The ratios are fixed; the anchors are not. That is why two people can measure the same chart and produce different retracement levels without either making a calculation error.
Begin with the move, not the tool
Hide the Fibonacci drawing control. Describe the impulse in a sentence first: where control visibly shifted, what price displaced, and where that displacement stopped. If the sentence requires several unrelated swings, the move is probably not a single impulse.
For an upward leg, the lower anchor belongs at the low from which sustained displacement began—not automatically the lowest wick visible on screen. The upper anchor belongs where that leg completed before a correction became structurally recognisable.
Require visible evidence
Use at least two observations: a break of a prior swing, expansion in candle range, rejection at the endpoint, or a corrective sequence after completion. Volume can add context where reliable, but it does not repair an incoherent price leg.
Freeze before reveal
In replay practice, save a screenshot before advancing candles. Write one sentence that would invalidate your anchor choice. Then reveal the next segment without dragging endpoints to improve the result. This preserves the distinction between a sound process and a fortunate outcome.
Compare, do not blend
Sometimes two swing definitions are reasonable. Draw each study separately. Label the timeframe and structural premise. Overlap between them may be informative; blending their anchors into a third, unexplained measurement is not.
Fibonacci retracement remains an interpretive aid, not proof of future movement. Its usefulness depends on transparent assumptions and disciplined review.