Skip to main content

Task Success Rate

Share of users who started a task and reached its finishing point, from GA4 funnels.

The proportion of users who started a given task in CanadaLogin and reached that task’s defined end point, measured from Google Analytics 4 funnel data.

This is a family of metrics, not a single number. Each funnel declares its own task success rate as a step_ratio in its YAML definition on GitHub. The entry and exit steps for each funnel are described below as of this writing, but the YAML is the canonical, versioned source - see funnel_src/funnels/ in the canadalogin-reporting-data-pipeline repo.

Only sign-up, password recovery and the two migration tasks end on a screen that confirms the task is done. Sign-in cannot, and the 2-step recovery exit is a later cleanup screen. See the task notes before comparing tasks or reading any one rate as a completion rate.

Calculation

Each funnel’s task success rate is already computed and stored as one row per window in funnel_metrics: the exit step’s activeUsers divided by the entry step’s, for that funnel run. There is no ratio to compute by hand - filter to the funnel and metric, and read value. Query funnel_metrics_current rather than funnel_metrics directly, so a future funnel version bump can’t silently blend two definitions together. Filter breakdown_value = 'RESERVED_TOTAL' for the overall rate across all relying parties; substitute a specific relying party name for a per-RP figure. The funnels catalog entry explains the RESERVED_TOTAL and window_incomplete conventions.

For one task, set funnel_id to the value given in that task’s notes.

One task:

SELECT
    window_end,
    value AS task_success_rate
FROM google_analytics.funnel_metrics_current
WHERE funnel_id = 'sign_in'
  AND metric_id = 'task_success'
  AND window_days = '7'
  AND breakdown_value = 'RESERVED_TOTAL'
  AND NOT window_incomplete
  -- set the reporting window; see the timespans note in cookbook/notes.qmd
  AND CAST(window_end AS DATE) BETWEEN DATE '2026-07-14' AND DATE '2026-08-03'
ORDER BY window_end;

Every task:

SELECT
    funnel_id,
    window_end,
    value AS task_success_rate
FROM google_analytics.funnel_metrics_current
WHERE metric_id = 'task_success'
  AND window_days = '7'
  AND breakdown_value = 'RESERVED_TOTAL'
  AND NOT window_incomplete
  -- set the reporting window; see the timespans note in cookbook/notes.qmd
  AND CAST(window_end AS DATE) BETWEEN DATE '2026-08-03' AND DATE '2026-08-31'
ORDER BY funnel_id, window_end;

One task:

tbl(con, in_schema("google_analytics", "funnel_metrics_current")) |>
  filter(
    funnel_id == "sign_in",
    metric_id == "task_success",
    window_days == "7",
    breakdown_value == "RESERVED_TOTAL",
    !window_incomplete,
    # set the reporting window; see the timespans note in cookbook/notes.qmd
    as.Date(window_end) >= as.Date("2026-07-14"),
    as.Date(window_end) <= as.Date("2026-08-03")
  ) |>
  select(window_end, task_success_rate = value) |>
  arrange(window_end)

Every task:

tbl(con, in_schema("google_analytics", "funnel_metrics_current")) |>
  filter(
    metric_id == "task_success",
    window_days == "7",
    breakdown_value == "RESERVED_TOTAL",
    !window_incomplete,
    # set the reporting window; see the timespans note in cookbook/notes.qmd
    as.Date(window_end) >= as.Date("2026-08-03"),
    as.Date(window_end) <= as.Date("2026-08-31")
  ) |>
  select(funnel_id, window_end, task_success_rate = value) |>
  arrange(funnel_id, window_end)

Interpretation notes

The queries above use window_days = '7', a rolling weekly rate recomputed nightly on a 7-day window ending each date. Three window lengths exist, and switching between them is a one-line change:

  • '1' is the same metric on a single day. It is the noisiest of the three, but it is the only one that shows the exact date something changed.
  • '7' is the usual default. Every window holds one of each weekday, so day-of-week effects wash out.
  • '28' is a rolling four-week window. It is the steadiest of the three and the one to use for monthly reporting.

