Define the questioned signature
Identify the exact mark under examination and the document or image from which it came. Keep the full source alongside any crop. Record whether the questioned item is a scanned handwritten signature, an image pasted into a document or a genuine digitizer capture; these are different evidence conditions.
Build an independent reference set
Collect known signatures with a documented source and a clear connection to the claimed writer. Keep each reference as a separate specimen. Where available, include material from different dates and comparable signing conditions. Repeated copies of one image add no independent writing variation to the set.
Check assessability before comparing
Inspect resolution, focus, compression, clipping and the amount of writing captured. A small, heavily compressed crop can conceal stroke detail. Keep the original scale and retain context when preparing a comparison view. A photograph does not contain the recorded timing or pressure channels of a digitizer capture.
Compare the pattern of variation
Compare entry strokes, letter construction, connections, proportions and terminal strokes across the reference set. Look for features that recur and features that vary within the known material. Use the number of references required by the selected workflow; some comparisons need two or more independently supplied exemplars.
Keep different outputs distinguishable
A descriptive similarity score, a raw comparison distance and a calibrated likelihood ratio answer different questions. Record which reference produced each result and keep the workflow’s own interpretation scale with the output. Do not turn a raw distance into a percentage of authenticity or average unlike output types together.
Create a source-linked comparison record
Keep the questioned specimen, individual references, selected detail views and the resulting comparison table together. Record the source, relevant processing and any excluded samples. A reproducible record should show which material was used and how natural reference variation affected the comparison.