Conditions: metrics, windows, attribution and comparisons

Conditions describe what must be true about an object for the rule to fire. For example: “spend today is above 60 USD, and there are no deposits from today's clicks”. They are configured at the builder step “Conditions”. Each condition is checked separately for each object in the scope (see “Scope”).

How the “Conditions” step works

1. At the top are the defaults for the whole rule: “Default window”, “Default revenue attribution” and “Exclude last hours”. Conditions that do not have their own value inherit them.

2. In the middle is the condition tree: individual conditions and “AND”/“OR” groups.

3. At the bottom is “Minimum volume”: a threshold below which the object is not evaluated at all.

A new rule starts with the default window “Yesterday” and one condition “Spend greater than 0”. Replace it with your own.

On the right, the builder shows the rule in one sentence, for example: “CPD over 3 d before today (click cohort) > $25.00”. Check against it to make sure you set exactly what you intended.

Condition tree: “AND” and “OR”

• “Add condition” adds a single condition. When there are two or more conditions, they are combined into a group.

• “Add group” adds a nested group of two conditions.

• A group has a type: “All (AND)” — all conditions of the group must be met, or “Any (OR)” — one is enough.

• A group must have at least two elements. If only one is left, the builder asks you to add another one or remove the group.

Example of a tree “expensive deposits or none at all, with noticeable spend”:

• All (AND)

  • Spend · Last 3 days (excluding today) · greater than 90 USD

  • Any (OR)

    • CPD · Last 3 days (excluding today) · cohort · greater than 45 USD

    • Deposits · Last 3 days (excluding today) · cohort · equals 0

“Yes”, “no” and “unknown”

Each condition has three possible answers: “yes”, “no” and “unknown”. “Unknown” means there is not enough data to decide. It is not the same as “no”.

The rule acts only if the result of the whole tree is “yes” and the object has passed the minimum volume. With “unknown”, the rule does not touch the object. In the preview and on the “Why not now?” log tab, the reason is “Not enough data to decide”. This way the rule does not make decisions based on data that does not exist.

Examples for the rule “Spend · Today > 60 AND Deposits · Today = 0”:

• spend 75 (“yes”), deposits 0 (“yes”) — the result is “yes”, the rule fires;

• spend 40 (“no”) — the result is “no” whatever the deposits show: in an “AND” group one “no” is enough;

• the ad set has no spend data for today yet — the result is “unknown”, the rule does nothing.

When a condition gets “unknown”

“No data” is not zero

• Facebook metrics (spend, impressions, clicks and everything calculated from them) are known only if the object has spend data in the window. If the ad set did not deliver during the window or the sync has not caught up yet, the value is “no data”, not 0. The rule cannot tell one from the other. Therefore the condition “Spend · Today < 5” will not catch an ad set that did not spend at all today.

• Tracker metrics (installs, registrations, deposits, revenue) are calculated honestly when there is traffic: if there were clicks but no deposits, “Deposits” = 0 and ROI = −100 %. The stop-loss “spend above X and 0 deposits” relies on this.

• Derived metrics with zero in the denominator are “no data”. The CPD of an ad set without deposits is unknown, not infinitely large. Therefore the condition “CPD > 45” does not fire on an ad set that has no deposits at all. To catch those too, add an “OR Deposits = 0” branch, as in the tree example above.

What a single condition consists of

The condition fields go in this order: metric → window → attribution → exclude last hours → operator → compare with. Attribution and hour exclusion appear only if the condition reads tracker conversions.

1. Metric

Metrics are grouped: “Spend and delivery”, “Tracker conversions”, “Derived metrics”, “Object attributes”. The list has a search, and only metrics available at the rule level are shown. For example, “Learning phase” exists only for ad sets. What each metric calculates is described in “Metrics and attributes reference”.

Object attributes (status, budget, budget type, “hours since …”) have no window: the current value is compared, with the caption “Current value of the object, without a window” next to it.

2. Window

A window is the period over which the metric is calculated. All windows are calculated in the ad account time zone. The “As in rule” option takes the default window.

• “Last 1 day” with “Including today” is the same as “Today”, and without it — the same as “Yesterday”.

• The window start is rounded up to a full hour: data within an hour is not split.

• “Lifetime” for tracker metrics is limited by the conversion history — about 190 days. If the object started spending earlier, a lifetime condition on conversions, revenue, ROI, ROAS, CPD or CPI will be “unknown”, and the preview warns: “Some objects are older than the tracker conversion history…”. Spend-only conditions are not limited by this.

• “Since budget change” can be selected only in the condition itself: it cannot be the default window or the window on the right-hand side of a comparison. Ads do not have it. If the object's budget has not changed since the ad account was connected, the window start is unknown and the condition will be “unknown”. In an “AND” group the rule will not touch such an object, and in an “OR” group the other conditions decide.

3. Attribution

Attribution answers the question of which time a deposit (registration, revenue) is assigned to. It exists only for metrics that involve tracker data. Spend, impressions, clicks, CPC, CPM and CTR have no attribution: they are always calculated by delivery hour.

