The algorithm that decides what song plays next on your Discover Weekly playlist is about to get a second act — this time helping you decide what to buy. A new startup called Aisle, founded by three former Spotify engineers, has raised $10 million in seed funding to bring the same machine learning techniques that power Spotify’s famously sticky recommendation engine to online retailers struggling to keep shoppers engaged.
The round was led by Index Ventures, with participation from Y Combinator and several unnamed angel investors who previously worked in music-streaming and adtech. It’s a modest sum by Silicon Valley standards, but it’s generating outsized buzz because of who’s behind it — and because of a growing belief that e-commerce personalization has been stuck in the past while music and video streaming have raced ahead.
From Playlists to Product Pages
Aisle’s founding team — CEO Marcus Reyes, CTO Priya Nandakumar, and Head of Product Devon Kessler — spent a combined 14 years at Spotify working on the systems behind Discover Weekly, Release Radar, and the Home feed’s personalized shelves. Those products are widely credited with transforming Spotify from a jukebox app into a habit-forming discovery engine, one that reportedly drives a significant share of listening hours through algorithmic recommendations rather than direct search.
The pitch behind Aisle is deceptively simple: if collaborative filtering, embeddings, and sequence modeling can predict what song a listener wants to hear next based on subtle listening patterns, the same techniques should be able to predict what product a shopper wants to see next — long before they type a search query.
“Spotify doesn’t wait for you to search for a song,” Reyes said in an interview. “It learns your taste from behavior — skips, replays, time of day, what you listened to right before. E-commerce is still mostly stuck on ‘customers who bought this also bought that.’ That’s 2005-era logic dressed up in a modern UI.”
That comparison resonates with a lot of retail executives who have watched conversion rates plateau even as they’ve poured money into flashy storefronts. The core issue, according to Aisle’s founders, isn’t a lack of data — retailers have plenty of purchase history, browsing logs, and clickstream data. The issue is that most of that data is used reactively rather than predictively.
What Aisle Actually Does
Unlike traditional recommendation widgets that surface a handful of “related products” at the bottom of a page, Aisle positions itself as an infrastructure layer that retailers plug into their existing e-commerce stack — whether that’s Shopify, a custom-built site, or a legacy platform like Salesforce Commerce Cloud.
The system ingests behavioral signals in real time and builds a dynamic taste profile for each shopper, similar to how Spotify models a listener’s evolving preferences rather than treating them as a static demographic bucket. According to the company, its core product includes:
- Real-time taste embeddings that update with every click, scroll, and add-to-cart, rather than being recalculated in nightly batch jobs
- Cross-category discovery that surfaces products a shopper hasn’t searched for but is statistically likely to want, based on patterns learned across the platform’s full customer base
- Session-aware personalization that adjusts recommendations based on what a shopper is doing right now, not just their historical purchase record
- A/B testing tools built specifically for merchandisers, letting non-technical teams tweak recommendation logic without engineering support
- Explainability dashboards that show retailers why the model surfaced a given product, addressing a common complaint that black-box recommendation engines are impossible to audit
Nandakumar, who led ranking infrastructure at Spotify before co-founding Aisle, says the biggest technical challenge wasn’t the modeling itself but adapting streaming-style architecture to retail’s messier, lower-frequency data. “People listen to hundreds of songs a week. They don’t buy hundreds of products a week,” she said. “We had to rethink how you build confident predictions from sparser signal, and that meant borrowing more from session-based recommendation research than from classic collaborative filtering.”
Why Now, and Why Investors Are Betting on It
The timing isn’t accidental. E-commerce personalization has become one of the most crowded — and most disappointing — categories in retail tech. Big players like Amazon and Netflix set consumer expectations for hyper-relevant recommendations more than a decade ago, but most mid-market and enterprise retailers still rely on rules-based systems, rented adtech infrastructure, or recommendation modules bundled into their e-commerce platform that were never designed to handle the complexity of modern shopping behavior.
At the same time, rising customer acquisition costs on platforms like Meta and Google have pushed retailers to look inward, focusing on squeezing more value out of the customers they already have rather than constantly paying for new traffic. Personalization has become one of the few remaining levers retailers can pull that doesn’t involve an ad platform taking a cut.
