Viewability Threshold Map 66 weeks | Dec 2024 – Mar 2026
Last 4 weeks average (US properties) — Threshold lines at 50%, 70%, 80%
CPM Index by Viewability Band
<50%
0.40× baseline
0.40× baseline
50–60%
0.55× baseline
0.55× baseline
60–70%
0.82× baseline
0.82× baseline
70–80%
1.00× (baseline)
1.00× (baseline)
80–90%
1.10× baseline
1.10× baseline
90%+
1.22× baseline
1.22× baseline
2-week lag: After a viewability threshold crossing, eCPM response takes ~2 weeks (DSP rolling window recalculates bid multipliers).
Property Time Series 66 weeks | Dec 2024 – Mar 2026
Weekly viewability %, eCPM, and supply volume over time. Reliable viewability data from Feb 3, 2025.
Viewability % (weekly) — Threshold lines at 50%, 70%, 80%
eCPM ($/1000 impressions) — Note: ~2 week lag after viewability change
Supply Volume (millions of impressions)
Data Coverage Note: Active View measurement coverage for some units (especially Yahoo Mail Desktop) was incomplete before Feb 2025. Viewability data is most reliable from Feb 3, 2025 onwards. Zero or near-zero viewability readings before that date reflect measurement gaps, not actual viewability.
Unit Removal Model 66 weeks | Dec 2024 – Mar 2026
Select units for removal and see live revenue & viewability impact. Default sort: viewability ascending (worst first).
| Country ⇅ | Sub-brand ⇅ | Section ⇅ | Slot ⇅ | Position ⇅ | Platform ⇅ | Supply Share ⇅ | Viewability ⇅ | eCPM ⇅ | Weekly Rev $K ⇅ | Band |
|---|
VIEWABILITY OVER TIME
Impact Through Dec 2026 (~39 wks remaining)
Select units to model removal impact
Break-Even Formula
Required CPM lift = 1/(1 − supply_removed%) − 1
e.g. remove 20% supply → need 25% CPM lift to break even
e.g. remove 20% supply → need 25% CPM lift to break even
Net revenue = −(removed supply × removed eCPM) + (remaining supply × (new eCPM − old eCPM))
Viewability Attribution 66 weeks | Dec 2024 – Mar 2026
Viewability and yield move together — but seasonality and supply changes confound the relationship. Use the event impact tool to isolate specific changes; the correlation shows the directional signal across the full period.
Viewability % vs. eCPM over time
Left axis: viewability %. Right axis: eCPM or Page RPM. DSP bid adjustments typically lag viewability changes by ~2 weeks — look for the right axis to follow the left axis with a delay. Vertical lines = key supply events.
Rate vs. Viewability Correlation view[T] → rate[T+2]
Event Impact + Counterfactual
Dashed grey = pre-event average held flat ("what if nothing changed?"). Comparison property re-scaled to same pre-event level — its divergence from the flat line shows how much of the movement was market-wide vs. property-specific.
Impact Summary
Select an event above
Rate Level & Stability by Viewability Band
Stability Index by Band
CV = coefficient of variation (std dev ÷ mean). Lower CV = more predictable rates week to week. High-viewability inventory commands both higher and more consistent eCPMs — DSPs discount uncertainty.
Full Funnel + Projections Jan 2019 – Feb 2026 + projections
Monthly revenue actuals (Jan 19 – Feb 26) with Budget, OL, and Seasonal projections through Dec 26
Monthly Revenue ($M)
Seasonal estimate: Uses FY25 monthly distribution applied to Jan/Feb 2026 run rate. Formula: seasonal_est[m] = FY25[m] × (Jan26+Feb26) / (FY25_Jan+FY25_Feb). Does not include structural changes.
Supply (Impressions Billions)
Viewability Impact Analysis 66 weeks | Dec 2024 – Mar 2026
Before/after analysis of threshold crossings and supply events — does viewability drive eCPM? Scatter excludes measurement-gap zeros (pre-Feb 2025).
Key Period Summary: Mail Desktop & Homepage Desktop
| Period | Property | Date Range | Wks | Avg Viewability | Avg eCPM | Supply (B) | Revenue ($M) | Notes |
|---|
Viewability vs eCPM Scatter (US Desktop, weekly)
Viewability & eCPM Over Time
2-week lag: After viewability crosses a threshold, DSPs take ~2 weeks to recalculate bid multipliers. Look for eCPM response lagging viewability changes.
Measurement Coverage Note: Active View measurement for Yahoo Mail Desktop was not fully operational before ~Feb 3, 2025. Weeks before that date show 0% viewability because measurement was not set up — this is historically accurate. The scatter plot and period analysis exclude zero-viewability weeks (min 1M measurable impressions filter applied). Viewability data is most reliable from Feb 3, 2025 onwards.