Windows overlap, so consecutive rows are not independent. One unusual day shows up in 7 weekly rows and 28 monthly ones. A NULL rate means nobody reached the entry step in that window. These rows are rare. Drop them rather than counting them as 0.0, or an average will read too low.

Task notes

All tasks

  • Users, not attempts. Every step counts distinct activeUsers in the window. A user who fails, retries and succeeds counts once, as a success. A user who fails three times and gives up counts once, as a failure. The rate is a share of people, not of attempts.
  • Traffic to the Manage CanadaLogin app is excluded. Someone can go through sign up or sign in without coming from a valid relying party, using a link that names the Manage CanadaLogin app as the destination. Real users always arrive from a valid relying party, so this traffic comes from employees, not actual users. Every funnel that is not a Manage CanadaLogin task therefore excludes traffic with the Manage CanadaLogin relying party ID. See Internal CanadaLogin components.
  • A series can change definition part way through. funnel_metrics_current returns the newest version for each date, and a new version is not always backfilled. Check funnel_version across the rows before reading a trend.

Sign in

  • Funnel ID: sign_in
  • Definition: sign_in.yaml
  • Entry: enter_email, the “Enter your email” screen
  • Exit: mfa_submit, pressing submit on the second-factor form

This rate measures submitting a second factor, but not necessarily success. The exit is pressing the submit button on the second-factor form. It is possible to submit an incorrect OTP code and count as a success in the task but fail to actually sign in. This is a limitation of what Google Analytics measures. This means the rate is marginally overstated, compared to a true sign-in success rate.

For an authentication success rate measured by IBM Verify itself, see Authentication Success Rate. It counts attempts rather than users and covers every sign-in, not only those that started in a tagged browser session, so the two numbers will not match.

Sign up

  • Funnel ID: sign_up
  • Definition: sign_up.yaml
  • Entry: terms, the Terms and Conditions screen
  • Exit: complete, the “Your CanadaLogin is complete” screen

Recover password

  • Funnel ID: recover_password
  • Definition: recover_password.yaml
  • Entry: reset_password, the “Reset your password” screen
  • Exit: complete, the “Your password has been reset” screen

Account checkers are excluded from version 5. Many people who start are checking whether they have an account, not recovering a password. They enter their email, enter the OTP code sent to them, and get error 26 (“This email is not associated with an account”). Counted as failures, they would make the rate mostly a measure of account checking. The funnel drops any user with a form_api_call_end event where error_id is 26 and form_id is passwordrecovery, using a GA4 user exclusion segment (the exclude_users block in the YAML). An excluded user leaves every step, not only the step where the error appeared.

The rate is a lower bound. Only the phone branch returns error 26. The passkey branch never looks up the email, and an address with no account fails the passkey prompt the same way a cancel does. Those checkers, and checkers who quit at the method choice screen, stay in and count as failures, even though they had no password to reset. The true success rate for account holders is higher.

Recover 2-step verification

  • Funnel ID: recover_2sv
  • Definition: recover_2sv.yaml
  • Entry: initiate_recovery, the “Initiate email verification” screen
  • Exit: confirmation, the “Delete sign-in methods you’re not using anymore” screen

Exit is a cleanup screen, not a confirmation. The last step is “Delete sign-in methods you’re not using anymore”, offered after a new method is set up.

Migration

Existing account:

  • Funnel ID: migration_existing_to_cl
  • Definition: migration_existing_to_cl.yaml
  • Entry: migration_not_skipped, answering the legacy sign-in prompt without skipping
  • Exit: migration_complete, the “Migration - Confirmation” screen

New account:

  • Funnel ID: migration_new_to_cl
  • Definition: migration_new_to_cl.yaml
  • Entry: migration_not_skipped, answering the legacy sign-in prompt without skipping
  • Exit: migration_complete, the “Migration - Confirmation” screen

The denominator is not the first step. In both funnels, task_success divides by migration_not_skipped, users who answered the legacy sign-in prompt without skipping it. The rate is the share of people who chose to link a legacy method and finished.

The prompt step has a time limit in migration_existing_to_cl. The legacy method prompt must follow the 2-step verification step within 30 minutes.

Alternative names

  • funnel completion rate
  • task completion rate
  • funnel success rate