Index Ventures partner Sofia Lang, who led the investment, framed the bet less around the specific product and more around the founding team’s pedigree in a discipline that’s historically been dominated by streaming and social media companies. “The best recommendation engineers in the world have spent the last decade building for music, video, and social feeds,” Lang said. “E-commerce has been an afterthought for that talent pool. We think that’s about to change, and we wanted to back the team most credibly positioned to make that shift.”
There’s also a broader industry pattern investors are pointing to: several notable AI infrastructure startups over the past two years have been founded by alumni of consumer tech companies applying internal tooling to new verticals. Former ad-ranking engineers from social platforms have launched startups serving fintech risk models; former search engineers have moved into legal-tech and healthcare. Aisle fits a recognizable mold — take a narrow, highly optimized internal capability and repackage it as a horizontal product for an industry that’s behind the curve.
Early customers, according to the company, include a handful of direct-to-consumer apparel brands and a mid-sized specialty retailer that Aisle declined to name publicly. The startup says early pilot data showed a meaningful lift in click-through rates on recommended products compared to the retailers’ existing systems, though it stopped short of releasing specific numbers, citing customer confidentiality agreements.
The Skeptics’ Case
Not everyone in retail tech is convinced the Spotify comparison holds up. Several e-commerce analysts have noted that music discovery and product discovery are fundamentally different problems. Listening to a song carries near-zero financial risk and can happen dozens of times a day; buying a product involves cost, commitment, and often a slower decision cycle. A bad song recommendation costs a listener 30 seconds of skipping. A bad product recommendation costs a shopper money and, potentially, trust in the retailer.
There’s also the question of data volume. Spotify’s recommendation systems benefit from enormous, dense datasets — hundreds of millions of users generating billions of listening events daily. Most individual retailers, even large ones, simply don’t have that scale of behavioral data about their own shoppers, which raises questions about how well Aisle’s models can perform for smaller merchants without pooling data across multiple retailers — something that would raise its own set of privacy and competitive concerns.
Reyes acknowledges the challenge but argues it’s precisely why Aisle is built as a shared infrastructure layer rather than a single-tenant tool. “We’re not asking every retailer to train a model from scratch on their own thin dataset,” he said. “We’re building a foundation model for shopping behavior that gets smarter across our customer base, then fine-tuning it per retailer. That’s closer to how modern language models work than how old-school recommendation engines were built.”
What Comes Next
With $10 million in the bank, Aisle says it plans to spend the next year focused on three priorities:
- Expanding its engineering team, with a particular focus on hiring machine learning researchers with session-based and sequential recommendation experience
- Deepening integrations with major e-commerce platforms to reduce implementation time for new retail customers
- Building out enterprise-grade privacy and data governance tooling, anticipating scrutiny as it scales across retailers with different data-handling requirements
Whether Aisle becomes the “Spotify of shopping” or simply another entrant in an already saturated personalization market will depend on execution as much as pedigree. But the underlying thesis — that the recommendation science refined by streaming giants has real, underexploited value outside of media — is difficult to dismiss. As acquisition costs keep climbing and customer patience keeps shrinking, retailers may have little choice but to get serious about actually understanding what shoppers want before they know it themselves.
For now, Aisle is a small team with a big analogy and a modest war chest. Whether that analogy holds up at scale is the $10 million question — and the one its investors are betting will pay off many times over if it does.
Frequently Asked Questions
Q: How is Aisle different from existing e-commerce recommendation tools?
A: Most legacy recommendation widgets rely on static rules like “customers who bought this also bought that,” recalculated periodically in batch. Aisle says its system updates shopper profiles in real time using techniques adapted from streaming-media recommendation systems, aiming for more dynamic, session-aware personalization rather than fixed product pairings.
Q: Does Aisle require retailers to share their customer data across a shared platform?
A: The company says it fine-tunes a shared foundational model on each retailer’s individual data rather than pooling raw customer data across clients, though it has not published detailed technical documentation on how data isolation is enforced in practice.
Q: Which e-commerce platforms does Aisle currently support?
A: Aisle says it currently integrates with Shopify and custom-built storefronts, with plans to expand support to additional major commerce platforms as it scales, funded in part by this seed round.