Property Focus 66 weeks | Dec 2024 – Mar 2026
One-click property isolation — same metrics as Property Time Series. Events filtered to selected property.
Viewability % — Threshold lines at 50%, 70%, 80%
eCPM ($/1000 impressions)
Supply Volume (millions of impressions)
Uplift Planner Last 4 weeks baseline
Model theoretical revenue uplift from viewability improvements. Combine levers to find the best path to your target — unit removal uses actual data; other levers use configurable assumptions.
Lever 1 — Remove Low-Viewability Units
Uses actual unit-level data from Tab 3 — removes units below the threshold and recomputes viewability from the remaining inventory.
50%
Lever 2 — Lazy Loading
Only render ad units when they scroll into the viewport. Boosts viewability but some below-fold slots are never served.
+8pp
-5%
Lever 3 — Ad Refresh Optimization
Tighten rotation cadence so each impression spends more time in view. Reduces volume but improves per-impression quality.
+5pp
-10%
Lever 4 — Viewport Targeting
Suppress ad calls when the slot is below the fold. Directly eliminates non-viewable impressions — highest viewability gain but highest supply cost.
+12pp
-15%
BEFORE / AFTER
Projected Impact — Apr–Dec 2026 (39 wks)
Enable levers to see projected impact
Path to Target
Select levers above to see the path to your viewability target.
How this works: Unit removal uses last-4-week actual data from the unit table — same logic as Tab 3.
Lazy loading, refresh, and viewport targeting use configurable assumptions (defaults = industry benchmarks; adjust for your property).
eCPM uplift uses linear interpolation across Yahoo-observed CPM Index anchors (50%→0.40×, 60%→0.55×, 70%→0.82×, 80%→1.00×, 90%→1.10×, 100%→1.22×) — movement within a band shows proportional uplift, not a cliff.
2-week DSP lag not modeled — bid adjustments take time to propagate after a viewability threshold crossing.
Glossary & Methodology
Rules, definitions, and known quirks you need to know when reading this dashboard.
Property Hierarchy
News Group ≠ "News" alone. Your revenue reporting uses News Group = Yahoo News (web) + Yahoo Homepage. This dashboard tracks them as two separate properties. When comparing viewability to News Group revenue, you need to consider both.
Yahoo Tech and Yahoo Autos roll into News Group revenue. In FP&A reporting, Tech and Autos are sub-brands within News Group. They appear as separate properties in this dashboard (Tabs 3, 4, 8) but their revenue is consolidated into News Group in all financial reporting. Engadget is separate and not included in News Group.
Engadget, Tech, and Autos have no viewability measurement. Their data exports contain Supply and Billable Revenue only — no Measurable or Viewable Impressions. Viewability % will show as blank for these properties in Tab 4 (Viewability Attribution). Tabs 3 and 8 will show units, eCPM, and supply share, but the viewability columns will be empty.
| Revenue Label | Viewability Properties | Notes |
|---|---|---|
| Yahoo Mail | Direct 1:1 match | |
| News Group | Yahoo News + Yahoo Homepage | Tracked separately here; also includes Yahoo Tech and Yahoo Autos sub-brands in financial reporting |
| Finance | Yahoo Finance | Direct 1:1 match |
| Sports | Yahoo Sports | Direct 1:1 match |
| — | Yahoo Shopping Hub | Hub property; viewability data present. Not broken out separately in standard revenue reporting. |
| — | Yahoo Health Hub | Hub property; viewability data present. Not broken out separately in standard revenue reporting. |
| — | Yahoo Travel Hub | Hub property; sparse viewability data (limited history). |
| Engadget (standalone) | Engadget | No viewability measurement in current data export. Supply + eCPM only. |
| News Group (sub-brand) | Yahoo Tech | Rolls into News Group revenue. No viewability measurement. |
| News Group (sub-brand) | Yahoo Autos | Rolls into News Group revenue. No viewability measurement. |
| — | Yahoo News App | Tracked in viewability data; not broken out in revenue reporting here. |
Viewability Bands & CPM Index
These multipliers are estimates, not DSP contracts. They reflect the approximate eCPM uplift associated with crossing each viewability threshold, derived from observed Yahoo data. Actual DSP bid adjustments vary by buyer.
