Vidsy · 2024 · Case study · 6 min

An inspiration library people actually opened.

Creative leads at Vidsy needed reference videos to brief creators and reassure clients. The library they had was a proof of concept everyone half-used. I redesigned it around how they actually searched, and let AI do the tagging nobody had time for.

Role
Sole product designer
Did
Research, workshop, prototyping, visual design
With
Creative leads, strategists, one engineer
Year
2024

Findability

63 → 92%

satisfied, survey before and after

Overall experience

74 → 96%

satisfied, same survey

Time on inspo

−16%

time tracking plus Slack monitoring

Worth

£133k

capacity freed, extrapolated to a year

If you only read one section: what I decided the world what was broken the bet how I found out what happened differently

The redesigned Vidsy inspiration library open on a laptop, on an olive-green desk with ceramics

01 · The world

Creative meets tech, with a bottleneck in the middle.

Vidsy makes video ads for brands like Pepsi, L'Oréal and Google through a network of independent creators. A brief comes in, a creative lead shapes it, creators apply and make videos, Vidsy tidies them up and ships. The whole thing runs on alignment: the client, the lead and the creator need to picture the same video before anyone shoots anything.

Inspiration is how that alignment happens. Show the client three reference videos and they relax. Show the creator the same three and they know what "playful but premium" means this week. So every brief starts with someone hunting for references, and that hunt was where time went to die.

  1. ClientBrief arrives
  2. Creative leadFinds inspo, shapes brief
  3. CreatorsApply, make videos
  4. VidsyTidy and ship
  5. ClientApproves, hopefully

02 · What was broken

A library everyone agreed was useful and avoided.

It's great in principle, but it can be a pain to use.Creative lead, interview

What worked

  • Organised by Vidsy's own frameworks
  • Self-serve; browse by category
  • Not limited to your own videos
  • Fast to skim a lot of content

What didn't

  • Slow, manual curation, so it fell out of date
  • No way to judge quality
  • Slow to load, constant log-in problems
  • Tags and details incomplete
  • Search clunky; filters combined the wrong way

The library

The old library: a password-protected WordPress page with a paragraph of instructions and a row of tiny video thumbnails
A protected WordPress page. Instructions at the top, six thumbnails the size of postage stamps, and a filter bar nobody trusted because the tags behind it were incomplete.

03 · The bet

How might we use AI to improve quality and throughput?

Before designing anything I wanted a number that made the project worth doing. This is the arithmetic that got it funded, laid out the way I showed it in the room.

Inspo's share of a lead's month

[x%]

From time tracking across creative leads. Placeholder until you confirm the figure.

×

Target reduction

30%

Conservative, based on what shadowing suggested a working search would save.

×

Value of a lead-hour

[£]

Average value generated per hour of a creative lead's time, annualised.

=

Capacity that unlocks

~£833k

Projected annual value if leads spent the saved time on briefs instead of hunting.

Projection deliberately rounded down. Section 06 shows what actually happened.

04 · How I found out

Six methods, in the order they paid off.

Each one answered a specific question. Listed as what it told me, not what it was called.

Research artefacts: a curation worksheet, a Google Analytics event table and a tagged spreadsheet of Slack requests
The Slack channel, coded. Every "does anyone have an example of…" became a row with a category, which became the search dimensions.
  1. 01

    Google Analytics

    What people searched for and which filters they touched. Told me which of our categories were real and which were wishful.

  2. 02

    Slack channel

    Leads asked each other for inspo when the library failed. I coded months of requests. The categories that emerged became the tagging schema.

  3. 03

    Interviews

    Six creative leads and strategists. The library wasn't disliked; it was distrusted. Different problem.

  4. 04

    Shadowing

    Watched people actually look for references. Most started outside the library: TikTok ad library, Twirl, MagicBrief. The competitor set wrote itself.

  5. 05

    Competitor analysis

    SWOT on platform ad libraries versus personal-library tools. The gap: nothing combined our frameworks with saving and sharing.

  6. 06

    Workshop

    How should the library be curated, and by whom? Answer: automatically, from internal ratings and client selections, or it would rot again.

05 · What I decided

Four decisions, each with its evidence.

I explored a trends view and an immersive player along the way. Some of that shipped, some was cut. What follows is what made it in, and why.

Curate automatically, or not at all

Shipped

Manual upload was the reason the old library went stale. New videos enter from internal ratings and client picks, with AI applying the framework and technique tags. Humans correct; they don't originate.

WorkshopInterviews
Filter sidebar showing Framework, Visual techniques, Project type and Industry groups beside a video grid
Filters mirror the tagging schema from the Slack analysis.

Search that widens instead of narrowing

Shipped

Two filters in one category used to return the intersection, usually nothing. Now the union. Clicking a title jumps to the original project.

Usability testingCompetitor analysis
Brand filter and clickable brand logo revealing all content from that brand
Brand became a first-class filter.

Trends as a way in

Shipped, reduced

Nearly everyone mentioned sites that teach while they show. A trends tab surfaces which frameworks are moving, with the numbers, so a lead can start from "what's working" rather than a blank search.

InterviewsAnalytics
Trends view for a brand showing framework percentages: Talking Heads 40%, Delight 38%, Interactive 26%, Gamification 12%, Breakout 2%
Real usage, not curation.

Playback and access that just work

Shipped

Unglamorous and the biggest single win. Hover to play, one-click download and link copy, no separate log-in. The immersive view with boards and comments was scoped down to the quick actions people actually used in testing.

Usability testingShadowing
Immersive view of a video with related videos, boards and a comment thread
The full immersive view, most of which was cut. The quick actions survived.

Stripping colour out of the grid and only adding it back on the filters was the last change before ship, and the one people commented on most. Nobody said "it's calmer." They said "I can find things."

06 · What happened

Measured the way I said I would.

Success criteria were set before design: continued usage in analytics, a before/after usability survey, and whether the Slack requests dropped. All three moved.

Findability · survey, before → after

63% → 92%
Occasionally the AI tagging is wrong but I'm still able to find something that works, a lot faster than before.Creative lead

Overall experience · survey, before → after

74% → 96%
I love that I'm able to find content that I've never seen before.Strategist
£133,280

A 16% reduction in time spent on inspiration, from time tracking and Slack monitoring over the first weeks, extrapolated across the rest of the year as capacity creative leads could spend on briefs. Under the 30% target, and reported as such.

Survey ran a few weeks after launch with the same questions as the baseline. Respondent count: [n]. Time tracking was self-reported; Slack requests were counted weekly.

07 · Differently

What I'd change with hindsight.

I'd have shipped the boring half first. Playback, access and the union search were most of the satisfaction gain, and they didn't need AI. The tagging model could have arrived a month later and nobody would have minded.

I'd also have set the 30% target lower in public and higher in private. We hit 16%, which is a real result, but the room heard "half of what she promised" before they heard "£133k."

And I'd have asked the one lead who never used the library why, before the redesign rather than after. He still doesn't. He has a folder.

Figures in square brackets are placeholders for Ruth to confirm. Screens are from the original case study.