How to Forecast Business KPIs: A Practical Guide to Metric Projections

Why Forecasting Matters for Startups

Every startup founder makes decisions based on where they think their numbers are heading. Will MRR hit $50K by Q3? Is churn trending up or was last month an outlier? Should you hire based on current growth or projected growth?

Most founders answer these questions with gut feeling and a mental trendline. That works until it doesn't. A simple projection based on your actual data is almost always more reliable than intuition, and it takes seconds instead of hours in a spreadsheet.

Five Projection Methods, One Dashboard

TotalKPI includes five projection methods built directly into every chart. No exports, no formulas, no separate tools. Toggle projections on and the forecast extends your data forward.

1. Linear Trend

The simplest and most common projection. Fits a straight line to your historical data using ordinary least squares regression and extends it forward.

Best for: Metrics with steady, consistent growth. Monthly recurring revenue that grows by roughly the same dollar amount each month. Support ticket volume that increases predictably with user count.

Watch out for: Linear projection assumes constant absolute growth. If your metric is actually growing by a percentage (compounding), linear will underestimate the future.

Read more about linear projections

2. Moving Average

Smooths out short-term noise by averaging recent data points and projecting that smoothed trend forward. You control the window size to balance responsiveness against stability.

Best for: Noisy metrics where you want to see the underlying trend. Daily active users with weekend dips. Ad spend that fluctuates but trends in a direction.

Watch out for: Moving average reacts slowly to sudden changes. If your metric just had a real inflection point, the projection will lag behind reality.

Read more about moving average projections

3. Exponential Smoothing (Holt's Method)

Weights recent data points more heavily than older ones, capturing evolving trends better than a simple moving average. Uses Holt's double exponential smoothing to model both level and trend.

Best for: Metrics where recent changes matter more than historical patterns. Churn rate that's been improving due to recent product changes. Conversion rates after a funnel optimization.

Watch out for: Can overreact to recent outliers. A single unusual data point will pull the projection more than it would with linear or moving average methods.

Read more about exponential smoothing

4. Logarithmic

Models growth that starts fast and gradually decelerates. The classic adoption curve shape: rapid early growth that levels off as you approach saturation.

Best for: User adoption metrics, market penetration, organic traffic growth for a maturing site. Any metric where you expect diminishing returns.

Watch out for: If your metric is still in an early exponential growth phase, logarithmic projection will be too conservative. It assumes deceleration has already begun.

Read more about logarithmic projections

5. Growth Rate (CAGR)

Calculates the compound annual growth rate from your historical data and projects it forward. This gives you percentage-based exponential growth.

Best for: Revenue and user growth in early-stage startups where you're growing by a percentage each month. Compound metrics where each period builds on the last.

Watch out for: CAGR projection assumes your historical growth rate continues unchanged. If growth is decelerating (as most metrics do eventually), CAGR will overestimate the future.

Read more about growth rate projections

Choosing the Right Method

There's no single best projection method. The right choice depends on what your data looks like:

The practical approach: try two or three methods on the same chart and see which one best matches the recent trend you'd expect to continue. TotalKPI lets you switch between methods instantly.

Projections + Correlations

Projections become even more powerful when combined with correlation analysis. If you've identified that blog traffic has a 0.85 correlation with trial signups (with a 2-week lag), you can project both metrics forward and see whether your pipeline is on track.

Overlay your projected traffic with projected signups on the same chart. If the projections diverge from the correlation pattern, something has changed and it's worth investigating.

Getting Started

1. Open any data source chart in your dashboard
2. Click the projections icon in the chart toolbar
3. Choose a method and projection period
4. The forecast line extends your data forward

That's it. No configuration, no formulas, no spreadsheet gymnastics. Your simple KPI dashboard now shows you where your metrics are heading.

For detailed documentation on each projection method, visit the projections docs.