| Viewability Band | CPM Index | Interpretation |
|---|---|---|
| <50% | 0.40× | 60% discount vs. 70–80% baseline |
| 50–60% | 0.55× | 45% discount |
| 60–70% | 0.82× | 18% discount |
| 70–80% | 1.00× | Baseline (Yahoo standard) |
| 80–90% | 1.10× | 10% premium |
| 90%+ | 1.22× | 22% premium |
Unit Removal Model — How Revenue Impact is Calculated
| Element | Definition |
|---|---|
| Volume basis | Last 4 weeks of data (most recent supply and revenue actuals). Annualizes to a weekly run rate. |
| Revenue lost | Weekly revenue of removed units × 39 weeks remaining (Apr 1 – Dec 28, 2026). |
| CPM lift baseline | eCPM of the remaining units only — not the blended rate including the removed units. Applying the lift to the blended rate would overstate the gain. |
| eCPM lift | Remaining units' actual eCPM × (new CPM index ÷ current CPM index) − current eCPM. |
| Revenue gain | Weekly remaining supply × eCPM lift ÷ 1,000 × 39 weeks. |
| Net impact | Revenue gain − revenue lost. Can be negative if low-viewability units carry disproportionate revenue. |
| 39 weeks | Weeks remaining in FY26 from Apr 1, 2026 through Dec 28, 2026. |
Known Data Caveats
Mail Desktop viewability is zero before ~Feb 3, 2025. Active View measurement was not fully operational before that date. These are measurement gaps — not actual 0% viewability. The scatter plot and period analysis exclude these weeks. All charts treat 0% viewability as null (no data point rendered).
Viewability % shown is display/native only. Video inventory is excluded from all unit-level analysis and the unit removal model.
eCPM is billable revenue ÷ supply × 1,000 — it includes both direct and programmatic. It is not a pure programmatic CPM and will differ from any DSP-reported CPM.
INTL viewability measurement coverage is lower. Many INTL units have sparse or missing measurement data. Use INTL figures directionally only.
Aggregation Rules
| Aggregation | How It Works |
|---|---|
| All Devices | Supply-weighted average viewability across Desktop Web + Mobile Web + Mobile App. Supply and revenue are summed. eCPM is recalculated from combined totals. |
| All Geo | US + INTL aggregated with the same supply-weighting approach. Does not adjust for coverage differences between geos. |
| Supply share | Each unit's share of total supply for that property + device + country combination, based on the last 4 weeks. |
| Viewability (time series) | Weekly viewable impressions ÷ measurable impressions. Weeks with zero measurable impressions are shown as null. |
Composition vs. Rate (Tab 4)
| Line | What It Answers |
|---|---|
| Solid — Actual eCPM | What Yahoo actually earned per 1,000 impressions each week. |
| Dashed — Baseline Mix | "What if supply shares had stayed frozen at week 1, but unit eCPMs changed as they did?" If this rises, rates improved. Gap between this and Actual = mix shift impact. |
| Dotted — Actual Mix | "What if supply shares shifted as they did, but unit eCPMs stayed at week 1 rates?" If this rises above week 1, the mix of units improved (more supply shifted to higher-value units). |
Baseline week is 2024-12-30 (first week in the dataset). All composition effects are relative to that week's unit mix.
Event Annotations
| Color | Type | Meaning |
|---|---|---|
| ● | Supply Up | Supply-increasing change (e.g. ad rotation extended, density increased) |
| ● | Supply Down | Supply-reducing change (rotation tightened, units removed) |
| ● | Critical | Event that caused a significant negative viewability or revenue impact |
| ● | Recovery | Event that restored or improved viewability |
Event dates snap to the nearest week boundary. Source dates may not fall exactly on a Monday week-start; the annotation is placed at the closest data point within ±14 days.
Data Pipeline Status
Feed health, coverage, and refresh schedule for all 18 parquet datasets.
Feed Inventory
Loading…
Refresh Commands
Fast Feeds — GAM (5 feeds)
Frequency: Run any time — safe to run daily
Feeds: Summary, Slots, Page Types, Formats, Yield Partner
Requires: Valid cert (run bq_auth.ps1 first)
cd "G:\Shared drives\Mail News Platforms AI Work\Paul N\Viewability Project\Scripts"
python refresh_bq_data.py --fetch
python generate_dashboard_data.py
Slow Feeds — AY / Prebid (13 feeds)
Frequency: Weekly (run once per week)
Feeds: Bidders, Login State, Refresh, Geo, Funnel, Deal Type, Browser, CTR, Ad Size, Content Type, Prebid, Adomain, Signals
Requires: Valid cert (run bq_auth.ps1 first)
cd "G:\Shared drives\Mail News Platforms AI Work\Paul N\Viewability Project\Scripts"
python refresh_bq_data.py --fetch-slow
python generate_dashboard_data.py
Auth note: Run
bq_auth.ps1 (enter PIN + touch Yubikey) before running any refresh.
The cert stays valid for ~4 hours — scheduled tasks will auto-renew the 1-hour JWT without any further interaction as long as the cert is still valid.
Demand Analysis — Bidder × Viewability
Bidder revenue and CPM across viewability regimes. Data: Dec 2025 – Mar 2026 (15 weeks). Source: Kristian G demand data joined with GAM viewability.