How Shippo Acquires Their Most Profitable Customers from Day One
By the numbers
Overview
Company
Industry
Headquarters
Campaign Type
Snapshot
Shippo powers shipping for more than 4.6 million customers. Finding more merchants ready to ship at volume is key to scaling further, but knowing who would actually scale came too late to influence bidding. A merchant might start a trial on day one and convert to a paid subscription a week later, but true shipping volume takes weeks or months to surface.
Shippo's growth team had already pushed beyond basic conversion tracking, building custom proxy events to give Google something better than trial starts for tCPA campaigns and even testing their own predictive model for tROAS campaigns. These strategies helped scale the business, but they didn’t fully capture who would become a power shipper. Voyantis helped them take it further, predicting which merchants will become power shippers at trial start and packaging those predictions to have the most impact on the auction.
Impact at a glance
Scaled spend 16% higher while reducing cost per new active shipper by 10%,
proving efficiency held as the campaign grew
Increased ROAS by 11% with 47% lower cost
to acquire high-volume shippers
Reduced cost per label by 22%,
showing new customers are actually shipping at volume
Achieved 91% model accuracy
predicting 90-day LTV
Scaling to additional Google campaigns and Meta
with platform-specific engineering and programmatic maintenance
The challenge
Shippo's growth team understood that Google needs early signals of value to avoid scaling campaigns toward the wrong users. They had already pushed further than most advertisers, building custom conversion events based on early shipping activity, like label creation, and developing their own predictive models to test value-based bidding. These gave Google something more closely correlated with value than raw sign-ups and helped scale the business effectively, but they still couldn't predict who would become a power shipper.
Shippo tested tROAS campaigns on Google with an internal value model to see how far they could push it. When tested against Voyantis, their internal model couldn't match the results. The experiment validated what the team suspected: having a predictive model wasn't the hard part. Getting Google to actually learn from them requires dedicated infrastructure and specialized expertise.
The solve
Redefine how and when to communicate customer value
The first step was redefining what value actually meant and how far into the future to look for it. Subscriptions matter, but they don’t tell the full story. Not every trial converts to paid, and among those who do, the gap between lowest and highest-tier subscribers is substantial.
Knowing who would scale made all the difference in more efficient acquisition and became the true measure of success. Voyantis worked with Shippo to define a 90 day LTV prediction horizon - long enough to capture whether a trial converts to paid and how much they'll ship once they do.
Identify future value from day-one behavior
No single input tells the story of who will scale. Predicting who would become a power shipper meant finding patterns in the earliest moments of the merchant journey, before behavior revealed anything. Voyantis's machine learning models evaluate thousands of features to surface the combinations that matter, including shipping activity in the first hours, time-of-day patterns, geographic location, email domain, and more.
Shippo's recent business model transition meant historical patterns didn't translate cleanly to current behavior. Despite that constraint, Voyantis's models achieved 91% accuracy in predicting 60-day LTV.
Engineer and maintain signals at scale
Getting predictions into Google's system – in the right format, at the right time, structured so the algorithm can learn from them – requires careful engineering. Unlike a conversion event that either happened or didn't, predictions are estimates that evolve as more data comes in.
Voyantis bridges that gap, carefully mapping auction nuances against Shippo's business goals to translate predictions into signals Google can learn from. For Shippo, that means shaping signals to filter out merchants unlikely to scale, capping outliers so high-volume whales wouldn't distort the algorithm, and weighting the value gap between subscription tiers so Google would pursue future power shippers more aggressively.
Voyantis's agentic AI monitors signals around the clock, detects drift before it impacts performance, and maintains fallback models when something unexpected happens. It's the kind of infrastructure that requires a dedicated in-house resource to maintain.
Voyantis’ impact
10%
decrease in cost per new active shipper
11%
increase in ROAS
47%
decrease in cost to acquire high-volume shippers
22%
decrease in cost per label
More power shippers at lower cost
Shippo deployed Voyantis signals on two campaigns at 16% higher spend, measuring performance against their existing tROAS approach. The results show that Shippo can scale spend and improve efficiency at the same time, acquiring more power shippers without paying more to get them.
Scaling across campaigns and platforms
Shippo is rolling out predictive signals across additional Google campaigns. Each campaign type has its own dynamics, and strategies that work in one place have to be rethought for another.
The partnership is built for that. Voyantis continuously refines models and signals as Shippo's business evolves, as auctions shift, and as new challenges emerge. With experience navigating platform differences across hundreds of clients, Voyantis helps Shippo move faster and compress time to value on every new rollout.
