Celestial Marketplace

Company Overview and Background

Celestial Marketplace is a luxury-focused e-commerce brand offering high-end designer pieces—think statement handbags, rare collectibles, and exclusive celebrity-backed drops. Their model includes a bidding function for certain products, but that’s only part of what sets them apart. They’ve built a recognizable name in high fashion circles for their curated, high-end items, but they needed help pinning down exactly what was happening once people arrived at their site.

They had plenty of ideas about which items got the most love on social media. They also had a decent handle on which advertising channels brought the most clicks. But once folks actually reached the website, Celestial Marketplace had little insight into user behavior or which channels truly led to sales, bids, or other key actions.

They came to Curve looking for help with the behind-the-scenes data “plumbing”—from standardizing their tracking to figuring out how to unify the different platforms they used for content, bidding, and checkout.

‍

How it All Started

When we first spoke, Celestial Marketplace openly admitted a few pain points that were blocking them from growing. Here’s the condensed version:

Multiple Platforms, No Unified Funnel

They hosted item descriptions and images on one platform, the bidding process on another, and the final purchase on Shopify. That made it a scramble to stitch together data. They could see broad traffic patterns—like which channels brought in visitors—but they had no clue whether Channel A or Channel B was leading to actual conversions or abandoned bids.

Weak Attribution After the Click

Once people arrived on the site, Celestial Marketplace had almost no visibility. They knew if someone landed on a product page, but didn’t have a reliable way to track if those same visitors ended up completing a bid or checking out. It was like having a half-finished puzzle: the corner pieces were there, but the middle section was a jumble.

“We use as our data layer Segment. We don’t use Google Tag Manager. I wish someone had put in GTM as like ‘hey guys, that’d be super easy.’ The company had also deployed Segment, but the strategy as to why Segment is also being utilized as that middle layer is unfamiliar territory, even for our engineers.”

Language and Nuance Barriers with a Global Dev Team

The team was scattered internationally. They had bright engineers across different time zones, but marketing requests often got lost in translation—literally. The marketing side would ask for a new type of tracking, and it might come back partially done or labeled differently, causing confusion about what was actually being measured.

This is not my expertise. We're looking to partner with someone who could pretty much say flat out wow, you were screwed. And here's what you need to do and explain it in layman's terms to our engineering team, in terms of if there's data that needs to come in

No Single Source of Truth

Between their content platform, a bidding platform, and a Shopify checkout, each had its own approach to analytics. The data was inconsistent or simply incomplete. Sometimes the marketing team saw a conversion rate that looked too good to be true. Other times, they couldn’t figure out if a user had placed three bids or just browsed around.

They had played around with Segment in the past, but it wasn’t fully hooked into their apps in a way that would unify a single user journey. So while they technically had some infrastructure, it was barely scratching the surface of what they needed.

‍

Our Initial Audit

We kicked off with a standard audit—basically, we hopped on calls and screen shares, asked to see how their analytics and platforms were set up, and mapped out a user’s path from start to finish. Once we saw the ins and outs, a few things stood out:

Tracking IDs Everywhere

The content platform assigned a user ID, the bidding platform assigned a different one, and Shopify assigned yet another. In practice, each system treated the same person like three separate visitors. It was no wonder the marketing team got conflicting or half-baked data.

Funky Event Definitions

Some “conversions” were just page views, while actual high-value actions (like placing a bid) weren’t tagged. This meant they were celebrating some meaningless stats and missing the truly important numbers.

Segment Dabbles, But No Execution

The marketing manager mentioned they’d tried to incorporate Segment but never had a roadmap on what data to send in or how to unify user identities across those three platforms. So, Segment was present… but not doing the heavy lifting they hoped.

We presented these findings plainly, walking through each breakdown in attribution and each conflicting ID that stood in the way. Once we had that laid out, everyone understood where the leaks were.

‍

Key Projects and What We Did

Unified Identity Through Segment

We decided to make Segment the central place where user data flowed in from each platform. Instead of letting each platform do its own thing, we created a single user identification process. So if someone started on the content site, placed a bid on the bidding platform, and then checked out on Shopify, they’d remain the same user the whole time.

• We worked with their engineering team to define a standard ID. For instance, if a visitor was logged in, we used their account email or an internal user reference as the anchor.
• We added custom traits when someone placed a bid or visited certain categories, so the marketing team could finally track funnel steps from that first page view to the final checkout.‍

‍ Fixed Up Conversions

We scrapped any events that didn’t mean much to their business goals, like “Viewed a Product” counting as a conversion. Instead, we gave them a clear path to measure all the moments that actually matter:

• Signed Up / Created Account
• Placed a Bid
• Won a Bid (or purchased an item right away, if no bidding was needed)

Now, when they’d do a marketing campaign, they could see how many new folks actually created an account and jumped into the bidding world.

Cross-Platform Reporting

We hooked up Mixpanel via Segment to streamline reporting. The brand had dabbled in Adobe Analytics in the past, but we suggested going with something simpler that was friendlier for their usage patterns. Mixpanel allowed them to see unique user flows across domains—tracking funnel steps for each region—and let them break down data in a flexible interface that the marketing side actually understood.

Keeping the Marketing and Dev Teams on the Same Page

A big part of our job was simply clarifying the requests. Instead of broad statements like “We want to see how many people place bids,” we gave the dev team step-by-step instructions:

• “Add a ‘Track’ call in Segment for placed_bid when the user clicks the final ‘Confirm Bid’ button.”
• “Include a property called auction_name so the marketing team can see how Auction A vs. Auction B performs.”

This spelled out the exact tasks they needed to implement without the guesswork. The marketing team found it easier to ask for new tracking once they understood that each event just needed a name and some relevant properties.

‍

Impact

Once everything was live, Celestial Marketplace started seeing their data in a whole new light. They weren’t guessing anymore about which ad campaigns led to meaningful actions, or which auctions were pulling in the strongest participation.

Now they had:

• A Clear Funnel: They could watch the path from social ad → content platform → bid → checkout, all in one place.
• Reliable Stats on Bidding vs. Straight Purchases: If an item allowed direct purchase (no bidding) vs. a full bidding process, they could see how that changed user behavior.
• Insights for Marketing: They could finally compare which region or channel brought the most account sign-ups vs. real bids. This meant more focused campaigns—directing budgets toward the channels that actually produced results.
• Shared Language for Teams: Both marketing and engineering now used the same terms for events (“placed_bid,” “added_to_cart,” etc.) instead of different labels on each platform.

‍

Why We Liked the Project

Even though the data was scattered, it wasn’t some massive legacy system that nobody dared touch. The team had a sense of urgency—they wanted to figure things out sooner rather than later. Their marketing folks took the time to explain their real business goals (like focusing on actual bids, not empty page clicks), and their dev team, despite being spread out, was quick to implement once the requirements were clear and spelled out.

On our side, it felt rewarding to see how those small changes—like establishing a universal user ID—had an immediate impact on how they read their analytics. Suddenly, they knew which ad sets were driving sign-ups vs. which ones led to folks browsing and ghosting. They could also see what was driving new watchers of a particular auction. Once the puzzle pieces snapped into place, their conversations went from “Where is the data?” to “Now that we have the data, how do we optimize?”