Skip to main content

Calls per Active User

Call volume relative to the active user base, comparable across weeks.

Call centre volume divided by the size of the active user base, so you can compare support demand across weeks even as the user base grows.

Raw call volume moves partly because the number of people using CanadaLogin changes. Normalizing by active users takes that out and leaves how often a typical user needed to phone in on a given day. The figure is scaled to calls per 100 active users to keep it in a readable range.

Calculation

For each call centre reporting week, take calls_accepted from call_centre.weekly_activity_report, subtract that week’s test_calls, and divide by the total user-days in that week, then multiply by 100. User-days is the sum of ibm_verify_events_transformed.events_auth_total_logins.unique_users over the days whose date falls inside the week’s date_range_start to date_range_end, so the numerator and the denominator cover the same span of days. A weekly call count over weekly user-days is a per-day rate. The two tables store their dates as text, so both sides of the join are cast to DATE.

Use calls_accepted, not calls_answered: accepted counts everything that reached the queue, so an abandoned call still counts as demand. test_calls is a subset of accepted and is large enough at current volumes to be worth subtracting.

SELECT
    w.date_range_start,
    w.date_range_end,
    CAST(w.calls_accepted AS BIGINT) AS calls_accepted,
    CAST(COALESCE(w.test_calls, 0) AS BIGINT) AS test_calls,
    CAST(w.calls_accepted - COALESCE(w.test_calls, 0) AS BIGINT) AS calls_excl_test,
    AVG(a.unique_users) AS avg_daily_active_users,
    CAST(w.calls_accepted - COALESCE(w.test_calls, 0) AS DOUBLE)
        / SUM(a.unique_users) * 100 AS calls_per_100_dau
FROM call_centre.weekly_activity_report AS w
JOIN ibm_verify_events_transformed.events_auth_total_logins AS a
    ON CAST(a.date AS DATE)
       BETWEEN CAST(w.date_range_start AS DATE) AND CAST(w.date_range_end AS DATE)
GROUP BY w.date_range_start, w.date_range_end, w.calls_accepted, w.test_calls
ORDER BY w.date_range_start;
weekly <- tbl(con, in_schema("call_centre", "weekly_activity_report")) |>
  select(date_range_start, date_range_end, calls_accepted, test_calls)

dau <- tbl(con, in_schema("ibm_verify_events_transformed", "events_auth_total_logins")) |>
  select(dau_date = date, unique_users)

weekly |>
  cross_join(dau) |>
  filter(
    as.Date(dau_date) >= as.Date(date_range_start),
    as.Date(dau_date) <= as.Date(date_range_end)
  ) |>
  group_by(date_range_start, date_range_end, calls_accepted, test_calls) |>
  summarise(
    avg_daily_active_users = mean(unique_users, na.rm = TRUE),
    user_days = sum(unique_users, na.rm = TRUE),
    .groups = "drop"
  ) |>
  mutate(
    test_calls = coalesce(test_calls, 0),
    calls_excl_test = calls_accepted - test_calls,
    calls_per_100_dau = calls_excl_test / user_days * 100
  ) |>
  arrange(date_range_start) |>
  select(
    date_range_start,
    date_range_end,
    calls_accepted,
    test_calls,
    calls_excl_test,
    avg_daily_active_users,
    calls_per_100_dau
  )

Interpretation notes

Both sides are per-day averages over the same Sunday to Saturday week: average daily calls over average daily active users. On a typical day, roughly this many calls came in per 100 active users. It can also be shown as a percentage of active users, e.g., 2 calls per 100 users is 2%, though it counts calls rather than callers.

Watch the units. Dividing a whole week of calls by one day’s active users reads about seven times too high. Normalizing by user-days avoids that and holds up for a partial week, since the day count comes from the data rather than a hardcoded 7. avg_daily_active_users is output for auditing only; the rate comes from the sum.

A call centre week runs Sunday to Saturday, from the report’s own date_range_start and date_range_end. The IBM Verify metrics anchor weeks on Monday, so the two do not line up; the join and the denominator follow the call centre boundaries.

A week appears only once daily active user data exists for it, so the series starts near the April 22, 2026 launch.

Alternative names

  • call centre contact rate
  • normalized call volume
  • support contacts per active user
  • calls per 100 users
  • support demand