Customer story
How UNTUCKit turned a two-month pilot into a long-term partnership
UNTUCKit wanted more shoppers engaging with search. They ran a two-month A/B test, saw the numbers, dropped the split, and went all in.
About UNTUCKit
UNTUCKit is a men's and women's apparel brand known for shirts designed to be worn untucked. It started with a simple fix to a common problem: most casual, collared shirts are cut too long to look right untucked, so they look sloppy. UNTUCKit shirts are built at the perfect untucked length with an impeccable fit throughout so you can look sharp and feel comfortable. What began online has grown into a major omnichannel brand, with more than 70 stores across the U.S. in addition to a very robust e-commerce business.
Why they started
UNTUCKit knew that shoppers who search convert at a much higher rate than those who don't, so getting more visitors to engage with search was a real business lever. One of the first things Nobi did was add an "Ask AI" button to their nav bar, giving shoppers a new way to engage with the site. Then they ran a two-month A/B test to see what the results actually looked like.
What the test found
Three places Nobi pulled ahead
Fixing misspellings
A shopper typed "smiths sirt." Their existing search returned no results and fell back to top sellers. Nobi figured out they meant the Smithson Shirt. It was the first result, and the shopper bought it.
Smithson Shirt not found. Sale lost.
Smithson Shirt surfaced. Sale saved.
Better product rankings
A search for "white button down" returned a t-shirt as the first result. Nobi returned only actual button-downs. The shopper bought the Gironde.
Semantic search
A search for "navy vest" (a product UNTUCKit didn't carry in navy) returned t-shirts. Nobi confirmed no navy vests were available but surfaced vests in other colors and other navy items. The shopper stayed on the site.
The result
The A/B test ran for two months. Then they stopped splitting.
Nobi's conversion rate came in at 17.6% against their existing tool's 15.0% — a +17.1% lift. Revenue per searcher climbed from $32.30 to $39.17, a +21.3% improvement. Average order value moved from $215 to $222.
The numbers were clear enough that UNTUCKit didn't need to keep running the test. They moved Nobi from a split to 100% of their traffic and started looking at what else they could do with it.
Head to head — separate test
In a direct head-to-head against their existing search tool, Nobi came out ahead.
Paid incumbent
16.5%
conversion rate
Nobi
17.4%
conversion rate
After the pilot
They went all in, and started shaping the product
Once they committed fully, UNTUCKit pushed deeper into Nobi than almost any other customer. Some of what they asked for became features every Nobi customer uses today.
They asked for a query channel report
UNTUCKit wanted to understand which channels their search traffic was coming from and which queries drove revenue. Nobi built it. That report is now in every customer's dashboard.
window.NobiFeatureHooks = {
onProductDisplay(product) {
return {
badge: getBadge(product),
subtitle: product.fit,
};
},
onResultsRender(results) {
return sortByInventory(results);
},
};They inspired Nobi's Hooks API
UNTUCKit's team wanted to control how Nobi displayed products, customizing each card to match their experience. That request became Nobi's Hooks capability, which lets any developer inject custom logic into the assistant without touching Nobi's code directly.
They run Nobi insights in weekly meetings
UNTUCKit's team reviews Nobi's search insights every week as part of their regular business review, looking at query trends, zero-result searches, and what shoppers are actually asking for. It became a standing part of how they think about their catalog.
They now have full search history
Before Nobi, UNTUCKit's search data only went back 12 months. Nobi stores everything from day one, so they can track long-term trends, compare seasons, and build a real picture of how search behavior changes over time.
See what Nobi can do for your brand
Start with a two-month test. Keep going if it works.