At a glance
Real-time alert that fires when Microsoft Advertising ROAS falls sharply against its own recent baseline. This is the first number to ping when the “Microsoft is our cheap hedge against Google” narrative is about to look bad in Monday’s report. It compares current ROAS to the prior 7-day, same-day-of-week window so that a normal weekday-to-weekend swing does not trigger a false alarm. A meaningful drop here means revenue per pound is collapsing faster than spend, and the cause is usually tracking, network drift, or a bidding learning phase rather than a real demand change.
| What it watches | Account-level ROAS (Revenue ÷ Spend) on Microsoft Advertising, recomputed continuously and compared to a rolling baseline. |
| Baseline comparison | Prior 7-day window matched on day of week. Tuesday is compared to last Tuesday, not to the weekend, so the alert is not fooled by the normal weekly rhythm of Bing search demand. |
| What “spend” means | Gross media cost, post Click Quality filter. The figure Microsoft will bill you for. |
| What “revenue” means | UET-tracked primary conversion value in your Microsoft Advertising account, not order revenue from the commerce platform. The two often differ. See Reconciling below. |
| Why same-day-of-week | Bing audiences skew desktop and B2B, so weekday and weekend ROAS differ structurally. A simple “vs yesterday” check would fire every Saturday. |
| Bot / invalid traffic | Spend is post-Click-Quality-filter; revenue is real-purchase only. Some invalid traffic slips through on the Search Partner Network and Microsoft Audience Network and drags ROAS down without converting. |
| Currency | Account currency. Multi-currency advertisers should run separate accounts per currency. |
| Cross-account aggregation | Per-customer-account. For a manager-account-equivalent view, watch the alert per child account. |
| Time window | Real-time, evaluated continuously against the prior 7-day same-day-of-week window. |
| Alert trigger | A ROAS drop beyond roughly a quarter against the matched prior window. The exact sensitivity is tuned per account so that low-spend days do not over-trigger. |
| Roles | owner, marketing, finance |
Calculation
Calculated automatically from your Microsoft Ads (Bing) data. See the At a glance summary above for what the metric tracks and the worked example below for a typical reading.Worked example
A UK home-and-garden retailer using Microsoft Ads as the cheaper incremental layer alongside Google. The alert evaluates on the morning of 14 May 26, a Thursday.| Window | Spend | Microsoft-attributed Revenue | ROAS |
|---|---|---|---|
| This Thursday (so far) | £640 | £1,920 | 3.0x |
| Last Thursday (matched baseline) | £610 | £3,050 | 5.0x |
| Prior 4-Thursday average | £625 | £3,180 | 5.1x |
- Spend is flat, so this is not a scaling problem. When ROAS drops because you pushed budget into less efficient inventory, spend rises alongside. Here spend held, so the revenue side broke. That points at tracking or attribution, not media buying.
- Check the tracking canary first. Open Clicks vs Conversions. If clicks held but conversions cratered, the UET tag broke and the UET Tag Tracking Broken alert should also be lit.
- Then check network drift. Microsoft Audience Network is on by default and can find cheap low-quality inventory overnight. Segment by network; if Audience Network share jumped, that is your ROAS leak and Wasted-Spend Burst (Audience Network drift) will confirm it.
- Rule out a learning phase. If a campaign was restructured in the last week (new conversion goal, audience swap), Smart Bidding re-enters a 7 to 14 day learning phase where ROAS is unstable. That is a wait, not a fix.
- ROAS down + spend flat + clicks flat + conversions down = tracking failure (highest priority).
- ROAS down + spend flat + Audience Network share up = network drift.
- ROAS down + spend up = you scaled past the efficient frontier; this alert is doing its job.
- ROAS down + recent campaign edit = learning phase; hold and recheck in a few days.
Sibling cards merchants should reference together
| Card | Why pair it with ROAS Dropped Below Threshold |
|---|---|
| Microsoft Ads ROAS | The level this alert watches. Open it to see whether the drop is brand vs non-brand and how far below your scaling floor you are. |
| Microsoft Ads ROAS Trend | The daily series. Confirms whether this is a one-day blip or a multi-day slide. |
| Microsoft Ads Clicks vs Conversions | The broken-tracking canary. Clicks holding while conversions fall means the UET tag, not demand, is the cause. |
| UET Tag Tracking Broken | If both alerts fire together, the ROAS drop is almost certainly a measurement failure, not a real efficiency loss. |
| Wasted-Spend Burst (Audience Network drift) | The other prime ROAS leak. Drift to low-quality placements drops ROAS without any tracking issue. |
| Microsoft Ads Total Spend | Tells you whether the drop came with rising spend (scaling) or flat spend (a break to investigate). |
| Google Ads ROAS | The benchmark. If Google ROAS held while Microsoft dropped, the cause is Microsoft-specific. |
Reconciling against Microsoft Advertising
Where to look in Microsoft Advertising: Microsoft Advertising → Reports → Performance → Account Summary, with both Spend and Revenue columns visible, segmented by day. Compare today’s ROAS to the same weekday last week. The drop should match what this alert reports to within a couple of percent. Other Microsoft Advertising views that look related but differ:- Smart Goals ROAS uses Microsoft’s modelled values rather than UET-measured. This alert uses primary UET-measured, so the absolute ROAS may differ even when the drop direction agrees.
- Conversion rate alerts in the Microsoft UI track a different metric; a ROAS drop can happen with stable conversion rate if average order value fell.
| Reason | Direction of divergence |
|---|---|
| Real-time event-ingestion lag. Microsoft conversions can ingest with a delay of a few hours. | Today’s ROAS may read low until late-arriving conversions land |
| Same-day-of-week baseline. The Microsoft UI does not offer this comparison natively, so a manual day-over-day check may look worse or better. | Either |
| Smart Goals modelled value. If you opted into modelled revenue, the Microsoft UI may show higher revenue and a softer drop. | Ours steeper |
| Card | Expected relationship | What causes legitimate divergence |
|---|---|---|
google_ads.gads_roas | If the Microsoft ROAS drop is real demand, Google often softens at the same time | A Microsoft-only drop points to a Microsoft-specific cause (tracking, network, bidding). |
shopify.total_revenue | Bing-tagged commerce revenue divided by Microsoft spend should fall in step with this alert | If commerce revenue held but the alert fired, the UET tag is the likely culprit, not real sales. |