Building the perfect bra takes thousands of data points.
That’s why Honeylove isn’t just another intimates brand. We’re a technology company that happens to make exceptional bras, tops, shapewear, and bodysuits. Technology shapes everything we do, from how we iterate garments based on customer feedback to how we optimize sizing across those thousands of data points.
When Honeylove was born in 2018, though, our data wasn’t consolidated. We were looking at analytics in Shopify, checking email campaign performance in one platform, and reviewing ad metrics in another. We weren’t connecting the dots as effectively as we could have.
Then we fell in love with BigQuery. In this post, we’ll cover how Honeylove uses BigQuery and Gemini to unify our data, automate key business insights, and leverage AI to boost product quality and service efficiency — as well as how other organizations looking to make the most of their data can follow our approach intimately.
Transforming insights with BigQuery and Gemini
The first step was getting all our data in one place. BigQuery gave us exactly what we needed: a performant, economical, unified data platform that integrates seamlessly with the tools our team already uses within the Google ecosystem, such as Google Ads and Google Sheets . This helped eliminate manual data silos and enabled us to quickly adopt AI and ML capabilities across the business.
The real transformation came when we started leveraging BigQuery ML functions for contribution analysis. We built models to analyze the key drivers behind some of our most critical metrics: conversion rate, customer satisfaction scores, website performance, and return rates.
What’s really powerful for us is that we can feed these contribution analysis results directly into Gemini to produce accessible reports and summaries. Before implementing this approach, 10 to 15 people would spend an hour before key meetings manually reviewing dashboards, trying to drill into the data and find meaningful insights. We’ve saved hundreds of hours per year just by automating this process with Gemini.
But the impact of BigQuery and Gemini goes beyond time savings. These tools help us find patterns and insights we would’ve missed entirely. Even if you have the best marketing analysts looking over dashboards, they just wouldn’t be able to slice it in the same way these reports allow us to do.
We’ve also been able to transform forecasting inventory and demand planning, another area where manual processes previously dominated. By deploying and training BigQuery ML’s ARIMA univariate forecasting models, we’ve used high-accuracy SKU-level demand forecasts that automatically adjust for seasonality and recent changes.
These automated forecasts consistently come within 5% of what we calculate manually — a huge improvement over third-party vendors that were sometimes off by 20% to 30%. Having that additional checkpoint gives us more confidence when making critical inventory decisions.
Unlocking value and creative with multimodal embeddings
Customer service tickets can be a treasure trove of valuable feedback and information for ecommerce brands. But only if you can extract insights from them, and with Google Cloud, we can.
We leverage Gemini embedding models and BigQuery vector search to transform the unstructured text of our tickets into actionable data. We generate vector embeddings for tickets already in our data warehouse using simple SQL commands, and then use those vectors for semantic searching through retrieval-augmented generation (RAG).
This allows us to ask precise, natural-language questions, such as “What do customers love about our bras?” or “What changes would you like to see to our bodysuits?” In response, Gemini instantly identifies similar use cases, enabling us to move beyond keyword matching and quickly find the root causes of any issues, which are often nuanced. This proactively guides product improvements and enhances service efficiency. We’re saving about 30 seconds per ticket, which might not sound dramatic until you multiply it across thousands of interactions.
We’re also experimenting with multimodal embeddings for video asset search across our ad and influencer content library. It’s been fun to test queries like “find me videos with dogs” or “find me a video with a red dress” and watch it actually work. The next step is to use those embeddings to compare new creative assets with existing ones and predict performance based on our historical data.
Growth creative has traditionally been driven by gut feelings rather than numerical analysis, but we hope to change that by using our huge library of existing ad creative to inform what we test and create in the future.
Building for the future with Google Cloud
Today, Google Cloud and BigQuery are a central pillar of our company. They allow us to spend less time on manual tasks and more time on high-value work that solves real-world problems, making us very efficient as a small team.
Working with the Google Cloud team is invaluable. They’ve been a true partner, and they continue to support our roadmap. We’re leaning further into BigQuery ML functionality, moving more of our data science work into automated, always-available models rather than offline analyses.
We’re also developing internal knowledge bots using the Vertex AI RAG Engine, connected directly to our internal documents hosted on Google Drive, to provide instant answers to internal policy and process questions. Additionally, we’re experimenting with Conversational Analytics API to provide a “BI in a box” experience so our teams can ask plain-text questions and get metrics and charts without needing an analyst.
As a technology-first company, this transformation continues to have a profound impact on what we do at Honeylove. It accelerated innovation in product quality, improved operational efficiency, and ensured that our customers receive a more intelligent and consistent service experience.