Jasmine Jia, associate director of data science at Blockthrough.

Machine learning isn’t magic – it needs strategy and a human touch | Ad Exchange

Data-driven thinkingis written by members of the media community and contains fresh insights into the digital revolution in media.

Today’s column is written by Jasmine Jia, Associate Director of Data Science at Blocking.

The term machine learning” seems to have magical effect as a sales buzzword. Associate this with the term data science,” and many companies believe they have a winning formula for attracting new customers.

Is it smoke and mirrors? Often the answer is yes.”

What is very real however, is the need for best practices in data science and for companies to invest in and fully support the talent that can apply these principles effectively.

Laying the groundwork for machine learning

Machine learning success starts with hiring talent that can leverage machine learning – a team of skilled data scientists – which is very expensive. Time adds to the cost. It takes a lot to build a data science team and integrate them with other teams across operations.

A successful machine learning pipeline requires data cleaning, data exploration, feature extraction, model building, model validation, and more. You must also continue to maintain and evolve this pipeline. And not only is the cost high, but companies rarely have the patience and time to manage this process while meeting their ROI goals.

Define best practices

With the right talent and the right pipeline in place, the next step is to establish best practices. It’s essential. Machine learning depends on how you implement it, what problem you use it to solve, and how deeply you integrate it into your business.

To paint a picture of how things can go wrong, just think of the times when lopsided data sets led to what the media called racist robots” and automated racism.” Or, on a lighter note, how about those memes showing machine learning mistaking blueberry muffins for Chihuahuas. Or mix images of bagels with photos of curled up puppies?

Best practices can avoid some of these common pitfalls, but defining them for the entire data analysis process is critical: before decision-making, during decision-making, and after decision-making.

Lets take this step by step.

Before: It is too common for companies to update an offer by adding a feature. But often they do so before they have finished collecting and analyzing meaningful data. No one took the time and resources to respond, Why are we adding this feature? »

Before answering this overarching question, other questions need to be addressed. Do you already see users adopting this behavior naturally? What will be the potential upside? Is it worth the expense and time to tap into your engineering resources? What is the expected impact? What would this new feature ultimately mean for the future success of this product?

YouIt takes a lot of data to answer these questions. But lets say that you have eliminated everything and decided that it was worth moving forward.

During: Youlaunched this feature. There must be a continuous flow of data that demonstrates whether or not the new feature has an impact at the network level, publisher level, and user level.

Do you see the same impact at all levels? Sometimes the benefits for one can be detrimental to another. You have to be careful. Factor analysis is essential. What are the factors at play that impact the analysis? Once identified, you need to determine if they are physically significant or not.

After: At this point, there are still more questions to address. What is the impact exactly? If you use A/B testing, can these short-term experiments provide reliable long-term predictions? What lessons can you learn? Whether it’s a failure or a success, how can it continue to evolve? What are the new opportunities? What new behavioral changes do yousee again.

Machine learning for the long haul

There’s a lot of data and oversight needed to make a machine learning program truly viable. Thisno wonder so manyYou don’t have the means to execute it properly and reap the benefits.

Here’s the kicker: the data team doesn’t make the decisions. The machine learning algorithm does not make the decisions. People make decisions. You can hire a fantastic team of data scientists, and they can create and refine a machine learning model based on 100% accurate masses of data. But for it to make a difference to your business, you need to develop a solid workflow around it.

The best way to do this? Make sure data science teams are deeply integrated with different teams in your organization.

Establish an ingrained data science practice and you’ll see that machine learning can work the magic.

Follow Blockthrough (@blockthrough) and Ad Exchange (@adexchanger) on Twitter.


#Machine #learning #isnt #magic #strategy #human #touch #Exchange

Leave a Comment

Your email address will not be published.