What could happen next?
Review campaign-level five-day estimates for CPA, spend, conversions, and ROAS, with ranges that make uncertainty visible.
Campaign performance, made clearer
Spendwise brings campaign history, near-term forecasts, and accuracy checks together—so you can see what may be changing before you decide what to do.
Forecasts are estimates based on your history—not promises about future results.
Campaign overview
Main project · sample preview
Current CPA
$24.26
Forecast CPA
$23.10
Likely range
$20–27
Blended CPA over time
Illustrative valuesBuilt around the questions that matter
Spendwise pairs a statistical forecast with evidence from your own account history—then keeps the source data, project, and access permissions together.
Review campaign-level five-day estimates for CPA, spend, conversions, and ROAS, with ranges that make uncertainty visible.
Backtest forecasts against held-out history. See error measures and sample coverage instead of relying on a generic accuracy claim.
Optionally generate an AI-assisted report with observed drivers and suggested checks. AI explains the data; it does not make a forecast certain.
Inside the forecast engine
Spendwise combines recent performance, campaign rhythm, and uncertainty into a forecast you can inspect. The calculation is statistical and explainable; AI is optional for turning the results into a written report.
The engine prioritizes current campaign behavior while reducing the influence of unusual spikes and one-off days.
Weekday patterns help the estimate reflect how your campaigns normally behave across the week.
Spend and conversion rate are projected independently before CPA is calculated, which makes sparse data less jumpy.
Rolling holdout backtests show how close earlier forecasts were to later outcomes in your own project.
Built to be measured, not blindly trusted.
Bring your real campaign history and see the error, coverage, and range before making a decision.
Read the output, not just the headline
Spendwise keeps the estimate, its uncertainty, and its historical test results together so your team can decide how much weight to give it.
The label compares forecast CPA with the recent seven-day CPA. A change within roughly 2.5% is treated as stable; lower CPA is favorable, higher CPA is worsening.
Improve, stable, and decline percentages are derived from the forecast range with a normal approximation. They are not guarantees or proven historical odds. AI can explain them, but does not calculate the forecast.
Rolling holdout tests compare old predictions with later actuals. Check sample count and error measures for your own history; a backtest does not guarantee future performance.
Simple by design
Keep your setup in your control. Connect only the projects and data you want to analyze.
Sign in with Google. Create a private workspace or an organization and invite teammates with the right role.
Connect Google Ads or Meta Ads from project settings, or import a supported campaign CSV.
Generate estimates, inspect their ranges, and compare past forecasts with what actually happened.
A practical starting point
Forecast quality depends on consistent, relevant history. The app shows when there is not enough data, rather than treating an empty or sparse project as a reliable forecast.
A campaign needs at least 21 active days with spend and conversions. Forecasts currently cover the next five days and use recent trend plus weekday patterns.
The current backtest needs at least 120 history days per campaign and eligible holdout windows. More samples make the measured results more informative, but not necessarily predictive of a future change.
Automatic column matching accepts campaign name, date, and spend or conversions as required fields. Revenue is optional and supports ROAS; profit is not required or currently forecast. CSV files can be up to 10 MB.
Ad account connections are configured in project settings. CSV uploads are available from the dashboard and saved to the project you selected.
CSV essentials
Campaign name, date, and spend or conversions. Profit is not required; revenue or conversion value is needed for revenue-based ROAS.
Provide the platform credentials for the ad account, sync its history on demand, and keep the connection scoped to a project.
Organization membership and project roles determine who can view reports, import history, or manage integrations.
Choose a supported AI provider in settings for written analysis. Forecasts and backtests do not require an AI key.
Good to know
A little more detail on data, forecast accuracy, AI, and working with a team.
Connect Google Ads or Meta Ads, or import a CSV with a campaign name, date, and spend or conversions. More consistent campaign history gives the forecasting model more information to work with.
No. Spend and conversions are enough for CPA. Revenue or conversion value is needed for revenue-based ROAS; profit itself is not currently a modeled metric.
No. Spendwise's forecasting model calculates the numeric estimates. AI is optional and can turn the available campaign and forecast data into a written report. An AI report does not guarantee higher forecast accuracy.
Review the rolling backtest, number of evaluated samples, error measures, and forecast ranges in your project. Results depend on the quality and amount of your history, and can change as campaign conditions change.
A campaign needs at least 21 active history days to generate a forecast. Backtesting needs longer histories—currently at least 120 days per campaign—and enough eligible forecast windows. More history does not guarantee a more accurate result, especially after major campaign changes.
Yes. The importer automatically matches common column names. Campaign name, date, and either spend or conversions are required. Campaign ID, source, status, revenue, impressions, and clicks are optional. CSV files are limited to 10 MB, and a downloadable template is available in the importer.
No. They are model-derived estimates based on the forecast range and a normal-distribution approximation, not a promise or a calibrated historical win rate. Compare them with the measured backtest and use the range to understand uncertainty.
Yes. Create an organization, add projects, and invite teammates with roles that control what they can view or manage. Campaign history and reports stay scoped to their project.