher frequencies.”

A good analogy would be like a thermostat that checks the temperature in the house on an hourly basis, he said.  

Calculating the most profitable bid on an hourly basis allows Teikametrics, which has managed more than $6 billion in retail sales to date for Amazon sellers, to allocate advertising dollars where the algorithm predicts the highest return on investment.

Keywords and ad groups with high traffic volumes and competitive auctions — such as consumer electronics, household items, and apparel — are among the categories that benefit the most.

Much of the transaction information comes from Amazon’s APIs. It gets integrated with the data from the manufacturer or sellers, McLean-Foreman explains.

The bidding works with Amazon Sponsored ad, but the internet is to roll it out for use with other Amazon ad units and offer it to brands for other retail marketplaces such as Walmart, and Wayfair.

Tracking metrics also has a place in the platform. While Amazon uses a metric called advertising cost of sale, Teikametrics developed a metrics called TACOS — the total advertising cost of sale — by calculating the total sale, not just the sale attributed to advertising. This calculates performance.