Fun.com

How FUN.com's flagship brand HalloweenCostumes.com increased Revenue per Visitor by 31% with Malachyte

By adapting search and product discovery in real time — even for first-time visitors — Fun.com turned chaotic intent into measurable revenue.

Product

Add to Cart

PDP

Metrics

RPV

CVR

AOV

Period

August-October 2025

Key Results

Revenue Per Visitor (YoY)

0.0%

Recommendation-Driven Revenue

0.0%

Recommendation-Driven Revenue

0.0%

Recommendation-Driven Revenue

0.0%

Recommendation-Driven Revenue

0.0%

Search Revenue Growth

0.0%

Search Revenue Growth

0.0%

Search Revenue Growth

0.0%

Search Revenue Growth

Orders Influenced by Recs

The Challenge

Fun.com's Halloween business lives and dies by relevance. In 2025, they faced:

1,500+ misspellings of "Wednesday Addams"

Legacy search couldn't handle typos, causing customer bounces during peak demand.

Hourly trend shifts driven by TikTok

Higher intent capture across journeys

25,000+ SKUs live during peak season

Massive catalog complexity made relevant discovery nearly impossible.

Legacy SOLR search failing on long-tail

High bounce rates from irrelevant recommendations and zero results.

"Search could not handle basic mistypes or mispellings without heavy handed redirects to broad listing pages."

— Director of Ecommerce, Fun.com

"Search could not handle basic mistypes or mispellings without heavy handed redirects to broad listing pages."

— Director of Ecommerce, Fun.com

"Search could not handle basic mistypes or mispellings without heavy handed redirects to broad listing pages."

— Director of Ecommerce, Fun.com

The Solution: The Golden Path

Fun.com didn't overhaul everything at once. They followed a phased rollout designed to prove impact quickly and expand with confidence.

Fix the highest-impact surface

Adapt experiences in session

Expand across the journey

Put teams in control

Search

Fun.com started where revenue was leaking fastest: search. MiQ replaced their legacy search engine with a real-time, intent-aware system that could adapt immediately — even on the first visit.

What changed

Misspellings understood automatically

No more zero-result pages from typos

Misspellings understood automatically

No more zero-result pages from typos

Search adapted to real-time trends

TikTok-driven demand captured instantly

Search adapted to real-time trends

TikTok-driven demand captured instantly

Failed searches rescued in-session

Long-tail queries now deliver results

Failed searches rescued in-session

Long-tail queries now deliver results

Legacy Search

MiQ Search

zombie cheerleaders

Results reordered in in 50ms

Issue: Shows zombies OR cheerleaders, not combo

Understanding: "Zombie cheerleader combo" prioritized

Legacy Search

MiQ Search

zombie cheerleaders

Results reordered in in 50ms

Issue: Shows zombies OR cheerleaders, not combo

Understanding: "Zombie cheerleader combo" prioritized

Legacy Search

MiQ Search

zombie cheerleaders

Results reordered in in 50ms

Issue: Shows zombies OR cheerleaders, not combo

Understanding: "Zombie cheerleader combo" prioritized

Legacy Search

MIQ Search

green dinosaur costumes

Results reordered in in 50ms

Issue: Mixed colors, green not prioritized

Understanding: "Green dinosaur" specificity understood

Legacy Search

MIQ Search

green dinosaur costumes

Results reordered in in 50ms

Issue: Mixed colors, green not prioritized

Understanding: "Green dinosaur" specificity understood

Legacy Search

MIQ Search

green dinosaur costumes

Results reordered in in 50ms

Issue: Mixed colors, green not prioritized

Understanding: "Green dinosaur" specificity understood

Legacy Search

MiQ Search

whostbusters

Results reordered in in 50ms

Issue: No results for typo, customer bounces

Understanding: "Ghostbusters" (typo corrected)

Legacy Search

MiQ Search

whostbusters

Results reordered in in 50ms

Issue: No results for typo, customer bounces

Understanding: "Ghostbusters" (typo corrected)

Legacy Search

MiQ Search

whostbusters

Results reordered in in 50ms

Issue: No results for typo, customer bounces

Understanding: "Ghostbusters" (typo corrected)

Measured impact

+0%

Search Session Conversion Rate

+0%

Search Session Conversion Rate

+0%

Search Session Conversion Rate

0.0%

Search Revenue

0.0%

Search Revenue

0.0%

Search Revenue

0.0%

-0%

Search Revenue Growth

0.0%

-0%

Search Revenue Growth

0.0%

-0%

Search Revenue Growth

Key Insight

Legacy systems can't learn from behavior. MiQ adapts with every action — even from a first visit.

