Case study · Marking

Smart Comments: reusable marking feedback that knows which criterion it belongs to

Not a satisfaction problem — a migration blocker. Three universities wouldn't adopt Cadmus Marker without a QuickMarks equivalent. Parity was the commercial requirement; I added two things Turnitin doesn't do while meeting it. Sold into all three.

My role
Product Manager — competitor teardown, pitch, design direction
Commercial outcome
Cadmus Marker sold into 3 universities
Reach
All 12 Cadmus Marker tenants
Beyond the brief
Parity, plus two additions of our own
The problem

Markers write the same sentence a hundred times a term

Cadmus Marker had an adoption problem with names attached: Queen's University Canada, the University of Western Australia, Zayed University.

Why it stuck

All three marked on Turnitin Feedback Studio, and their academics had built their working habits around it.

The actual evidence

Markers wouldn't move to a tool that made their job slower

QuickMarks weren't peripheral — they were load-bearing. Switching meant going back to typing the same feedback repeatedly, on cohorts of hundreds.

They said no — not to Cadmus, but to the migration. Internally that read as a satisfaction gap. It was a blocked sale.

The reframe that mattered: parity is usually a vanity metric — a box ticked because a competitor has it. Here parity was the deal. Until markers could reuse comments, Cadmus Marker was a downgrade from what they already had, and no amount of product elsewhere compensated.

Not because reusable comments are interesting — because three named universities couldn't be onboarded without them. That turns a marking convenience into a revenue dependency.

Median time to mark

Defined in the pitch, before build. Median time to mark should fall — normalised against word count and rubric criteria, so a slow script isn't credited to the feature.

Competitor teardown

I marked with the competitor's version before designing ours

Teardown

Parity meant knowing exactly what academics were being asked to give up. I marked with QuickMarks the way they would — which is where the opening showed up.

Turnitin Feedback Studio · QuickMarks

Curated sets, managed separately from marking

  • Reusable comments live in QuickMark sets — curated collections built and managed as their own artefact.
  • The set is built ahead of the cohort, as a separate task from marking it.
  • The model assumes you know your comments before you meet the cohort.
  • The practical result: the comment you need mid-script often isn't in the set, so you type it and move on.
What we built

Any comment, promoted in-flow, scoped to rubric & score

  • A marker clicks Add to Smart Comments on a comment they've just written, without leaving the script.
  • No upfront curation step. Reusable comments accumulate from feedback the marker was writing anyway.
  • Reuse is scoped to the rubric and score the comment was written against, so suggestions arrive in context.
  • A marker can also mark comments for reuse deliberately, or remove one — so the set stays theirs to control.
Turning point

Matching QuickMark sets alone would have unblocked the three sales — and fit comfortably inside the appetite. Which is why what came next was a choice, not a necessity.

What we added on top

Two workarounds that made ours better than the thing we were matching

Two frictions became obvious, and neither needed a bigger project — just a different placement. Both went into the pitch as additions to the parity scope:

1 — Convert a comment into a smart comment while marking. The comment a marker needs is the one they just wrote, not one predicted a week earlier. Reuse costs nothing to set up.

2 — Attach it to the rubric criterion it was written against. A flat pool gets worse as it grows. Tying each to its criterion and score band means suggestions arrive pre-filtered.

Where they came from

Neither was in the brief. Both came from using the competitor's product rather than reading its feature list — the difference between “we have this too” and “ours is easier”.

1 · The comment a marker just wrote
Rubric criterion with a marker comment and a Convert to smart comment action
1
2 · Now reusable, still in context
The same comment after conversion, now labelled as a smart comment
2
1Convert to smart comment — one action on a comment that already exists. No separate management screen, no upfront curation.
2Once converted it's reusable, and it stays attached to Evaluation of Group Process & Collaboration at the 50–64% PASS band — so it comes back only where it's relevant.

Sandbox tenant with a test rubric.

How it works

Four steps, two of them ours

Step 01
Write feedback

The marker comments on a script exactly as they would have anyway. No new habit required.

Step 02 · added
Convert it in place

One action turns that comment into a smart comment. No separate manager, no upfront setup.

Step 03 · added
It stays with its criterion

Stored against the rubric criterion and score band it was written for, not into a flat pool.

Step 04
Reuse in context

Next time that criterion is being marked, the relevant comments are the ones offered.

Marking view — smart comment actions on the rubric
MFA marking view with the rubric menu open, showing Pre-load Smart Comments and Import Smart Comments
1
1Pre-load and Import — reuse surfaced where the rubric lives, not in a separate management area.
Capture, in place
Rubric criteria panel with a comment field and a Browse smart comments control
2The comment box sits directly under the criterion being scored — so a comment already knows its context when it's promoted.

Sandbox tenant. Criteria shown are placeholders from a test rubric.

Read the steps

1 and 4 are parity. 2 and 3 are the additions — the reason a marker gets value on their second script rather than after an afternoon of setup.

Scope

What shipped, and what I deliberately held for v2

The temptation here is to ship a comment library. I explicitly didn't.

The scoping decision I'd defend hardest

Build a way to reuse comments — not a comment library

Collections, sharing, permissions, curation, a management surface — the entire appetite spent before a single marker saved a keystroke. The pitch stated the exclusion outright, while requiring the implementation stay compatible with one later.

Naming a future you're deliberately not building stops a minimal v1 being mistaken for a short-sighted one — and gave engineering its one constraint: don't foreclose the library.

Shipped in v1 live

  • Add to Smart Comments — promote any comment written during marking, without leaving the script.
  • Scoped reuse — comments come back in the context of the rubric and score they were written against.
  • Marker control — mark a comment for reuse, or remove it from the reusable set.

Held for v2 next

  • Pre-selecting comments per assessment — narrowing which stored comments are offered, so the suggestion set stays relevant as the library grows.
  • Importing legacy comment banks — bringing in exported QuickMarks files from Turnitin, so institutions don't abandon years of accumulated feedback to switch.
Fast-follow

Narrowing what gets offered per assessment stops the set becoming a long list — but it isn't what unblocks the first marker. Named, not built.

Impact

Three sales, and platform-wide reach

3universities
Queen's University, Canada University of Western Australia Zayed University

Cadmus Marker sold into all three once the blocker was removed — and the two additions made the migration an upgrade on what they were leaving.

12 / 12
Now live for every Cadmus Marker tenant, not only the three that needed it. Because there's no curation step to complete first, there's no adoption cliff — a marker gets value the second time they'd have retyped something.
Live across all 12 Marker tenants 3 universities unblocked & sold Parity met, plus 2 additions Baseline not captured pre-launch — time saving not yet measurable

What I'd do differently

The measure is right and still unproven — a reduction needs a before and an after, and the after doesn't exist yet. I can show the commercial outcome and the reach, not the time saved.

What I'd repeat

Three things.

1 · Reframe

A “satisfaction” complaint was a blocked sale. Three institutions stalling on one gap is a commercial signal, not a feature request — framing it that way got it prioritised.

2 · Use, don't read

A feature comparison would have said Turnitin “has reusable comments” and we'd have shipped a slightly nicer version of the same model.

3 · Separate the two

Parity was the commitment; the additions were named as extras. Keeping them distinct let the enhancements be argued on merit instead of quietly inflating a parity promise.