Case Studies
 / 
Accuracy, Latency, & Cost
June 23, 2026

Factory Cuts Droid Web Search Latency by 5x and Pushes Reliability Past 99.9% with You.com

Lance Shaw

Product Marketing Lead

Share
  1. LI Test

  2. LI Test

How Factory updated their Droid agent with the You.com Web Search API to ground it in faster, more reliable, and more cost-efficient context.

AT A GLANCE
Customer Industry Product Solution
Factory AI Coding Agents Droid — agentic SDLC platform You.com Search API

Factory helps enterprises build their software factory, an end-to-end system where agents turn signals into shipped software. Factory's Droid coding agents implement features, validate code, monitor production, and triage incidents across the TUI, web/Desktop app, Slack, and Linear. Current web search is critical infrastructure for Factory, grounding every workflow in up-to-date documentation instead of stale training data.

After hitting reliability and latency ceilings with their previous provider, Factory benchmarked You.com against a competing offering and migrated to the You.com Search API. Reliability jumped to above three 9s. P50 latency dropped from over five seconds to sub-one second. Sessions are cheaper because Droid converges on the right answer faster when it starts with the right context.

5x+

Faster search latency

P50: >5s → sub-1s

99.9%+

Web search reliability

Critical for proper grounding

Lower

Session costs

Fewer iteration loops

The Customer

Factory helps enterprises ship software through a continuous feedback loop. Its coding agents span the full software development lifecycle, implementing features, validating code, monitoring production, and triaging incidents. Engineers sit at the center as the governance layer, setting direction, enforcing standards, and owning the outcomes the software factory produces.

Because of this, the Droid underlying model and tooling stack have to be both fast and dependable. Web search in particular plays a quiet but critical role: it grounds Droid in current, accurate documentation whenever the model's training data falls short.

The Challenge

Factory's previous web search provider was creating two persistent problems for the Droid harness. Reliability occasionally dropped below three 9s, meaning a nontrivial slice of agent sessions experienced failed search calls. P50 latency was above five seconds, adding noticeable drag to every grounded action. For an agent that may call web search several times in a single session, those two issues compound— slowing convergence to a correct answer and forcing Droid into more iteration loops on integrations and unfamiliar APIs.

The decision to move to a better Search API came down to operational performance: Factory needed a search API that could meet agent-grade reliability and latency SLAs at scale.

The Solution

Factory benchmarked You.com for reliability, latency, and quality and, based on its superior performance, chose to migrate Droid's web search tool to the You.com Search API. Inside Droid, the API is used wherever the agent needs to ground its work in fresh information — most often:

  • Pulling up-to-date SDK and API documentation when implementing or extending an integration.
  • Fetching code snippets and architectural overviews when navigating an unfamiliar system or library.
  • Looking up configuration steps for MCP servers and other developer tooling that Droid is asked to enable.

The Results

After the migration, Factory measured marked improvements on the two metrics that drove the switch:

  • Reliability: Web search success rate moved from below three 9s with the previous provider to three 9s or higher with You.com, with some windows reaching four 9s.
  • Latency: P50 web search latency dropped from over five seconds to sub-one second—a 5x+ improvement on the median call.
  • Session efficiency: With You.com, sessions are cheaper. With better upfront grounding, Droid avoids redundant iteration loops and converges on correct answers sooner. Factory is seeing lower costs for sessions, and Droid is reaching a correct answer much sooner because it has access to the right context.

"The reliability and latency improvements when using You.com were the decisive factors. We cut P50 web search latency from over five seconds to sub-one second and pushed our success rate to three 9s or higher. Anecdotally, sessions cost less because Droid gets to a correct answer faster when it has the right context up front."

Jonathan Lucuix-Andre

Technical Staff, Factory

Build Agents Grounded in a Better Context

As the leader in agentic software development, Factory is obsessively focused on the speed and reliability of every part of the software development lifecycle. You.com's Web Search API will remain a key driver as Factory helps enterprises build their own software factories.

If your agents depend on the open web for grounding, retrieval, or research, the You.com Web Search API delivers the reliability and latency profile that agent-grade workloads demand — backed by the same search infrastructure that powers customer deployments at Factory and across leading AI teams.

Featured resources.

Paying 10x More After Google’s num=100 Change? Migrate to You.com in Under 10 Minutes

September 18, 2025

Blog

September 2025 API Roundup: Introducing Express & Contents APIs

September 16, 2025

Blog

You.com vs. Microsoft Copilot: How They Compare for Enterprise Teams

September 10, 2025

Blog

All resources.

Browse our complete collection of tools, guides, and expert insights — helping your team turn AI into ROI.

Bar chart showing model accuracy on DeepSearchQA; Frontier leads at 83.67%, followed by others ranging from 81.9% down to the lowest score of 21.33%.
Product Updates

Introducing the You.com Research API—#1 on DeepSearchQA

You.com Team

February 26, 2026

Blog

A person standing before a projected screen with code, holding a tablet and speaking, illuminated by blue and purple light.
AI Agents & Custom Indexes

Why Agent Skills Matter for Your Organization

Edward Irby

Senior Software Engineer

February 26, 2026

Blog

Illustration with the text “What Is P99 Latency?” beside simple line-art icons, including a circular refresh symbol and layered geometric shapes.
Accuracy, Latency, & Cost

P99 Latency Explained: Why It Matters & How to Improve It

Zairah Mustahsan

Staff Data Scientist

February 25, 2026

Blog

Modular AI & ML Workflows

How to Add AI Web Search to n8n

Tyler Eastman

Lead Android Developer

February 24, 2026

Blog

Abstract circular target design with alternating purple and white segments and a small star-shaped center, set against a soft purple-to-white gradient background.
Modular AI & ML Workflows

Give Your Discord Bot Real-Time Web Intelligence with OpenClaw and You.com

Manish Tyagi

Community Growth and Programs Manager

February 20, 2026

Blog

Blue graphic background with geometric lines and small squares, featuring centered white text that reads ‘Semantic Chunking: A Developer’s Guide to Smarter Data.’
Rag & Grounding AI

Semantic Chunking: A Developer's Guide to Smarter RAG Data

Megna Anand

AI Engineer, Enterprise Solutions

February 19, 2026

Blog

Clothing rack seen through a shop window, displaying neatly hung shirts and tops in neutral and dark tones inside a softly lit retail space.
AI Agents & Custom Indexes

4 AI Use Cases in Retail That Demonstrate Transformation

Chris Mann

Product Lead, Enterprise AI Products

February 18, 2026

Blog

Graphic with the text “What Is a Forward-Deployed Engineer?” beside abstract maroon geometric shapes, including concentric circles and angular line designs.
AI Agents & Custom Indexes

The Forward-Deployed Engineer: What Does That Mean at You.com?

Megna Anand

AI Engineer, Enterprise Solutions

February 17, 2026

Blog