Intercoder Reliability and Validity of WebPlotDigitizer in Extracting Graphed Data.

Drevon et al. (2017) asked a simple question. Can a free web tool pull numbers from published graphs as well as two humans can?

They took 36 single-case graphs and had coders click every point with WebPlotDigitizer. Then they checked how closely different people agreed and how accurate the clicks were.

Agreement was high. Coders matched on almost every point they pulled.

The tool also hit the true values. The tiny errors that did occur were too small to change a study’s conclusion.

Manolov (2026) extends this work. Their 2026 tutorial now lists WebPlotDigitizer as the first no-install step in any SCED workflow.

Howard et al. (2019) and Tyrer et al. (2006) look like contradictions at first. They warn that visual analysis and statistics can disagree. The key difference is timing. Daniel et al. test data extraction, not the later step of deciding what the data mean. Once the points are pulled, you still need the rules those papers provide.

Lanovaz et al. (2017) hands you one such rule: keep at least three points in A and five in B before you trust dual-criteria lines drawn on the graph you just digitized.

If you meta-analyze single-case data or check a graph in a grant proposal, you no longer need to eyeball the figure. Open WebPlotDigitizer, trace the line, and paste the numbers into Excel. The study shows the risk of coder error is now lower than the risk of skipping the tool.