Every ATS has a pipeline. But almost none are designed by recruiters.

Over 12 years at Korea Search, Jobplanet, HYBE, and global firms, I've used dozens of ATS platforms. Greenhouse, Lever, domestic solutions -- you name it. Every tool had a "pipeline" feature, but I felt the same frustration every time. Whoever built this has never actually recruited.

So I built my own. The "ideal pipeline" that had lived only in my head for 12 years, finally turned into code.

The Problem with Existing Pipelines

The biggest problem with existing ATS pipelines is that they're designed from the administrator's perspective. They're reporting tools for HR managers to track hiring status -- not candidate management tools for recruiters.

Specifically:

Using these tools, I always maintained a separate spreadsheet. Formal updates went into the ATS; actual judgments and records lived in Excel. This double-work devours a recruiter's time.

"Declined" and "Rejected" Are Different

This was my biggest frustration, and the first problem I solved when designing the Convince-X pipeline.

In traditional ATS platforms, when a candidate exits the process, they're typically marked as "Rejected." But this single label conflates two fundamentally different situations.

Without this distinction, you can't identify where the hiring process breaks down. High declines signal offer competitiveness or position attractiveness issues. High rejections point to sourcing stage criteria problems. Different causes demand different solutions.

In Convince-X's pipeline, removing a candidate from the process requires choosing between "Declined" and "Rejected" and recording the reason. As this data accumulates, patterns emerge in the analytics tab.

One more feature: Reactivation. A candidate who declined three months ago might become interested again. A rejected candidate might be perfect for a different position. In traditional ATS platforms, the "Rejected" pool is a graveyard. In Convince-X, reactivation triggers an email notification to the candidate.

Stage Design Philosophy

Deciding on default stages took more deliberation than expected. This isn't just UI -- it's a perspective on the hiring process.

The final default stages:

Why these 6 stages? They reflect how hiring actually works in the Korean market. Global ATS platforms default to American-style stages like "Phone Screen" and "Onsite," but in Korea, the 1st/2nd interview distinction feels far more natural.

There's a reason Negotiation is a separate stage. In Korea's hiring market, the salary negotiation phase is surprisingly long and critical. For senior tech roles, this stage alone can take 2-3 weeks. Burying it within "Offered" hides where the bottleneck actually is.

Kanban Board UX

Trello already nailed kanban boards, so just copy the pattern, right? But recruitment kanban differs from project kanban in several important ways.

First, cards need different information. Project kanban cards work fine with a title and labels. Recruitment kanban cards need the candidate's name, current company and title, position applied for, last update time, and email sequence status -- all visible at a glance.

Second, a side panel is essential. Clicking a card should slide out the full candidate profile. Not a modal -- a side panel. Because recruiters need to keep the panel open while comparing with other cards. Recruiters are always comparing candidates.

Third, drag-and-drop must work in both directions. This seems obvious, but moving a candidate backward from "Negotiation" to "Applied" must also be supported. Negotiations genuinely fall apart and sometimes the process restarts from scratch.

Analytics Tab: What the Numbers Reveal

The pipeline's third tab is Analytics. I placed it next to the kanban board for a simple reason: decisions should be data-driven.

Traditional ATS analytics typically show "headcount per stage." That's a status report, not analysis.

What Convince-X's analytics tab shows:

For example, a 20% conversion rate from Applied to 1st Interview means sourcing criteria are too broad. A low conversion from Negotiation to Offered may signal compensation competitiveness issues. These insights need to surface from the numbers so recruiters can decide their next move.

Between In-House and Headhunters

One of the hardest aspects of pipeline design was accommodating different user types.

In-house recruiters and headhunters (search firms) do the same hiring work but follow different workflows.

In Phase A, I intentionally chose a neutral design. Position Name and Client are plain text fields. In-house recruiters can leave Client blank; headhunters can enter client company names. Not enforced, but available when needed.

Explicit user-type differentiation with distinct UIs is a Phase B item. Right now, we're in the data-gathering stage. Once we see patterns in how different user types actually use the tool, we'll branch accordingly.

Still Phase A

Let me be honest. This pipeline is still Phase A -- the first iteration.

Core features -- kanban board, drag-and-drop, side panel, declined/rejected distinction, analytics tab, reactivation -- all work. But there's more to build.

What I'm most excited about is AI insights. As pipeline data accumulates, AI could automatically surface patterns like "candidates for this role frequently drop off at the 2nd interview" or "this client's Negotiation stage averages 3 weeks." That's the kind of insight genuinely valuable to recruiters.

One more thing. This pipeline doesn't exist in isolation. Convince-X's AI talent search finds candidates, the email sequence engine handles outreach, and when responses come in, the pipeline manages them. Search → Outreach → Pipeline -- it's the final puzzle piece of the entire workflow.

The thought that kept recurring while building Phase A: tools reflect the perspective of their maker. When an engineer builds it, the data model is clean. When a recruiter builds it, field-level judgment is baked into the structure. Like distinguishing between declined and rejected.