collectionHQ Prioritise Items

I turned a 7,000 item report into the hundred that mattered

collectionHQ could hand a librarian 7,000 titles when they only had an hour. I designed a wizard that asks how many they want.

Role: Senior UX Designer
Timeline: Nine weeks, shipped November 2024
What I owned: Research, Facilitation, Product design, Usability testing
Status: Shipped

The shipped Prioritize Items wizard open over a Dead Items Removal Action Plan, page 1 of 126. Six rows let the librarian include or exclude core items, DEI items, last copies, items in active experiments, items selected for transfer and items by popular author. Underneath, a live count reads "Your selections will leave 1,629 items", then asks "Reduce size further?". A notice by the Prioritize button warns that active filters and grouping will be removed. The customer's region and library are blurred. — 1Six things you can include or leave out
2A live count of what's left
3It all sits inside the decade-old report

The problem

One collectionHQ report could come back with 7,000 titles, and the librarians opening it had thirty to sixty minutes between desk shifts. That's completely overwhelming.

My role: I ran the research and the ideation workshop, designed the wizard, and moderated both rounds of testing.
The team: I wasn't the product owner. He had the final say, and I worked alongside him, a business analyst and the development team.

The result

7,000titles a single report could return

How this was measured

That's the biggest list collectionHQ could hand a librarian in a single report.

Product data, quoted in the research brief, 2024

15 to 200items they can actually get through in a session

How this was measured

That's the range people gave me when I asked how many items they get through. A lot of them didn't have a number at all, and just worked through a list over weeks.

Ten librarians, nine libraries, UK and US, March 2024

19moderated sessions across two studies

How this was measured

There were ten in the research study and nine in usability testing, and I moderated both.

Condens reports, March and June 2024

Context

The tool was powerful, and that was the problem

collectionHQ analyses a library's whole collection and tells staff what's stale, overstocked or underperforming. It's incredibly powerful, but it's huge, and a single report could come back with 7,000 titles.

The people opening that report have a spare thirty to sixty minutes a day, in between desk shifts, programmes and outreach. Some only get two hours a week. So when 7,000 titles come back, it's completely overwhelming, and you don't know where to start.

A collectionHQ Dead Items Removal Action Plan (Fiction), page 1 of 126, as a dense table with orange column headers: author, title, ISBN, barcode, publication date, date added, last use date, recommended action, and a row of EBSM metric tags on each item such as Dead, Collection Check, Outdated, Overstocked and Last Copy. The customer's region and library are blurred. — This is the report as it stood, and the interface I had to work inside. It was over a decade old, it failed colour contrast, and it was confusing unless you already knew your way around it. The coloured metrics column on the right is the other half of this project, and it's the half customers chose to ship first.

What librarians did instead was apply their sorts and filters, print the report or export it to a spreadsheet, and then cut the spreadsheet down by hand until it matched the time they had. All of that ate into the hour they were supposed to be spending at the shelves.

Decision 01 of 03

Let them ask for a number instead of a sort

The bind

The product already had sorting and filtering, and neither of them helped. They both answer "show me this kind of item", but neither one answers "give me as many as I can finish today", and that was the only question anybody actually had.

What I chose

So I designed the wizard around prioritising instead of filtering. A librarian says what to leave out, ranks the metrics that matter to them and, crucially, types in the number of items they want. There's a live count that updates as they go, so they can see if they've cut too far before they commit.

What it was based on

In the first study, ten librarians described the same spreadsheet workaround without me prompting them.

What it cost

It only works if you know your number, and plenty of librarians don't have one. They just work through a list of any size over weeks, so for them this doesn't change anything.

The lower half of the shipped wizard. 'Your selections will leave 1,629 items', then 'Reduce size further?' set to Yes, a box reading 150 labelled 'Highest priority items', and a ranked list of EBSM metrics: 1st Grubby, 2nd Outdated, 3rd In overstocked collection. Each row has a drag handle on the left and up and down arrows on the right. — This is the bit that matters. You type in the number you can actually get through, and then you put the three metrics in the order you care about.

The bottom of the shipped wizard. Under 'Prioritization breakdown' it reads 'Your 150 items will be prioritized in the following way': 11 items tagged Dead, Grubby, Outdated and Overstocked, 2 tagged Dead, Grubby and Outdated, and 137 tagged Dead, Outdated and Overstocked. Below that, 'Sort items in action plan' is a dropdown offering 'Priority order as above' or 'Keep existing action plan order'. — Before you commit, the breakdown shows how your 150 items get picked. You can also keep the report's existing order if that's how you like to work.

Four sticky notes under the heading 'Current user workarounds': 'Limits reports to 500 items total to share among team', 'Reports prioritised by branch or call number to avoid printing 600 pages', 'Limits to 250 items at most before fatigue sets in', and 'Reduces time period on Last Use Date to 2-3 years to shorten report'. The name under each note is blurred. — These are four of the workarounds librarians described in the first study, in their own words. Every one of them is a way of cutting the report down to what a team can actually get through.

Decision 02 of 03

I built it inside a decade-old interface, not around it

The bind

The collectionHQ interface was over ten years old. It looked dated, it failed colour contrast, and it was confusing if you didn't already know where everything was. Rebuilding the whole product around one new feature was never on the table.

