A move from 100 to 110 is a 10% simple return and a log return of ln(1.10) = 0.0953. A move back from 110 to 100 is a minus 9.09% simple return and a log return of minus 0.0953 - exactly the opposite sign, which simple returns never give you.
Two properties make them the working unit for analysis. They add: a sequence of log returns summed over a year equals the year's log return, so time aggregation is arithmetic. And they are better behaved statistically, which is why volatility, correlation and most quantitative work is computed on logs rather than percentages.
Convert back before reporting. A cumulative log return of 0.35 is not a 35% gain but e^0.35 - 1 = 41.9%. For small moves the two are close, which is why the difference is easy to miss until the numbers get large.
Related: arithmetic-return, geometric-return, cumulative-return, volatility