Statistics are wonderful, but sometimes they mislead us. We know that the human brain doesn’t use statistics to make decisions. Sometimes, we think we understand them and we try to use them. The dangerous thing, though, is when something seems like good statistics but brings us to the wrong conclusion. The best example that I know of to explain this problem is Rosaria Syndrome.
Rosaria is an extremely thorough, organised and motivated accounting expert. Rosaria worked in accounting at Translated for many years. Her job was to pay a thousand translators every month. She took the amounts and bank details from their invoices and issued the wire transfers.
Translated has always been very careful to pay its translators quickly and on time, because this has always been an effective way to attract and retain the best linguists.
Rosaria makes very few mistakes – she’s probably one of the people I know who makes the fewest mistakes – but naturally she occasionally got the amount and the bank details wrong. To issue a thousand payments, Rosaria needed about 10 days. All this meant that sometimes a translator would not receive the payment or would not be able to predict when during the 10 days the payment would be issued.
To solve these problems, Translated decided many years ago to automate its invoice and payment systems. In just 10 minutes a few days after the end of the month, all payments were issued to you, with very high levels of accuracy. The quality of Rosaria’s work improved, and she had more time to deal with translators.
About a year later, Rosaria asked to meet with me. Without mincing her words, she said to me: “Marco, if we don’t change something, we’ll fail. If we keep making mistakes in our payments, translators won’t want to work with us anymore. We’re going to become the worst translation company to work with.” And then she began listing the noticeable problems that translators had had in receiving payments.
Rosaria was well aligned with Translated’s values, and she took the matter seriously.
I think Rosaria noticed my astonished face as I listened to her words, and at a certain point she stopped listing the problems. What Rosaria didn’t know was that, a few days beforehand, Translated had been chosen by the translators of ProZ (the world’s largest translation community) as the best company in the world to work with, with the main reason being its fast, prompt payments. In fact, her work was award-winning.
Rosaria was convinced that we were the worst payers in the world, but translators thought we were the best.
Why was she so convinced that everything was going wrong? Because her brain had done what we all have a tendency to do: we’re fooled by what seems like good statistics but isn’t.
Rosaria’s work had changed in two ways after we automated the system:
her time went from 8 days for payments and 2 days for troubleshooting to 0 days for payments and 1 day for troubleshooting. The number of problems had halved, but now 100% of her time was spent solving them. She no longer received thank you emails from translators, replying to her message to tell them that their payment had been issued. Now the system sent the notifications, and obviously, no one sends a thank you in reply to a machine-generated message.
Since she no longer received positive feedback and saw only the problems, her brain was conditioned to overestimate the negative cases, to underestimate the positive ones, and to no longer have any indication of the total number of cases. In fact, the number of payments had doubled along with the number of translators.
Rosaria Syndrome is innate in humans. In medical terms it has an incidence of 100%, which means that each one of us has it. In fact, to tell the truth, I have had similar discussions with most of the people I know.
It is highly likely that in the future we will increasingly spend our time dealing with the exceptions, bugs and problems that machines are not able to handle. In this future, Rosaria Syndrome will become more significant.
To combat Rosaria Syndrome, do not give in to the temptation of statistics that seem like the easy option, trust your intuition and logic, talk to people more, and always ask yourself if you are observing the real phenomenon or just a subset – this last part is the most difficult bit.