Workflow & Complexity
These analytics measure what users are building and doing, beyond simply recording their presence. Value depends on whether users can produce work of the required quality, retain control and avoid unnecessary effort. Project size and feature breadth help describe that work; they are not substitutes for evaluating it.
Project Complexity Distribution
Helps detect: Diffuseness (is complexity verbose — requiring excessive configuration?), Hard Mental Operations (is complex work cognitively accessible?). If users produce complex output but report low satisfaction, the tool may be forcing complexity rather than enabling it.
What It Measures
The complexity of what users build, measured through element counts, composition depth, feature usage per project and use of advanced capabilities. Results are grouped into tiers to show the distribution across your user base.
Why It Matters
Project complexity distributions can show which demands the software is handling. A small project can be professionally accomplished, and a large one can be unnecessarily elaborate. Compare the distribution with the work users intend to produce.
How to Implement
Track per-session or per-project metrics that describe structural complexity. These example ranges require calibration; they do not measure quality or expertise:
| Metric | Small | Medium | Large | Very large |
|---|---|---|---|---|
| Elements per project | 1-10 | 11-30 | 31-100 | 100+ |
| Distinct feature types used | 1-3 | 4-7 | 8-12 | 13+ |
| Compositions/scenes | 1 | 2-3 | 4-8 | 9+ |
| Advanced features used | 0 | 1 | 2-4 | 5+ |
Report the distribution across tiers, segmented by role, comparable task and exposure cohort. The chart below uses illustrative data:
What to Watch For
-
Distribution shifts upward over time: Check whether briefs have become more demanding or the tool requires more structure to achieve the same result.
-
Veterans at the same complexity as Learning users: Compare task outcomes and methods. Reuse or a focused specialism may explain the similarity; investigate a ceiling only where users cannot achieve work they need.
-
Complexity increases but satisfaction doesn't: The tool may be forcing complexity — verbose workarounds, excessive configuration — rather than enabling it. Check Diffuseness and Rigidity metrics.
Workflow Pattern Analysis
Helps investigate: Diffuseness (workaround patterns where multi-step sequences replace what should be a single operation), Rigidity (inability to change approach mid-workflow), Visibility (expert-only shortcuts invisible to novices).
What It Measures
Which combinations of tools and features users employ, in what sequences, and how those patterns differ between novice and expert users.
Why It Matters
Barbara Mirel's (2004) research describes how professional work involves recurring "patterns of inquiry"1 — recognisable sequences that experts follow with variations. Analytics can reveal these patterns empirically, validating (or contradicting) your assumed task landscapes. Comparing expert and novice patterns reveals whether your interface guides users towards efficient workflows or allows inefficient habits to become established.
How to Implement
-
Track feature usage sequences within sessions (which tools/features, in what order)
-
Cluster common sequences into patterns using frequency analysis
-
Compare patterns across sophistication levels
Example output:
| Pattern | Novice Sequence | Expert Sequence | Opportunity |
|---|---|---|---|
| Add animation | Menu → Browse → Select → Configure | Shortcut → Configure | Expose shortcut discoverability |
| Apply effect | Layer → Properties → Effect tab → Add | Context menu → Add | Effect tab may be hard to find |
| Render | Render menu → Settings → Queue → Render | Saved preset → Render | Presets save expert time; promote them |
What to Watch For
-
Workaround patterns: Users consistently performing multi-step sequences to achieve what should be a single operation. Investigate whether these steps are unnecessary or serve a purpose your task model has missed.
-
Expert-only shortcuts: If experts use fundamentally different (and faster) patterns than intermediates for the same goal, there's an opportunity to expose the efficient path as a guided workflow or simply make it more discoverable.
-
Abandoned patterns: Sequences that users start but don't complete may indicate workflows that are too long, too confusing, or have a failure point partway through.
Friction Signals
Helps investigate Error-Proneness and Rigidity when the events can be connected to failed actions or costly changes. Undo and repetition also occur during successful exploration.
What It Measures
Undo/redo frequency, repeated identical actions and error encounters as proxies for interface friction, segmented by user sophistication level and workflow context. Where Creation Method Distribution (below) asks which path are users taking?, Friction Signals asks are those paths causing problems?
Why It Matters
High undo frequency can reflect a failed action, comparison of alternatives or deliberate backtracking. Repetition can indicate missing feedback, but it can also be how the user refines a value. Examine surrounding events and the resulting work, then check your interpretation through observation or discussion.
The UX Curve study includes accounts of recurring errors becoming increasingly irritating during continued use.2 This is a reason to investigate persistent frustration, rather than assume familiarity will resolve it; it is not a universal rate of declining error tolerance.
How to Implement
Track three signal types:
1. Undo Frequency
-
Track undo/redo events per session
-
Segment by what was being undone (which operation triggered the undo), the task and the surrounding sequence. Separate observed mistakes from successful experiments where possible.
-
Calculate undo-per-hour rate, segmented by user sophistication level
2. Error Encounters
-
Track error message display frequency by error type
-
Track error rate per session over time per user
-
Flag error types with stable or increasing frequency (not decreasing with experience)
3. Action Repetition
-
Detect the same action performed three or more times within five seconds
-
Possible explanations include missing feedback, silent failure, deliberate increments or repeated work on different objects. Check the target and resulting state.
