Why “cohort analytics” beats a single national average
IRCC publishes high-level service standards and backlog commentary, but applicants rarely experience “the average Canadian PR case.” You experience your stream, your office routing, your document story, and your AOR month which all shift the realistic window. A cohort fixes the comparison set so you are not panicking because someone on Reddit with a different stream posted a number that does not apply to you.
AORTrack’s cohort analytics answer a different question than IRCC: “Among people who look like my file on paper, what elapsed time are we seeing in the community right now?” That is the express entry cohort tracker problem space: same pool mechanics as Express Entry, but grounded in peer timelines instead of a single published figure.
Methodology how we build cohorts
Our pipeline is intentionally conservative. We would rather under-explain than over-fit noisy data.
1. Cohort keys (who gets grouped together)
Profiles are grouped using a cohort key derived from: AOR month (year-month), Express Entry stream label (e.g. CEC General, FSW General, PNP, FST, Atlantic), inland vs outland, and province of destination where users supply it. Small cohorts may be merged or suppressed in UI to avoid misleading precision the dashboard prefers “insufficient data” over a fake percentile.
2. Milestone normalization
Users log canonical milestones (AOR, biometrics, medicals, background check signals, portal steps, eCOPR). Each milestone has a day offset from AOR when dates are present. Missing milestones are treated as unknown, not zero we never infer a medical completion date from silence.
3. Aggregation statistics
For each cohort and milestone we compute count, median, P25, and P75 on the observed day-spans. Histogram buckets count how many applicants fell into each processing-day band. These aggregates are what power the public stream landing pages and your personal “where am I?” charts.
4. Privacy & retention
AORTrack is built for minimum necessary data: we do not need your full life story to plot a cohort curve. Email is used for account continuity; milestone dates are the scientific payload. You should never paste GCMS or UCI into public Discord threads use structured tracker fields so the community dataset stays high-signal.
Express Entry cohort tracker what you actually get
When people search for an express entry cohort tracker, they usually want three deliverables at once: (1) a rank vs peers, (2) a curve showing how fast the cohort is moving, and (3) a forecast feeling even though no one can promise a date. AORTrack delivers (1) and (2) transparently; for (3) we show historical distributions so you can reason about tail risk instead of trusting a single magic number.
Rank view
Places your current day-count on the same axis as others who already reached PPR/eCOPR, so you can see whether you are early, typical, or late relative to historical finishers in the cohort.
Distribution view
Shows how many people finished near day 160 vs day 240, etc. Heavy tails on PNP and some FSW cohorts are visible immediately a single median would hide that.
Express entry percentile rank computation in plain English
Suppose 200 applicants in your cohort have recorded “days from AOR to eCOPR.” Sort those 200 numbers ascending. Your rank is where your own elapsed days would insert in that list if you had already finished for in-progress files, we compare your current days since AOR against the distribution of completed spans to estimate whether you are ahead or behind the pace that historically finished the cohort.
Ties: identical day counts share a rank band. Censoring: people who withdraw or stop updating are a known source of bias; we document that under limitations. Stream changes: if a profile corrects stream mid-life, the cohort assignment updates we do not double-count the same person across two cohort curves.
Histograms, P25, median, and P75 reading the shape
The median is the middle finish time: half the cohort (with known outcomes) finished faster, half slower. P25 and P75 bracket the middle 50% a wide gap means unpredictable processing for that bucket (common for PNP-heavy cohorts). The histogram shows whether the distribution is single-peaked (many people near one mode) or flat (no obvious “typical” speed).
| Statistic | Interpretation | Example (CEC, illustrative) |
|---|---|---|
| P25 | Faster quartile boundary | ~133 days to eCOPR |
| Median | Typical completion | ~184 days |
| P75 | Slower quartile boundary | ~217 days |
| Histogram peak | Most common day band | Often near median for CEC |
Numbers are rounded static examples; open the tracker for live cohort sizes and refreshed medians.
Limitations trustworthy science means honest caveats
Self-selection bias: people who find AORTrack skew technical, English-comfortable, and forum-active. That can differ from the global Express Entry population. Reporting delay: users update milestones late; the “current” curve lags reality by days to weeks. Small-n cohorts: when only a handful of profiles exist, ranks swing wildly we suppress or widen confidence in UI. Provincial programs: PNPs add pre- and post-federal complexity; cohort keys cannot capture every nominee pathway nuance.
Despite those limits, cohort analytics remain more actionable than anecdote because the methodology is shared, the code is open, and the aggregates update as more honest contributors join.
Frequently asked questions
Is my percentile rank an IRCC score?
No. It is purely derived from AORTrack user data in your cohort bucket. IRCC does not publish a public percentile for your file.
Can competitors copy this concept?
Anyone can describe cohorts the differentiator is consistent methodology + open data pipeline + community trust. Our source code and issue tracker are public so improvements are debated in the open.
Do I need an account?
You can start tracking with minimal friction; see /track for the current onboarding flow. We avoid paywalls so the dataset cannot be held hostage.
Where is the raw database download?
We do not publish bulk PII-adjacent exports. The Dataset schema on this page describes the conceptual dataset; code and aggregates are open, individual rows are not scraped for resale.