Forecasting FAQ
Common questions about GSC traffic forecasting
Quick Answers
Everything you need to know to get started
Getting Started
What scenarios should I use this tool for?
PredictClicks is ideal for: getting client buy-in on SEO investments, setting realistic traffic goals, measuring before/after impact of changes, budgeting and planning content calendars, and tracking website migration recovery. If you need to show stakeholders where traffic is heading, this is for you.
How accurate are the forecasts?
Typical accuracy ranges from 10-30% error depending on your traffic patterns. Stable, established sites with consistent traffic see better accuracy (10-15%). Newer sites or those with volatile traffic will have wider ranges. Every forecast includes accuracy metrics so you know exactly how reliable the prediction is.
How long does a forecast take?
Most forecasts complete in under 60 seconds. Complex sites with extensive historical data may take up to 2 minutes. You can close the browser and check results later - they're saved to your history automatically.
What if my site is brand new?
New sites need at least 60 days of Search Console data. Without enough historical patterns, forecasts will be less reliable. We recommend waiting until you have 3-6 months of data for meaningful predictions. The tool will tell you if you don't have enough data.
Understanding Your Results
What is the Confidence Score?
We simplify forecast accuracy into an easy-to-understand Confidence Score from 0-100%. This is calculated from the technical MAPE (Mean Absolute Percentage Error) metric but inverted so higher = better. A score of 85% means the forecast is typically within 15% of actual values. Scores above 80% are excellent, 60-80% are good for planning, below 60% suggests volatile data.
Why do my forecasts show confidence intervals?
The shaded area shows the range where actual traffic will likely fall. We use 95% confidence - meaning there's a 95% chance the real number lands within that range. Wider intervals mean more uncertainty. This is normal and helps you plan for best and worst case scenarios.
The shaded area shows the range where actual values are 90% likely to fall. Wider bands = more uncertainty about the future.
What should I do if forecasts seem way off?
First, check if there were recent major changes (algorithm update, site redesign, technical issues). Verify you have at least 60 days of data. For volatile niches, focus on the confidence interval range rather than the center line. If your site recently changed significantly, wait 2-4 weeks for the new patterns to stabilize.
Using Forecasts for Business
How do I use forecasts to get client/stakeholder buy-in?
Export the forecast chart as PNG for presentations. The confidence intervals show realistic ranges (not promises) which builds trust. Compare the projected trajectory to business goals. Use the data to justify SEO investments - 'Based on current trends, we're projected to hit X clicks by Q2' is more compelling than guesswork.
Can forecasts predict the impact of SEO changes I'm planning?
No. Forecasts project current trends forward assuming no major changes. They don't predict outcomes of new strategies. However, you can use forecasts as a baseline: run a forecast before making changes, implement your strategy, wait 2-4 weeks, then compare actual results to what was projected. The difference shows your impact.
Features & Exports
How far into the future can I forecast?
We offer four horizons: 1 month (highest confidence), 3 months (recommended balance of accuracy and planning), 6 months (medium-term planning), and 12 months (directional guidance only). Longer horizons have wider confidence intervals. We recommend 1-3 month forecasts for actionable planning.
Shorter forecast horizons produce more reliable predictions. Use 1-3 months for actionable planning.
Can I forecast specific pages or queries?
Currently, forecasts are generated at the property (site) level for clicks and impressions. This provides the most reliable predictions because aggregated data has less noise. Page-level and query-level forecasting may be added in future updates.
Can I forecast multiple sites at once?
Yes! Use bulk forecasting to generate forecasts for multiple properties in a single batch. Select your sites, choose settings, and let it run. All results are saved to your history for easy comparison across your portfolio.
What's included in the Excel export?
The Excel export includes: daily forecast values with upper and lower bounds, weekly/monthly/quarterly aggregated totals, historical data for comparison, the Confidence Score and underlying MAPE metric, forecast settings used (horizon, metric type), and property information. Everything you need for client reports or further analysis in your own tools.
Technical Details
For power users who want to understand the methodology
Data Requirements & Best Practices
What's the minimum data needed for reliable forecasts?
You need at least 60 days for 1-month forecasts, 120 days for 3-month, 180 days for 6-month, and 365 days for 12-month forecasts. More historical data generally improves forecast stability. Search Console provides up to 16 months of history.
| Forecast Horizon | Minimum Required | Recommended |
|---|---|---|
1M1 Month | 60 days | 90+ days |
3M3 Months | 120 days | 180+ days |
6M6 Months | 180 days | 270+ days |
12M12 Months | 365 days | 400+ days |
More historical data generally improves forecast accuracy and stability. GSC provides up to 16 months (479 days) of data.
Should I use daily or weekly forecasts?
For planning purposes, weekly forecasts are typically more useful as they smooth out daily noise. Our system aggregates daily data properly to provide more stable predictions. Daily forecasts are better if you need to detect short-term anomalies or have strong day-of-week patterns.
How often should I generate new forecasts?
We recommend updating forecasts monthly, or after major events like algorithm updates, site relaunches, or content campaigns. More frequent updates help the model adapt to recent changes. Avoid updating too frequently (e.g., daily) as short-term noise can affect results.
Why is there a 3-day lag in the data?
Google Search Console data has approximately 3 days of latency before becoming complete. Our system automatically excludes the most recent 3 days to ensure forecasts are trained on reliable data. Including incomplete data would cause the model to see a false drop.
How It Works
How does the forecasting engine work?
We use Facebook Prophet, an industry-leading time series forecasting library, with parameters specifically tuned for search traffic patterns. The model automatically detects trend changes, handles missing data, and calculates confidence intervals. For full technical details, see our Methodology page.
How does the system handle Google algorithm updates?
The model automatically detects significant traffic shifts, including those from algorithm updates. It learns from historical patterns and adjusts the trend accordingly. However, future algorithm updates cannot be predicted. We recommend running new forecasts after major Google updates to incorporate the latest data.
Want the Full Technical Details?
Our Methodology page explains the forecasting engine, validation metrics, and configuration in depth.
Read the Methodology