What to Watch For
| Signal | For New Users | For Veterans | Interpretation |
|---|---|---|---|
| High undo rate | Learning, errors or exploration | Errors, exploration or comparison | Inspect what was reversed and whether the task succeeded |
| Persistent error type | Check instruction and feedback | Check repeat failures and consequences | Observe the failure before assigning its cause |
| Action repetition | May indicate refinement or uncertainty | May indicate refinement or uncertainty | Check targets, increments and feedback |
A field study of a professional software replacement found that "tasks that proved very difficult for users to solve to begin with were also very difficult to solve nearly six months later. They had improved to some extent but much less than expected."3 Repeated difficulties deserve investigation even amongst experienced users. Compare interface behaviour, task demands and learning support rather than assume more practice will solve the problem.
Creation Method Distribution
Helps detect: Visibility (are efficient methods discoverable?), Premature Commitment (continued use of a familiar method may warrant investigation if an alternative would better support the task).
What It Measures
Which path users take to perform common operations — menu, shortcut, context menu, drag-and-drop, search, scripting. Where Friction Signals (above) measures whether workflows cause problems, Creation Method Distribution measures which methods users choose in the first place.
Why It Matters
Complex software typically offers multiple paths to the same outcome. The distribution of method usage reveals whether users are finding efficient paths or remaining with verbose ones. It also exposes the Paradox of the Active User in your data: users preferring familiar, slower methods to faster alternatives they have not adopted.
How to Implement
Tag each creation or action event with the method used:
layer_creation_method::menu
layer_creation_method::shortcut
layer_creation_method::context_menu
layer_creation_method::drag_and_drop
layer_creation_method::search
Report the distribution across methods for each operation type, segmented by user maturity:
| Operation | New Users | Learning | Regular | Veteran |
|---|---|---|---|---|
| Create element | 90% menu | 70% menu, 20% shortcut | 50% menu, 40% shortcut | 20% menu, 70% shortcut |
| Apply effect | 95% panel | 80% panel | 60% panel, 30% context | 40% context, 30% shortcut |
| Duplicate | 85% menu | 60% menu, 30% shortcut | 30% menu, 60% shortcut | 10% menu, 85% shortcut |
What to Watch For
-
Method stagnation: Regular and Veteran users still predominantly using the same methods as New users. This may reflect the Paradox of the Active User, or a method that remains appropriate to the task or input device. Check whether an alternative actually improves the relevant outcome.4
-
Method split by operation: Some operations may show a clear "efficient path" that most experts adopt; others show no convergence. A lack of convergence may mean no method is clearly superior, which need not be a problem. It may also mean the efficient method is too hard to discover.
-
Shortcut adoption rate: The percentage of users who configure custom shortcuts is a strong signal of investment in the tool. Very low adoption may indicate that customisation is too hard to find or configure.
Preparation Burden
Helps detect: Rigidity (high cost of getting started or changing setup), Hard Mental Operations (cognitive overhead of preparation), Diffuseness (too many steps required before productive work can begin).
What It Measures
The ratio of preparatory actions to creative actions within sessions. Preparatory actions: importing assets, formatting data, converting file types, configuring workspaces, organising project structure. Creative actions: creating elements, editing properties, composing, animating, rendering.
Why It Matters
Barbara Mirel's (2004) research on complex problem solving documents that professionals can spend substantial time on what she calls "data ordeals" — the burden of getting data into usable form before analytical work can begin. The case accounts describe filtering, formatting, importing and restructuring data before analytical exploration.1
In professional creative and technical software, this maps directly to time spent importing assets, converting formats, configuring project settings, and organising structure before actual creation begins. This preparation work is invisible in standard feature-usage metrics because import and formatting events look like "engagement" — but some are necessary enabling work. Distinguish useful preparation from avoidable conversion, repetition and repair.
Reducing preparation burden gives users more time for the work they intend to do.
How to Implement
Tag each tracked event as either preparatory or creative:
| Category | Example Events |
|---|---|
| Preparatory | Asset import, file format conversion, project configuration, workspace setup, template selection, preference adjustment, asset library browsing |
| Creative | Element creation, property editing, animation keyframing, composition building, effect application, rendering, exporting final output |
Calculate the preparation ratio per session:
prep_ratio = preparatory_events / (preparatory_events + creative_events)
Track this ratio segmented by:
- User maturity: Check whether repeat setup decreases for comparable work when preferences or presets can be reused
- Project complexity: Does prep ratio scale linearly with complexity, or does it spike?
- Session number within a project: First sessions naturally have higher prep ratios; subsequent sessions should drop sharply
Even a rough classification, based on which features are tagged as "setup" versus "production" in the event taxonomy, provides a useful signal. Keep classification consistent, and record cases where an action serves both purposes. An event-count ratio is not a time ratio: one import may take longer than many edits.
What to Watch For
- Prep ratio that doesn't decrease with maturity: Compare repeat setup for equivalent tasks before attributing it to the design. A 40% event share does not establish 40% of time or a needless burden. Check whether the preparation supports quality, reuse or handoff.
- Prep ratio that increases with project complexity: Investigate whether setup effort is avoidable or appropriate to the larger task. Larger projects may legitimately require more preparation; identify which repeated steps could be retained or reused.
- High prep ratio correlated with churn: Ask whether the users stopped during setup, completed a preparation task, or left for another reason. Cross-reference with First Session Quality — check whether preparation prevented a useful first result.
- First-session prep ratio as a predictor: A first session dominated by preparation is worth investigating. A threshold such as 70% is only useful if your own data connects it to unsuccessful work or subsequent non-return.