What I chose

So I built the wizard to sit inside that interface and match it, and then I fixed accessibility within the feature itself. That meant contrast ratios, keyboard operation and ARIA labelling. Staying on brand with a dated product was the constraint, so I worked within it instead of fighting it.

What it cost

The feature itself is accessible, but the product it opens from still wasn't, and a wizard can't fix the page behind it.

Research

Where the design came from

I ran a three part remote ideation workshop with Product, Engineering and Customer Support. We did quick-fire concepts, then sketching, then dot voting to choose, and the Product Director had the casting vote if there was a tie. The prioritisation wizard came out of that room, not out of my head.

Then I prototyped it and showed it to Engineering before I showed a single user. That meant we caught the build concerns while they were still cheap to deal with.

A Miro board from the ideation workshop. On the left, an earlier sketch with notes listing what to filter by (Core, DEI, Last Copy) and what to prioritise by. On the right, a photographed paper sketch of a report screen with a 'Limit report size?' toggle, a report amount, a ranked 'Prioritize by' list and a 'Dynamic CTA with total items', surrounded by sticky notes such as 'parameter screen feels logical' and 'Removes extra clutter and information overload on busy report screen', and green voting dots. — The sketching round, with the room's votes in green. The sketch on the right already looks like the wizard, with a report size, a ranked list and a button that shows the total.

What the research told us

The librarians who know their number gave me anything from 15 to 200 items per session. A lot of them don't have a number, and just work through a list over weeks or months.
They already prioritise by last use date and oldest publication date. Whether a book's a bit battered or tatty matters less, except for children's books.
Most of them exclude DEI items, and most keep last copies. A few exclude series, and there wasn't any way to do that in collectionHQ at all.
If you prioritise a dead items report by condition, you get a combined weeding list. Several of them said that's something they'd wanted for years.
Most of them stop weeding over the summer, when circulation peaks and staff are on leave.

Decision 03 of 03

I let customers choose which of my features shipped first

The bind

The EBSM metrics column and this wizard were two halves of one project, and they were both mine. We didn't have the resource, so we couldn't do both at the same time. I could've just gone with my instinct.

What I chose

So I put the question to customers in the usability sessions instead. If you can only have one, which should come first? Almost twice as many chose the metrics column, because it applies to more reports, it's visible instead of hidden behind a button, and it explains why an action was recommended. So that's what we did, and the metrics column went first.

What it was based on

I asked nine librarians across eight libraries directly, in June 2024.

What it cost

That delayed Prioritise Items, the feature I'd been building towards, by about three months. I'd designed a question that could go against me, and it did.

If you could only have one first

Five of the eight libraries chose the metrics column

5 EBSM metrics column first
3 Prioritise Items first

If you could only have one first
EBSM metrics column first	5 of 8
Prioritise Items first	3 of 8

One block per library, counted from my June 2024 research report.

An excerpt from the research report: 'Almost twice as many customers say they'd prefer the EBSM Metrics column than the Prioritisation Wizard, if they could only have one first.' The bullets note that customers struggled with the question, that the column applies across more reports, that it is visible rather than hidden behind a button, and that it explains recommended actions. The recommendation: deliver both longer term, but if resources only permit one, the EBSM metrics column first. — The finding, straight from my research report. The recommendation went against the feature I'd been building towards.

Testing

What testing changed

Every participant found the new button without being prompted, worked their way through the wizard, explained back what it was doing and applied the result. It passed, but it didn't come out of testing unchanged.

Several people tried to drag the metrics into order instead of using the up and down arrows, so drag and drop went in alongside them. The link for removing a prioritisation was too easy to miss, so it became a secondary button, and I added a second way to get to it from inside the wizard. And one report size across every branch turned out to be unusable, because branches differ in collection size and staffing, so administrators can now set sizes per branch.

An excerpt from the research report: 'However, several users expected to be able to drag and drop the three metrics within the priority order section of the wizard', above the prototype's ranked list of Grubby, In overstocked collection and Outdated with up and down arrows, and the recommendation to include drag and drop functionality. — One of the findings that changed the design. Drag and drop went in alongside the arrows.

A video call during a usability session. The Prioritise Items prototype is on screen, reading 'Your selections will leave 1,043 items'. Norrie's video is top right; the participant's video and name are blurred. — One of the moderated sessions on the prototype. The participant's face and name are blurred.

It calculates right at the bottom how many will leave in, because I can immediately see if I've limited it too far.
Usability participant, June 2024

A video call during a usability session. The prototype shows a breakdown: 49 of the requested 200 items will be picked in the following way, in rows of Dead, Outdated, Overstocked and Grubby tags, then 'That leaves 151 items to choose from'. The participant is blurred. — The breakdown, which shows how the 200 items you asked for would be picked across each mix of metrics.

Please launch this. It'll save me so much time from having to print off 30 page reports and use a hi-lighter to mark the items I think are most important.
Ascension Parish Library, Louisiana

The outcome

The outcome, and why there's no number

This is a difficult one to quantify. Librarians had to weed before this shipped, and they still have to now. So it didn't move adoption, it didn't change usage numbers and it didn't close any support tickets, because none of those were ever the problem. I'm not going to dress that up.

What changed is that the preparation went away. Nobody has to export a report and trim it in a spreadsheet any more. I only really have anecdotal evidence for that, which is customers telling us it saves them a lot of time.