Cohort in simple terms. The click happened on Monday at 23:40, the deposit arrived on Tuesday at 03:10:

• by event time, the deposit goes to Tuesday;

• by click date — to Monday, that is, the day when the click was paid for.

Therefore the cohort ROI for Monday honestly compares Monday's spend with what it brought, even if the deposits arrived on Tuesday or Wednesday. The flip side: the cohort “matures”. On Tuesday morning, the ROI of Monday's cohort is still lower than it will be on Wednesday, because some deposits are on their way.

Options in the condition's “Attribution” field:

• “Metric default: …” — if the rule has no attribution of its own. CPD, ROI, ROAS and profit default to cohort, other tracker metrics to event time;

• “As in rule: …” — if attribution is set at the rule level;

• “By event time” or “By click date (cohort)” — explicitly for this condition.

Attribution also appears when the tracker is involved only on the right-hand side. For example, in the condition “Spend greater than 2 × CPD”, the attribution applies to CPD.

4. Exclude last hours

The latest hours are not complete yet: deposits and revenue keep arriving. The “Exclude last hours” field (0–72 h) moves the right edge of the window N hours back from the current moment. The shift applies to all metrics of the condition, that is, to spend, to revenue and to the right-hand side of the comparison. The “As in rule: N h” option takes the rule's “Exclude last hours” value.

• “Today” at 15:30 with a 4 h exclusion — hours from 00:00 to 11:00 are counted (the incomplete hour is dropped).

• “Yesterday” and “N days excluding today” end at midnight anyway, so the exclusion affects them only early in the morning. At 03:00 with a 6 h exclusion, yesterday is counted until 21:00.

• After the exclusion, the window must still have at least 2 hours left. Otherwise the builder shows “Excluded hours leave the window empty”:

  Window	Maximum excluded hours
  Today	22
  Yesterday	46
  Last N hours	N − 2
  Last N days, including today	24 × N − 2
  Last N days excluding today	24 × (N + 1) − 2
  Lifetime, Since budget change	up to 72

Important. For the “Today” window with an N-hour exclusion, the window is empty for roughly the first N hours of the day. During this time the condition is “unknown”, and the minimum volume is “Below the minimum volume”. The rule simply does not act until morning.

5. Operator

List attributes (status, delivery status, budget type, learning phase) have only “equals”.

6. Compare with

The “Compare with” field offers several options for the right-hand side. The set depends on the metric.

Value. A plain number in the metric's units.

• “CPD · Last 3 days excluding today · cohort · greater than 25” — more expensive than 25 USD.

• “ROI · Yesterday · cohort · less than −20” — ROI below −20 %.

• “Delivery status · equals · Paused”.

Range. Appears with the “between” and “outside the range” operators: two fields, “From” and “To”. “From” must be less than “To”.

• “ROI · Yesterday · cohort · between 0 and 30” — the object is in profit, but no more than 30 %.

• “CPM · Today · outside the range 5–40” — a suspiciously cheap or expensive impression.

Another metric × multiplier. Expected value = multiplier × the selected metric. Fields: “Whose metric” (“Same object” or “Parent”), “Metric”, “Window” (the condition window by default), “Multiplier” (greater than 0, up to 1000). Only metrics in the same units as the left-hand side are offered: money with money, ratios with ratios.

• Same object: “Spend · Today · greater than 2 × CPD · Last 7 days excluding today”. The rule fires if today the ad set spent more than two of its usual costs per deposit.

• Parent: “CPI · Last 3 days excluding today · greater than 1.3 × parent CPI”. The ad set is 30 % more expensive than its campaign. The parent of an ad set is the campaign, the parent of an ad is the ad set. Campaigns have no parent.

The parent is compared by its weighted average value, that is, the metric calculated over the whole parent: campaign CPI = all its spend / all its installs. It is not the arithmetic mean across ad sets. Therefore only ratio metrics can be compared with the parent: costs (CPC, CPM, CPI, cost per registration, CPD) and rates (CTR, CR, ROI, ROAS). Counts, spend, revenue and profit are not compared with the parent.

Same metric in another window ± %. Expected value = metric in the baseline window × (1 + change ÷ 100). Fields: “Baseline window” and “Change” in percent: greater than −100, up to 1000. “−40 %” means “40 % below the baseline”, “+50 %” means “50 % above”.

• “CTR · Today · less than · CTR over the last 7 days excluding today −40 %” — the creative is burning out.

• “CPM · Today · greater than · CPM yesterday +50 %” — a CPM spike.

The baseline window must differ from the condition window, otherwise the builder shows “The comparison needs a different window”. “Since budget change” cannot be the baseline window.

Important. If the baseline is negative (ROI, profit), the percentage is calculated from its absolute value, and “−” always moves the threshold down, to the worse side. Example: ROI for the past days is −20 %, the condition “ROI today less than baseline −30 %” fires at ROI below −26 %. The same applies to the multiplier: “less than 0.5 × parent ROI” with a parent ROI of −20 % means “below −30 %”.

How to enter values

A threshold outside the allowed range (for example, “ROI less than 2000” when you meant 20 %) will not be saved: the builder highlights “Threshold is outside the metric range”.

