For years, Joe Golden has helped his mom pick out tech gadgets, offering advice on the latest products that he felt best suited her needs. But as consumer electronics continue to evolve and choices multiply, he felt the product research process was growing overly complex, and existing online resources were just not keeping pace.
“Shopping for complicated products is too difficult and time-consuming for people,” said Golden, who previously co-founded a photo product site called Collage that was acquired in 2021. “Ultimately, I saw lots of people not ending up with the right product for themselves because of these difficulties.”
He decided to take matters into his own hands, launching a tech product recommendation engine called PerfectRec. The Seattle startup, which launched its tool earlier this year, uses AI and human product expertise to rank on-the-market electronics for users based on a quick customer questionnaire.
The search engine helps shoppers cut down on the time it takes to make a purchasing decision, part of a broader goal of becoming the “starting point for shopping online,” Golden said. The market opportunity could be worth “hundreds of billions of dollars,” he said.
Golden said PerfectRec scours the internet to find its product experts in “unlikely places.” The startup characterizes these reviewers as the “kind of people who run product-specific websites or forums and spend their free time making product recommendations.” In turn, Perfect Rec pays them as freelancers or part-time employees for their product expertise, contributing to the development of its own metrics and rankings that feed into its product review engine. The startup also has about 10 full-time employees, who are mostly software engineers and data scientists.
Several online resources exist to help consumers assess products. Traditional media sites like the New York Times’ Wirecutter, Gizmodo, and CNET review tech releases.
Amazon is also trying new product review systems. The e-commerce giant is rolling out an AI feature that automatically creates customer review summaries and a product ratings tool on search pages, displaying star ratings and the percentage of reviews with that specific number.
Chatbots like ChatGPT and Google’s Bard can also recommend products based on natural language prompts, while Microsoft unveiled AI shopping features earlier this year.
Golden said the startup plans to differentiate in part through its business model. He said publications often earn commissions through affiliate links, which creates a “conflict of interest” because it might incentivize them to promote specific products and sellers, and encourages users to make immediate purchases rather than holding off for a better deal in the future.
PerfectRec, which does not currently generate revenue, plans to monetize user-provided data by selling it to advertisers. For instance, the startup will collect information about a user’s laptop preferences, including budget, dimensions, and brand. This data can then be used to help companies refine their customer targeting strategies, said Golden, who wrote part of his University of Michigan economics dissertation on search advertising.
Golden said chatbots like ChatGPT and Bard can produce hallucinations, or false information disguised as fact. He said these AI tools are often trained on outdated data, which means they may not incorporate recent product releases. He said PerfectRec’s multiple-choice questionnaire is more efficient, especially on mobile devices, compared to typing queries.
PerfectRec, which has not yet raised any outside funding, reaches around 1,000 users a day.
