Interpolation estimates a value inside the range of your known data. Extrapolation estimates a value outside it. Both can use exactly the same formula. The difference is where your x sits, and that changes how far you can trust the answer.
Example: you measure a plant at 10 cm on day 2 and 16 cm on day 6. Estimating its height on day 4 is interpolation, and the answer is 13 cm. Estimating day 30 is extrapolation: the straight line says 52 cm, which assumes it keeps growing at the same rate for another three and a half weeks.
Interpolation vs extrapolation at a glance
| Interpolation | Extrapolation | |
|---|---|---|
| Where x is | Between known points | Before the first or after the last known point |
| What you assume | The pattern holds between the points | The pattern carries on past the data |
| Typical error | Small, held in check by nearby points | Grows the further out you go |
| Common uses | Reading tables, filling gaps, resizing images | Forecasts, trends, projections |
| Linear formula | y = y1 + (x − x1)(y2 − y1) / (x2 − x1) | The same formula |
| Linear answer stays between y1 and y2 | Yes | No |
Same formula, two uses
Take two known points: sales of 40 (thousand) in 2019 and 52 in 2021. The slope is (52 − 40) / (2021 − 2019) = 6 per year.
- Interpolate 2020: 40 + 1 × 6 = 46
- Extrapolate 2024: 40 + 5 × 6 = 70
The arithmetic is identical. The 2020 estimate is pinned by real data on both sides. The 2024 estimate depends entirely on the trend carrying on for three more years, through whatever happens to prices, competition and demand.
Why extrapolation goes wrong faster
Suppose the real relationship is curved, y = x², but you only know two points, (1, 1) and (2, 4). The straight line through them is y = 3x − 2. Here is how far it drifts:
| x | Straight line | True value | Error |
|---|---|---|---|
| 1.5 (interpolation) | 2.5 | 2.25 | 0.25 |
| 3 | 7 | 9 | 2 |
| 5 | 13 | 25 | 12 |
| 10 | 28 | 100 | 72 |
Inside the data the error is small, because the curve cannot wander far between two nearby points. Outside, nothing ties the line to the real curve, and the error keeps growing with distance. That is the core reason extrapolation needs more caution.
When to use interpolation
- Reading values between rows of a table, such as steam tables, statistical tables and interest rate tables.
- Filling a missing reading in a time series that has data on both sides.
- Estimating rates between known maturities, as in yield curve interpolation.
- Resizing images and resampling signals, where new pixels or samples are estimated from their neighbours.
When extrapolation is reasonable
- Short-range forecasts, when you have a solid reason to expect the trend to hold.
- Projecting a measurement a little past the last test point, for example the next step in a lab series.
- Finding where a trend line crosses an axis, such as a starting value or a break-even point.
To extrapolate more safely:
- Stay close to the data. A step 10% beyond your range is far safer than one 200% beyond it.
- Check the shape. If the data curves, a straight line will drift, so fit a curve that matches the pattern.
- Look for limits. Growth slows, tanks fill and prices hit floors. Ask what would break the trend.
- Give a range. When the result matters, quote a low and high estimate rather than one number.
Our linear extrapolation guide has worked examples, including one where a straight line fails badly.
Tools for both
- The interpolation calculator on our homepage handles linear interpolation between two points.
- The extrapolation calculator projects beyond your data with a straight line or a best-fit line.
- The Lagrange interpolation calculator fits a curve through up to 15 points.
New to the method? Start with how to interpolate, or see every formula on the interpolation formula page.
FAQs
Is extrapolation less accurate than interpolation?
Usually, yes. Interpolation is checked by known data on both sides, while extrapolation relies on the pattern continuing where you have no data at all. The further past your data you go, the larger the likely error.
Can you use the same formula for interpolation and extrapolation?
Yes. The linear formula, and polynomial or spline formulas, work for any x. What changes is not the maths but how much you should trust the result.
What is an example of interpolation in real life?
Reading a value between two rows of a steam table, estimating a missing temperature reading from the readings before and after it, or a photo editor creating new pixels when it enlarges an image.
What is an example of extrapolation in real life?
Projecting next year’s sales from the last few years, estimating a child’s adult height from a growth chart, or a weather model estimating conditions beyond the last measured point.
Is regression interpolation or extrapolation?
Neither on its own. Regression fits a trend line to data. Using that line inside the range of your data is interpolation, and using it outside the range is extrapolation.