SEO Forecasting: Estimating Traffic and ROI Before You Invest
- Filed November 10, 2026
- Analytics, reporting and testing, guide 6 of 8
An SEO forecast is a conditional estimate: if these pages reach these positions for queries with this much demand, and if visitors convert at this rate, then the investment returns roughly this much, within this range. Every part of that sentence is an assumption, and a useful forecast makes each one visible so the reader can disagree with it. Google’s own guidance on hiring SEO help warns readers that anyone guaranteeing first place should be avoided (Google on hiring an SEO), and its guidance on third-party tools adds that tool predictions “are their own and like predictions generally, may not happen” (Google on third-party SEO tools). A forecast is a planning tool, not a promise.
Two kinds of forecast
Baseline forecast: what organic traffic will do without new investment, projected from historical data. It accounts for trend and seasonality and is the comparison point for any plan.
Opportunity forecast: the additional traffic and value a specific plan might add, built from demand, expected positions, click-through rates and conversion rates.
The investment case is the difference between them, minus the cost.
Building the baseline
Use the full 16 months of Search Console data the report keeps (export it regularly if you want a longer record) and longer analytics history if available. Time series methods separate trend from seasonality; the forecasting textbook by Hyndman and Athanasopoulos describes exponential smoothing and ARIMA as “the two most widely used approaches to time series forecasting”, with complementary strengths (Forecasting: Principles and Practice). A spreadsheet with a seasonal pattern and a trend line is a simpler starting point; what matters is that the method and the history behind it are stated.
Check the history for events that should not be projected forward: migrations, outages, one-off news spikes, tracking changes.
Building the opportunity forecast
For each target page or query group, list:
| Input | Source | Uncertainty |
|---|---|---|
| Monthly search demand | Keyword tool estimates, Search Console impressions for related queries | Tool volumes are estimates and differ between tools |
| Expected position range | Judgment from competition and current positions | High |
| CTR at that position | Your own Search Console data by position for similar queries | Varies by SERP features, brand, query type |
| Share of clicks lost to SERP features | Observation of the actual results pages | Varies by query |
| Conversion rate | Your analytics for similar organic landing pages | Depends on page and offer |
| Value per conversion | Finance or CRM data | Lower when records exist |
| Time to reach position | Judgment | High |
| Cost | Content, development, links, tools, staff | Known or estimable |
Use your own CTR curve from Search Console where possible. Published industry CTR curves come from other sites and other query mixes.
A transparent scenario calculation
Take a hypothetical software company considering 20 comparison pages targeting queries with a combined estimated demand of 12,000 searches per month.
| Assumption | Conservative | Expected | Optimistic |
|---|---|---|---|
| Average CTR achieved (from own data at likely positions) | 2% | 5% | 9% |
| Monthly clicks (12,000 × CTR) | 240 | 600 | 1,080 |
| Visitor-to-lead rate | 1.5% | 2.5% | 3.5% |
| Monthly leads (clicks × rate) | 3.6 | 15 | 37.8 |
| Lead-to-customer rate | 10% | 15% | 20% |
| Monthly customers | 0.36 | 2.25 | 7.56 |
| Value per customer (first-year revenue) | $4,000 | $4,000 | $4,000 |
| Monthly value at full ramp | $1,440 | $9,000 | $30,240 |
All inputs are invented for the illustration. The arithmetic for the expected case: 12,000 × 5% = 600 clicks; 600 × 2.5% = 15 leads; 15 × 15% = 2.25 customers; 2.25 × $4,000 = $9,000.
Ramp: assume no traffic for the first three months, then a linear ramp from month four to full value in month nine. Months four to nine add up to 3.5 months of full value (1/6 + 2/6 + … + 6/6), and months ten to twelve add 3 more, so the first year earns 6.5 months of full-ramp value.
Cost: for example, $40,000 for research, writing and design. If the work is outsourced, how the agency is chosen decides whether that cost buys the planned work.
Reading the scenarios over 12 months (6.5 × monthly value):
- Conservative: 6.5 × $1,440 = $9,360, well short of the $40,000 cost.
- Expected: 6.5 × $9,000 = $58,500, above the cost within the year.
- Optimistic: 6.5 × $30,240 = $196,560.
These are first-year revenue figures, not profit; margin, sales costs and churn change the picture. The decision depends on whether the company can tolerate the conservative case and on how confident it is in the position assumptions.
Sensitivity
Identify which assumption moves the outcome most. Every input in the example multiplies the result, so each one changes value proportionally: halving the expected CTR halves value, and a 10 percent error in demand changes value by 10 percent. What differs is the width of each plausible range. CTR spans 2 to 9 percent in the scenarios, a factor of 4.5, which makes it the assumption that moves the outcome most. Spend research effort on the assumptions that matter most, such as checking the actual results pages for those queries to see how many clicks are likely to go to AI Overviews, ads or local packs.
Forecasting errors to avoid
- Using tool search volume as exact demand. Tool volumes are estimates, sometimes grouped across variants.
- Using generic CTR curves. Results pages with AI Overviews, shopping units or local packs change click distribution.
- Ignoring ramp time. New pages take time to be crawled, indexed and to earn positions.
- Ignoring cannibalization. New pages can take clicks from existing pages instead of adding new ones.
- Single-number forecasts. A point estimate hides the range of plausible outcomes.
- Forgetting the baseline. Value is the difference from what would have happened anyway.
Reporting a forecast
Present the three scenarios, the assumptions behind each, the sensitivity of the result to the main assumptions, and the date by which actual data will be compared with the forecast. Then compare: a forecast checked against reality improves the next one, and a forecast never checked is just an opinion with numbers. Where the plan is a template change, a controlled SEO split test checks the effect more directly than a before-and-after comparison.