Dani Weidner Open to Staff roles

Feature / Research to shipped

Creating and editing links at scale

Enterprise customers needed links at scale, and the only path was a single-item form or a homegrown spreadsheet. I led the research that reshaped the plan, cut the feature the research showed nobody needed, and shipped an upload flow built around fixing errors.

In one minute

SituationCustomers needed thousands of links at once, or tightly named batches of a handful, and fixing a wrong tag meant asking an admin to edit links one at a time.
What I didRan usability research across three user groups, cut a complex in-platform editor, and designed creation and editing around recovering from errors.
OutcomeBulk creation shipped, handling up to 10,000 records per file with each row validated.

Demo. Upload a file, fix only the failed rows, then edit links in bulk with the affected set in view. Fictional brand and data. Open full screen ↗

Decision10,000Records per file, each row validated with its own status
Decision5 of 7Key customer accounts named error recovery the top blocker, which set the priority
Scope14Participants across media, entertainment, and financial services

Decisions

Decision 01 / 04

Test with three groups, not one

A CSV wizard existed in prototype with no evidence it worked for the people who would use it daily. I ran moderated sessions with internal team members and unmoderated sessions with current customers and external growth marketers.

I synthesized the findings into one cross-group view with severity scored by product familiarity, and turned them into Figma Make prompts to prototype the fixes.

Decision 02 / 04

Cut the complex in-platform editor

The research showed an in-platform editing approach was hard to use and rarely needed, so it was cut, and the simpler path shipped first: download a template, fill it, upload, validate.

Tradeoff: less flexibility at launch, in exchange for the flow customers would use every day.

Decision 03 / 04

Design the flow around the rows that fail

Validation runs up front, and a two-tab preview separates the rows that will work from the rows that won’t. Users download just the failed rows, fix them, and re-upload, instead of starting the whole file over.

Error states are universally the biggest blocker. Users hit errors they couldn’t resolve. The flow works, but it doesn’t teach.Research synthesis, cross-group findings
STEP SYSTEM VALIDATION ROW ERROR ONE FLOW TEMPLATE, FILL, UPLOAD, VALIDATE, CREATE DOWNLOAD TEMPLATE FILL THE FILE UPLOAD VALIDATE ERRORS RETURNED UP FRONT, BY ROW CREATE MANY LINKS AT ONCE UP TO 10,000 PER FILE, EACH WITH ITS OWN STATUS TWO ROWS FLAGGED, THE REST CREATED
Fig. One flow in, thousands of links out, every row with its own status
Decision 04 / 04

Show the affected set before any edit applies

Bulk editing followed as a stepped wizard with a persistent rail listing the selected links, so users see exactly what will change across analytics tags, routing, and social metadata before anything applies. Nothing edits silently.

Next / Case 03
A/B testing for people who run experiments for a living