How to trick WebWork Tracker and keep your activity level up

How WebWork Tracker counts activity

WebWork's own documentation describes a straightforward model. Within each active minute it adds up your keyboard, click and scroll actions and compares that count against a normative value the team selects. Generally the presets are Relaxed at 25 actions per minute, Normal at 50 and Strict at 100, with the resulting activity level capped at 100%. Roughly speaking, below 30% reads as Low, 30-60% as Medium, and above 60% as High.

Two things follow from this design. First, the level is per-minute and then averaged, so the daily figure is the mean of many small scores rather than one live reading, and a single quiet stretch pulls the average down more than most people expect. Second, because there is a cap, once you comfortably clear the normative value extra frantic input buys you nothing — the number rewards steady coverage far more than short bursts of intensity.

What actually feeds the number

Only a narrow set of signals moves the level, and knowing which ones matter is the whole game. Keystrokes count, clicks count, and scrolling counts. What WebWork records is that these events happened, not their content — it is measuring that input occurred, not logging the words you typed or the pages you read.

This is why raw mouse drift is a weak signal here. A cursor sliding back and forth generates movement, but the formula is built around discrete actions — keys, clicks, scrolls — so pure motion with none of those barely lifts the number. Everything else that real work involves, the reading, the thinking, the calls, is simply invisible to the metric.

• Keystrokes — any key, regardless of what you type.
• Clicks and scroll actions inside your apps.
• Not the content of your input, and not what is on screen.
• Not raw cursor drift on its own — the formula wants discrete actions.

What your manager actually sees

The activity level never travels alone. In the dashboard it sits beside the screenshots for the same period and the list of apps and sites used during tracked time, and a reviewer reads them as one record. A high level next to a screen that has not changed between two frames raises more questions than an honest, moderate number ever would.

That pairing is the reason chasing the percentage in isolation backfires. The figure is a coverage signal, not a productivity score, and when the picture beside it contradicts it — busy number, frozen screen — the contradiction is what a manager notices, not the number itself.

• Activity level per period plus daily and weekly reports.
• Screenshots at a frequency your team configures (some plans allow blurred frames or none).
• Apps and websites used during tracked time.

Why honest work can score low

Because only discrete input feeds the level, plenty of genuine work scores low for structural reasons. Reading a spec, watching a recorded call, reasoning through a bug, or sitting in a meeting produces almost no keys, clicks or scrolls, so a research-heavy day can land well below a data-entry day even though both are entirely real.

There is no universal 'good' number, because the honest baseline depends on how much of the job runs through the keyboard and mouse. Continuous-input roles like support, QA and data entry naturally sit high; build-and-think roles like development, design and analysis sit in the middle; and reading- or meeting-heavy days sit lower still. The useful comparison is against your own quiet weeks, not a colleague in a different kind of role.

Why a jiggler falls short here

A hardware dongle or a two-line script repeats one path forever, with no clicks, no scrolls and no pauses. Against WebWork's formula that is close to the worst possible input: it is the one signal the metric weights least, and it never produces the discrete actions the number actually rewards, so the level tends to stay stuck in the Low band no matter how long it runs.

It also fails the second test. The screen behind a jiggler never changes, so two identical frames half an hour apart, sitting next to a suspiciously flat activity line, read as automation rather than work. Vendors publish the patterns they flag and update them over time; no tool can promise a specific tracker will never notice a mechanical loop, and a jiggler gives it the clearest possible pattern to catch.

What reads as a person, not a loop

A human at the keyboard produces an uneven, mid-range line: input that mixes keys, clicks and scrolling, climbs during hands-on work and dips for calls and reading. A believable day for most roles lives somewhere in a natural band around 60-85%, not pinned at the ceiling. A level held flat at or near 100% for hours, day after day, is more suspicious than a mid-range number, because no human hand behaves that evenly.

If you want your report to read as real, the goal is variety and spread, not a maximum. Discrete actions of different kinds, arriving at irregular intervals with genuine pauses, along paths that are not identical every time — that is the shape of a person. A single repeated action at a metronome's rhythm is the shape of a script.

Where Reaction fits

Reaction sits in one narrow place: the stretches when you step away from the keyboard but the session should stay alive — the coffee, the stretch, the school run. It moves the cursor in smooth curves, types safe navigation keys, scrolls real documents and switches between the apps you already use, all with pauses, so the input reads as varied rather than mechanical. Crucially, you set an activity band instead of chasing 100% — a level pinned at the ceiling is the least believable number there is.

It starts only when the machine is genuinely idle and steps aside the instant you touch the mouse. What it does not, and cannot, do is make low-input work look like high-input work while you are actually at the desk doing it — that is a limitation of the metric, not a feature to sell. All it does is stop a normal break from reading as an absence. Use it within your agreements; see the acceptable-use note below.