Personalized Recommendations

After stabilizing search, Fun.com extended MiQ to product detail pages — turning static recommendations into live, in-session discovery. After just 3–5 interactions, PDP recommendations adapted in real time.

What changed

Duplicate items eliminated

Recommendations became context-aware

Duplicate items eliminated

Recommendations became context-aware

Recommendations aligned to theme

Intent captured within the same session

Recommendations aligned to theme

Intent captured within the same session

Discovery improved without rules

No manual merchandising needed

Discovery improved without rules

No manual merchandising needed

Before (Legacy)

After (MiQ)

Anchor Product

Toy Story Rex Costume for Toddlers

$59.99

Recommended for You

CTR 25%

0.04

0.10

0.04

0.02

Before (Legacy)

After (MiQ)

Anchor Product

Toy Story Rex Costume for Toddlers

$59.99

Recommended for You

CTR 25%

0.04

0.10

0.04

Before (Legacy)

After (MiQ)

Anchor Product

Toy Story Rex Costume for Toddlers

$59.99

Recommended for You

CTR 25%

0.04

0.10

Measured impact

+0.0%

Recommendation-Driven Revenue

+0.0%

Recommendation-Driven Revenue

Key Insight

Cold-start PDPs became adaptive systems — learning within the same visit.

Key Insight

Cold-start PDPs became adaptive systems — learning within the same visit.

Unified Discovery & In-Session Learning

With search and PDPs aligned, Fun.com extended MiQ across additional discovery surfaces, powered by the same real-time understanding.

What changed

Homepage recommendations

Personalized landing experience

Homepage recommendations

Personalized landing experience

Add-to-cart recommendations

Cross-sell at conversion point

Add-to-cart recommendations

Cross-sell at conversion point

Cross-journey consistency

Unified intelligence everywhere

Cross-journey consistency

Unified intelligence everywhere

Measured impact

+0%

Revenue per Visitor

+0%

Revenue per Visitor

+0%

Revenue per Visitor

Orders influenced by recommendations

0.0%

-0.0%

PDP CTR improvement

0.0%

-0.0%

PDP CTR improvement

0.0%

-0.0%

PDP CTR improvement

"We haven't seen this kind of growth, at scale in years."

— CMO

"We haven't seen this kind of growth, at scale in years."

— CMO

"We haven't seen this kind of growth, at scale in years."

— CMO

Copilot & Real-Time Controls

As confidence grew, Fun.com's merchandising team began using MiQ's fine-tuning controls to guide real-time outcomes, without engineering.

What changed

Boosting trending SKUs

e.g. Barbie, Vampire Cowboy

Boosting trending SKUs

e.g. Barbie, Vampire Cowboy

Suppressing out-of-stock items

Real-time inventory control

Suppressing out-of-stock items

Real-time inventory control

Testing strategy changes safely

A/B test merchandising rules

Testing strategy changes safely

A/B test merchandising rules

Merchandising Control Panel

Boost Halloween Costumes

Targets: Product 1

50%

Prioritize seasonal costumes by 50%

Suppress Out-of-Stock Items

Targets: Product 3

50%

Reduce OOS item visibility by 50%

Margin Optimization

Targets: Product 2

50%

Prioritize higher margin products by 50%

Updated Rankings Preview:

ANCHOR PRODUCT

Luna Lovegood Costume

$69.99

Fixed

Harry Potter Ravenclaw Robe

$39.99

↑ Boosted

Professor Trelawney Costume

$129.99

↑ Boosted

Luna Lovegood Accessory Costume Kit

$9.00

Normal

Rainbow Glasses

$29.00

↓ Suppressed

Key Insight

MiQ puts your merchandising team in the driver's seat — no code, just impact.