Media plan forecast ranges: low, expected and high cases with stated assumptions
Part of Programmatic media planning
The short answer
A media plan forecast range gives a low, an expected and a high case for each planned metric, each built from named sources and stated assumptions about budget, flight, geography, inventory and metric definitions. After launch, delivery is read against the range rather than a single target, and any mid flight revision is dated and kept beside the original forecast.
Key takeaways
- Planning tools say their outputs are estimates, so a single forecast number states more certainty than the source gives.
- Each case changes named assumptions; a range without stated assumptions cannot be checked or improved.
- Compare forecast and delivery only on matching metric definitions, dates and campaign setup.
- Keep the original forecast when reforecasting, and tie each revision to the dated change that caused it.
Why should a forecast never be a single number?
Definition
- Forecast range
- A set of planning cases for one metric, such as impressions, reach or video views, each tied to a written set of assumptions, presented together so the reader sees both the expected outcome and how far it could move.
The tools planners use describe their outputs as estimates. Google's About Reach Planner (opens in a new tab) page says "All values quoted are estimates and don't guarantee future performance" and that actual performance depends on factors such as ad quality, ad relevance and campaign settings. Its forecasts page (opens in a new tab) says forecasts are modeled on trends in the ad market and the historical performance of similar campaigns. A single number on a plan hides that uncertainty and turns an estimate into something that reads like a commitment.
How can a site plan communicate low, expected and high cases?
Build each case by changing a small set of assumptions you can name, and leave everything else fixed. In Display & Video 360, the plan builder (opens in a new tab) lets a planner set location, dates, demographics and a budget per product line, and use Add Another to add lines for different combinations of settings and budgets. Those lines can hold the three cases side by side. Google notes that plan outputs include a graph of budget against matching inventory, and that initial CPMs may look low because they reflect broad targeting.
| Case | Assumptions that change | What the reader should take from it |
|---|---|---|
| Low | Only the core list delivers; a narrower inventory match after the site list, suitability settings and frequency cap are applied; the higher cost assumption | The outcome if the constraints bind; the floor for planning other media around |
| Expected | The approved list and settings as planned, with the tool's central estimate after targeting is added | The figure the plan is built around, not a promise |
| High | Full list availability across the flight and the lower cost assumption, with no change to suitability or list rules | The ceiling under the same rules; not a case for loosening them |
The high case should never come from widening the brief. If more delivery needs adjacent topics or a looser suitability tolerance, present that as a separate, labeled option with its own forecast, so the client can choose it deliberately instead of finding it inside the top of a range.
Which assumptions should a forecast state?
Write the assumptions where the reader sees them, next to the range, not in a separate tab. A reviewer should be able to rebuild each case from the list below.
Assumptions to state beside each forecast range
- The tool, its version or run date, and the data period behind it; Reach Planner forecasts use the most recent data from a period equal in length to the planned dates, up to 92 days.
- The audience measure: Google says on-target reach "does not represent a forecast of total unique users (total reach)" and "can be higher than the census population" because it counts visitors.
- The population source and its vintage: the US Census Bureau says the latest vintage "supersedes all previously produced estimates for those dates".
- The metric definition, such as TrueView views or impressions, written in the platform's words.
- Budget, flight dates, geography, formats and currency.
- The site list version, its size and the suitability settings and frequency caps applied.
- What changes between the low, expected and high cases, and nothing else.
For Canadian plans, take population inputs from Statistics Canada's population estimates by age and gender (opens in a new tab), table 17-10-0005, and record its release date. A population figure is a census baseline, not an advertising audience: keep it apart from the platform's modeled reach. In the US, the Census Bureau's population estimates FAQ (opens in a new tab) defines a vintage as a specific version of estimates, so name the vintage you used.
How do I compare forecast views with delivered views?
Delivered figures only test a forecast when they measure the same thing over the same setup. Google's Reach Planner page warns that if the campaign setup differs from the plan, "the metrics in your reporting may be different than in your forecast". View counts need extra care. Google's YouTube view metrics page (opens in a new tab) defines public view count as the sum of organic views and ad public views, and YouTube's engagement metrics page (opens in a new tab) says that, beginning August 24, 2026, views are counted the moment a video starts to play, while "TrueView views are billable, and YouTube public views are not billable".
Five steps to read delivery against a forecast range
- Match the metric. Compare forecast TrueView views with reported TrueView views, never with the public view count on the video.
- Match the setup. List each difference between the plan and the live campaign: budget, dates, list version, formats, caps, suitability settings.
- Match the window. Use the same dates for both, and wait until the platform's data window has closed before calling the final reading.
