QORPO Blog / Development

How to use Unreal Insights bookmarks to compare a repeatable hitch

Q
QORPO Studio16 November 2026 · 4 min read
Glass timeline bars with two brass timing markers
Concept illustration created for this article, not a gameplay screenshot.

Use Unreal Insights bookmarks to identify the same player action in each performance capture. Put a marker before the action you want to compare, then inspect the frames and thread activity around it. The bookmark gives you a place to start reading a trace. The timing data is what supports a diagnosis.

Pick a short sequence with a visible trigger

Imagine a model test where a player approaches an old gate, presses an interaction button and sees a brief hitch as the gate opens. Start from a known position with the same save and camera angle. Walk the same route, interact once and continue past the gate. The pass should be short enough that another person can repeat it without interpreting a long set of instructions.

Record what the player sees. “The camera pauses when the gate starts moving” is more useful than “the map runs slowly.” If you can repeat it only on the first use, record that too. A return through the gate and a fresh start can have different loading or cache conditions. The QORPO shader stutter checklist explains why those conditions belong in a comparison.

Place bookmarks around the action

Epic's trace quick start shows how bookmarks appear in Timing Insights. The Unreal Insights reference describes trace channels that can include bookmarks and frame data. Confirm that the build records the channels you need before committing to a long capture.

Choose markers that identify observations. In the gate model, one marker could identify when the interaction was requested. A second could identify when the gate began moving, if your team can instrument that separately. Use the same placement and names across builds. A bookmark attached to a timer or to a different point in the input path will change what the captured interval means.

If a second person is helping, agree on where they see the transition start. A markup in a clip and a marker in a trace are useful together if you know which game action they refer to. Document the point so future captures do not drift as the test route changes.

Inspect frames, then follow the work

Open the trace and find the marker. Look at the frames immediately before and after it. Compare which tracks contain long work and when that work starts. A long frame on the game thread, a long render thread interval and a delay before any visible activity each point to a different investigation. Treat the timing around the bookmark as a clue to follow.

The Unreal asset loading article gives a related way to map an input to when its assets are ready. If the hitch seems linked to loading, inspect the loading work and compare it to the action. If another thread is the one spending time, keep investigating that thread. A coincident marker alone cannot attribute the hitch.

Compare one change with the same capture conditions

Suppose the team moves a small asset preparation step earlier. Run the old and new build with the same route, device, quality settings and startup state. Compare the interval after the interaction and check the earlier part of the route too. The work may have moved to a different time, which matters if that new time is also visible to the player.

A second pass of the same build helps you see variation before interpreting a difference across revisions. If the result occurs inconsistently, keep representative captures of both outcomes. Note the exact part of the interaction where they diverge.

If the comparison follows work that starts earlier, add a marker to that earlier step in both builds. A capture that starts too late can hide the cost you moved. Keep enough lead-in and follow-through to show what changed around the action.

What should a reviewer receive?

Send the two build identifiers, the route, the marker names and the time range you inspected. Describe which tracks changed and what the player saw. If the evidence only covers one test machine or one saved position, say so. The goal is a trace someone can open, a moment they can find and a comparison they can repeat. See the QORPO development blog for more workflows built around packaged tests.