Overview
Campaign Monitors is an Extole extension that checks the health of your referral programs on a schedule and sends a notification when a metric moves in a way the monitor considers unusual. It catches tracking breaks, stopped feeds, reward failures, and sudden shifts in participation, but it works from arithmetic on report data, so some alerts describe a change in what is being counted rather than a change in what your customers did. This page explains what each monitor measures, when it alerts, and what to check when one fires. Extole installs Campaign Monitors on your account under Partners > Extensions, where it appears as Campaign Monitors in the Utilities category. Each monitor is a component of the extension with its own settings. The default values below apply unless your Extole team or an admin on your account changed them. Alerts arrive as My Extole notifications. See Subscribing to My Extole Notifications to choose who receives them and through which channels. Extole’s support team receives the same alerts and may contact you about one.How a Monitor Works
Every monitor is a scheduled report plus an evaluation of that report’s results:- A report runner builds a series for each metric over the trailing quarter, one value per period. The period is a week for weekly monitors and a day for daily monitors.
- When the report completes, the monitor evaluates every series against its checks and settings.
- For each metric that fails a check, the monitor sends one notification naming the metric, the period, the values it compared, and the check that fired.
- The whole quarter is re-evaluated on every run. A spike or a stall that happened weeks ago stays in the series, and some checks fire again on each run until that period ages out of the window. A repeat alert for a condition you already understand is the same condition, not a new one.
- A condition that persists re-fires. A metric that dropped and stayed low alerts again on the next run where the check still holds.
Program Performance Overview
The Program Performance Overview monitor watches the referral funnel. It runs weekly, and also daily when Enable Daily Monitoring is on, which is the default. It reads production events on your live, paused, ended, and stopped campaigns, and skips campaigns taggedstaging, makegood, or snoozed.
It evaluates two scopes:
- Program level: every metric on each program’s dashboard funnel, such as promotion clicks, shares, share clicks, signups, and conversions. This includes metrics that count unwanted activity, such as blocked or test users, because the program level has no exclusions.
- Account level: metrics computed across all programs. The monitor alerts on a decrease in Outcomes or NTF Customer Contribution Rate (%), and weekly on an increase in Non-Production Steps or Bot Steps Rate (%). It never alerts on All Steps, All Production Steps, Production Steps Rate (%), All New Customers, New Attributed Customers, All Blocked Steps, Blocked Steps Rate (%), All Bot Steps, All Test Steps, or Test Steps Rate (%), which it reads only as context.
- Outcomes counts the attributed events that the campaign marks as outcome events. Which events carry that mark is campaign configuration. When a campaign is replaced or republished and the outcome mark moves to a different event, the Outcomes series changes on the publish date even though customer behavior did not.
- NTF Customer Contribution Rate (%) is New Attributed Customers as a percentage of All New Customers. On a period with no new customers at all, the rate reads 0%.
The Checks
Before the checks run, the monitor prepares each series:
- A series whose average is below the Low Data Volume Threshold is combined into larger buckets of Periods Combined for Low Volume periods, so a few events in a quiet week do not read as noise.
- A series with fewer periods than Minimum Data History is not evaluated.
- When Enable Seasonal Adjustment is on, a repeating pattern of the expected length is removed before the significant-change check. It is on for daily runs, which have a weekly pattern, and off for weekly runs.
Daily Monitoring Differences
The daily monitor uses the same settings with their daily values, plus two behaviors of its own:- Dropped to zero is tiered by volume. The monitor averages the days before the zeros began. If that average is five or less, zeros never alert. If it is between six and 20, the monitor alerts once, on the third consecutive zero day. If it is 21 or more, it alerts once, on the second consecutive zero day.
- A long zero streak is treated as expected. A metric that has been at zero for more than 14 consecutive days is left out of evaluation, so a step you stopped using does not alert every day.
Settings
Settings marked basic appear on the monitor by default; the others appear under its advanced settings.What a Program Performance Alert Looks Like
A notification is named<account>: Unusual traffic decrease alert for <program> or <account>: Unusual traffic increase alert for <program>. An account-level alert has no program suffix. The message names the metric, the check that fired, the period that ended, and the series the monitor compared, for example a list of the previous weekly values followed by the latest value.
Reward Rates
The Reward Rates monitor watches reward volume and reward health weekly. For each period it reads the counts of production rewards earned, fulfilled, sent, failed, and canceled, and the same counts for non-production (test and bot) rewards.
Because the baseline is the average of the earlier periods, a decline that happened gradually over several weeks still trips the large-decrease check once the latest period is far enough below that average, and it trips again every week while the level persists. Reward volume follows the funnel, so a drop in rewards earned usually traces back to a drop in shares or conversions that began earlier.
Other Monitors
The extension carries five more monitors. Each has a short description and its settings on its own page in the extension.Reading an Alert
Start from the report the alert links to, and compare the values in the alert with the metric’s own dashboard before you change anything. These patterns account for most alerts that do not need a fix:- A repeat of a known condition. Check the alert’s history for the same program and metric. If the same condition has alerted before and was already explained, nothing new happened. The alert stops when the period that caused it leaves the trailing quarter.
- A program-level alert on blocked, bot, or test counts. A metric such as users blocked can stall at zero or spike after one week of blocked traffic. The monitor treats it like any other program metric. Your referral traffic is unaffected.
- A rate at 0% because both counts were zero. On a day with no new customers at all, NTF Customer Contribution Rate (%) reads 0%. The question is whether your customer events reached Extole that day. Compare your source system’s count for that period with the Events report.
- A metric whose definition changed. When a campaign was replaced or republished near the first bad period, compare which events carry the outcome mark in the old and new versions. An Outcomes series that moves on a publish date is a definition change. Check the conversion counts directly to see what customers did.
- A stalled metric. Identical values period after period usually mean nothing new is being recorded: an upload that stopped, an integration that froze, or a step the program no longer uses.