Machine Learning

Tests displayed: Is $ 1 changed with more than $ 100 behavior?

I am currently leading a small data team to a small technical company. For every little thing, we have many independence, when, and how we run. In this series, I open our attempted attempts, the issue highlights the important idea associated with the test.

And here we will share the amazing effect from the starting effect on our bonus program and use it to discuss your decision (at least when including people).

Background: It's time and we need to rent a million nurses

Intlycare helps health facilities related to nurses. We are an awarding machine for the glorified nurses and therefore we always want to hire successfully. The nurses come to us from many sources, but those who come in transmit receiving high reviews and staying with us for a long time.

The year was 2020. Intlycare was a child's company (many principles). Our app was a new job and many features were first. Some examples …

  • We had an intelypos method to share the transmission link with friends but had no financial motives for doing so.
  • Our application process was a major lump. We need a small mountain of reviews above the phone conversation and indicators. Only a small set of applicants made all operations method.

During the psychiatric, we crossed the imagination of the transmission and agreed that adding financial incentives would be easy to examine. Something like, “Get $ 100 when your friend starts to work.” Zero's making there, but the idea doesn't have to have a novel to be good.

Knowing that many people may also look at a bonus, and know that our request process is nothing short of Gauntlet, and we wondered if it would be better than giving doctors tiny prize where their friends evacuate The application.

A little prize of something simple vs a prize for something difficult? I mean, depending on many things. There is only one way to know the best, and that try.

Exploration testing

We are occasionally assigned to some doctors in one of the two experiences:

  1. The clinic earns additional $ 1 / hour in their next shift when their transfer begins the application first. (Very easy. The initiating system takes 1-2 minutes.)
  2. The clinic earns $ 100 when their transfer completes their first change. (It is difficult.

We held a third of the doctors as a way of control and allowed doctors to know the rules through a series of emails. It is always risking the test of the spills like this, but the thought of a single group that stole all the passengers were appeared to be widened, so we felt well with each expansion.

Determined by non-social: Many people heard about the two choices and asked, “Did you think about trying motivations to motivate?” (Example: I cancel you, do something, we both Receive the award). Studies show that they are much better than individual motives and are common (Instacart, Airbnb, crime, …). We've considered this, but our financial team was very sad in our thoughts to send $ 1 to one hundred people who may not have workers.

I think QuickBooks don't like that? Sometimes, you just accept that it is better not to live with a finance team.

As $ 1 / hr reward could not be a refined texts without being a major head, we pay limited to both programs that refer to each person. This gives us two referral programs when a significant difference is a matter of time and money paid.

It turns the time and introduction of incentives. More. Social advice is also important. More. If you try to grow-hack your transfer system, you will be smart to look at both types before raising payment.

Undy side: Thinking about exams

Product Data Science, separately, is interested in how people interact with things and things. Usually the app or application, but it can be a visible factor such as the pair of headphones, thermostat, or mark on the main road.

Science comes from the converting the product and looks at how people can change their behavior as a result. And you can't do better than watch your customers work with your product in the field to learn how change was helpful.

But you can't test everything. There are no endless testing and any team provided by the workforce will need to determine the objects under the fullest set of ideas. Where do you start?

  • Start with product itself. Ask the people who are accustomed to how they love, what they wish was different, Sean Ellis, NPS, Mother Test, etc. This is a common point of product and personal product groups.
  • Start by human status. Many scientists will behave in decades and write some patterns in human behavior. These scientists have different names (moral values, ethics, ethics, etc.).

In my humble demonstration, the transactions of these first areas is overwhelming. Conducting science writes many behavior patterns that can inform you how much your product can change.

It means a few honored honorable romantic …

  • Aversion loss: People hate losing more than loving
  • Peak-End: People remember the best things when they end with a good note
  • Social Market habits: Everything changes when people pay for goods and services instead of asking kindness
  • Framing: People make decisions based on How Details are introduced
  • Left-digit appointment: Center ideas are influenced by leading digits ($ 0.99 = 🔥, ​​$ 1.01 = 🥱)
  • Avent Bias: People Hate You Wait

These types of psychologists do not belong to product silver and marketing. We've test many of these in different situations so that it is not necessarily, but others work. Maybe you'll get $ 160m under your bed strands.

To zoom in the persias

Many people show a moral pattern where they choose small, fast rewards over the great rewards in the future. Social scientists call like such options and measure you with difficult questions such as …

People who quickly choose, small rewards have a certain amount of current selection.

Current employment may vary from humans and circumstances. You can be a patient in many cases. But if you are hungry or tired … not so much. This is one of the reasons why A / B test is so useful in comparison, means, the survey or discussions.

With our efforts, the question is not as much different in lab questions:

  • Would you like to have ~ $ 8 sometime this week or $ 100 Sometimes in the coming weeks and maybe?

Test results

Many of us (current company installed) think that $ 100 offer can do much better. It sounds great. It looks better in email. It's easy to explain.

I thought that we could get more dogs From transmission of $ 1 / hr application, but did not assume that that would continue to be transfers. I found people to take money and run.

I was surprised that the $ 1 / HR program brought many apps by transfer and more doctorseverything at a principal low cost.

Photo by the writer

A $ 100 offer has led to 65% increase in the comparisons of our control group. This is great !. Many people like this type of program because we only give rewards when the transfer is effective. Sounds like money is well used.

$ 1 / hr, however, is best. Transfers increased by 81% compared to our control group. And although the rewards are small and paid before, the transfer of these aclypro materials continue to work in the same amount as a group of £ 100.

Yes, most of the aclypro has received the transfer rewards and fails to cross the finish line, but the figures are still valid. Or in the incomplete conversion price from the app to function, the total cost of the new Atlyprospo was less than half of the cost of $ 100.


ABT: always check regularly

The role of business play can change for many reasons. As the company becomes large and very familiar, the transfer is very important. You don't need people to spread the word because the word is already out.

We ran to 2020 during the global epidemic. We were conducting a little test year later and we saw the same results. Can we see the same results now, in 2025? It's hard to say. Items are very different from health care workers these days.

This is also the reason why the test is very important. Something like transmission bonuses can work for company wonders for a while while the company's DUD B. UPSIFIs product tests that the world is your laboratory. Lowly is what it is difficult to know what is in other products.

Important tolerance for those who did it so far

  • If you are running exams to improve your product and collect data from people using your product, behavioral science is a fruitful facility to check for testing ideas.
  • Sometimes visions that seem very dependent has a profound effect on!
  • Time and launch is produced by motives. Before you get rid of some payment size, perhaps you have made thought of the time and effort needed to get that payment.

(This post was repaired from My 2022 post office Hosts Home Hostbook

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button