Deposit delay: why “today” is dangerous

Deposits arrive with a delay: hours, and sometimes days, after the click. Therefore any revenue metrics “for today” — deposits, registrations, redeposits, revenue, profit, ROI, ROAS, CPD — look worse during the day than they will in the evening.

Example. The ad set is consistently profitable: it spends about $120 a day and brings about $180 of revenue (ROI +50 %), and deposits arrive on average 3–8 hours after the click. Rule: “ROI · Today · By event time · less than 0 → pause”.

• At 09:00, spend since midnight is $35. Deposits that arrived since midnight total $20, mostly the tail of yesterday evening's clicks. ROI = (20 − 35) ÷ 35 ≈ −43 %. The rule pauses a working ad set.

• The cohort is no better in the morning: by 09:00 today's clicks have brought deposits of, say, $10, ROI ≈ −71 %. The cohort matches money more honestly, but for “today” it has not matured yet.

• By event time, “today” also mixes days: today's spend is compared with deposits from yesterday's clicks. In the evening, such an ROI may, on the contrary, show a “profit” for an ad set that brought nothing today.

The builder warns about this right next to the condition: “Deposits arrive with a delay: today's revenue by event time is understated. Use the click cohort, the “Yesterday” window or exclude the last hours”. Cohort windows up to the current hour have a hint: “The click cohort is still maturing…”.

Minimum volume

Without a volume threshold, the rule fires on noise: an ad set spent $3, has no deposits, ROI is −100 % — and it is already “unprofitable”. “Minimum volume” filters out such objects before the conditions are checked.

• The threshold metric is “Spend”, “Clicks”, “Impressions” or “Installs”. The value is compared as “at least” (≥). When enabled, the default is “Spend ≥ 20 USD”.

• The volume is calculated in the rule's default window with its excluded hours, not in the windows of individual conditions.

• An object below the threshold is not evaluated and not touched. In the preview and log it shows “Below the minimum volume: … required …”.

• If there is no data for the window yet (for example, the “Today” window in the first hour of the day), this is also “Below the minimum volume”.

If there is no threshold, the builder warns: “No minimum volume: the rule may act on noise. Set a minimum volume or add a spend, clicks or impressions threshold to the conditions”. The warning does not appear if the mandatory conditions include “Spend”, “Clicks” or “Impressions” with the “greater than” or “at least” operator and a threshold above zero. Mandatory conditions are those in “AND” groups from the very top of the tree. A condition inside an “OR” and an installs threshold in the conditions do not remove the warning.

Tip. A threshold in a condition and the minimum volume behave differently. An object below the minimum volume gets “Below the minimum volume”, and it is immediately clear that it simply lacks data. An object that did not pass the threshold in a condition gets “Conditions not met”. For analyzing the log, the minimum volume is more convenient.

Default window and attribution

Important. If “By event time” attribution is selected at the rule level, ROI, ROAS, CPD and profit are also calculated by event time, although cohort is usually used for them. The builder shows the hint “The rule-wide attribution (…) overrides this metric's default (…)” for such conditions. If a metric needs its own attribution, set it directly in the condition.

Limits

How “window × attribution” pairs are counted

Each unique combination of window, excluded hours and attribution is a separate data query, and one rule can read no more than 6 such combinations. Above the tree there is a counter “Window × attribution pairs used: N of 6”. All windows of the rule are counted: condition windows, windows on the right-hand side of comparisons and the minimum volume window. Metrics without attribution (spend, impressions, clicks) in a window where the tracker is already read do not add a separate pair.

• “Spend · Today greater than 60” + “Deposits · Today · cohort equals 0” — 1 pair.

• “ROI · Yesterday · cohort” + “ROI · 3 days excluding today · cohort” + “Spend · 3 days excluding today”, minimum volume in the “3 days excluding today” window — 2 pairs.

• The same “Yesterday” window, but one condition by cohort and another by event time — 2 pairs.

• The same “Today” window, but one condition excludes 2 hours and the other does not — 2 pairs.

Two identical conditions in one group cannot be saved: “Duplicate condition”.

Ready-made formulations

Ready-made rules with these conditions are collected in templates. It is convenient to take them as a basis and change the thresholds.

How to check conditions before launch

• The preview (the last step of the builder) shows for each object the metric, value, condition and result — “yes”, “no” or “unknown” with a reason — as well as what the rule would do right now. Nothing is applied. More details are in “Preview, test and live modes”.

• In the log, each action has a “Why?” button — a snapshot of the conditions at the moment of the decision: value, window with dates and time zone, attribution, threshold and result. The “Why not now?” tab checks any object by ID. More details are in “Log: runs, ‘Why?’ and rollback”.

Important. Tracker metrics see deposits per Facebook object only from sources where the campaign, ad set and ad ID macros are set up. If no source of the ad account passes them, live in this ad account is blocked; otherwise the rule would decide that everyone has zero deposits. How to set it up is described in “Setting up traffic sources”.

What's next

• Metrics and attributes reference

• Actions: pause, enable, budget, notification

• Schedule, guards and priorities

• Why the rule did not fire: reasons, limitations and FAQ

• All articles in the “Autorules” section