- Place the result. Record whether delivery fell below the low case, inside the range or above the high case, line by line.
- Explain the miss. For each result outside the range, name the assumption that failed, and carry the corrected assumption into the next forecast.
| Result | What it says about the forecast | Next step |
|---|---|---|
| Inside a narrow range | The assumptions held and the range was informative | Keep the method; record the result |
| Inside a very wide range | Correct but of little use for planning | Find which assumption drove the width and narrow it with better inputs |
| Below the low case | A constraint bound harder than assumed | Check inventory match, list availability and settings changes |
| Above the high case | An assumption was too cautious, or the setup widened | Check whether targeting expanded beyond the plan before calling it a gain |
The table borrows its logic from forecast evaluation research. Bracher, Ray, Gneiting and Reich, writing in PLOS Computational Biology (opens in a new tab) in 2021, describe an interval score made of three quantities: the width of the interval, a penalty for observations below its lower end and a penalty for observations above its upper end. The paper concerns epidemic forecasts, not advertising, but the principle carries: a good range is both narrow and right.
How should a forecast be updated mid flight?
Reforecast when an assumption changes, not on a whim. Keep version 1 of the forecast untouched, issue version 2 with its date and the change that caused it, and compare final delivery with both. Google says Reach Planner is updated weekly (opens in a new tab) and that recommended plans "don't guarantee actual performance", so a rerun weeks later can move even with no change to the plan; record the run date on each version.
| Trigger | Why it moves the forecast | Evidence to attach |
|---|---|---|
| Budget or flight change | Both are direct inputs to the plan builder | The approved change and its date |
| New site list version | Matching inventory changes with the list | Version numbers before and after, with the domain counts |
| Stricter suitability settings | Google says changes that result in more restrictive serving may lead to underdelivery | The setting change from event history (opens in a new tab) |
| Format mix change | Format mix is an input to Reach Planner forecasts | The new mix and the reason |
How does PlacementGraph present projections for a plan?
What does a forecast range look like on one plan?
Fictional example
A Utah summer music festival's display and video forecast
A fictional three day music festival in Salt Lake City plans a spring ticket campaign in Utah and southern Idaho, with display on music, local events and dining sites and YouTube in-stream on music channels. The planner builds three lines per product in the plan builder: low with the core list only and a higher cost assumption, expected with the approved list, high with full list availability. Beside the range, the plan states the run date, the data period, the population vintage, list version 1 and the metric names.
In week three the client adds budget for a second ticket release. The planner issues forecast version 2, dated, citing the budget change, and keeps version 1. At the end of the flight, display impressions land inside the version 2 range. YouTube TrueView views land below the low case; the setup comparison shows that a stricter suitability setting was applied in week four, so the next forecast models that setting in the low case from the start.
Frequently asked questions
Should the low case be a worst case?
No. A worst case, such as a campaign paused for two weeks, is a risk to note separately. The low case is a plausible outcome under the plan's own rules when constraints bind. Mixing the two makes the range so wide that it stops guiding decisions.
Can a forecast range be used as a delivery commitment?
Not without a contract that says so. Planning tools describe their outputs as estimates. A guaranteed deal or reservation with a publisher sets committed volumes through its own terms; a forecast range in a plan describes what the buyer expects, with assumptions, and the plan should say which of the two it is.
How wide should the range be?
As wide as the assumptions honestly justify, and no wider. Width comes from the inputs that are uncertain, so narrow it by improving those inputs, such as confirming list availability with sellers, rather than by quietly dropping the low case.
Should a forecast use the client's past campaigns?
Yes, as one named source, if the past campaign used comparable lists, formats, settings and seasons. State which campaign, its dates and how it differs. A past result from a broader list or a different suitability setting is a weak input for a tighter plan.
Sources
- About Reach Planner (opens in a new tab)Google Ads Help
- About forecasts in Reach Planner (opens in a new tab)Google Ads Help
- About suggestions in Reach Planner (opens in a new tab)Google Ads Help
- Create a plan for a campaign (opens in a new tab)Display & Video 360 Help
- About YouTube ads and view metrics (opens in a new tab)Google Ads Help
- How engagement metrics are counted (opens in a new tab)YouTube Help
- Evaluating epidemic forecasts in an interval format (opens in a new tab)Bracher, Ray, Gneiting and Reich, PLOS Computational Biology, February 2021
- Frequently Asked Questions About Population and Housing Unit Estimates (opens in a new tab)US Census Bureau
- Population estimates on July 1, by age and gender (17-10-0005) (opens in a new tab)Statistics Canada
- Brand suitability (opens in a new tab)Display & Video 360 Help
- Event history (opens in a new tab)Display & Video 360 Help
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.