Synthetic data: More than just make-believe

Digital marketers are constantly working with real data. What the online shopper does say a lot about what he wants. But as you know, you have to be very careful with personally identifiable information (PII).

You can anonymize online shoppers by removing their names from the records before analyzing the data. Or you can use an algorithm to synthesize the observed online behavior and use this ‘synthetic data’ for your analysis.

This may seem like an exaggeration. Why worry if you have the correct data? Synthetic data will not replace actual data, but specific use cases that a digital marketer may find useful.

Privacy of Synthetic Rights

“The use case is important,”, founder of multiple, an artificial intelligence consultancy in Germany. In that case, data protection legislation, such as the European GDPR, becomes a factor.

As a marketer, you need data to conduct experiments and optimize prices. This data also includes personal data. Personal data cannot be stored at any level. Synthetic data must bypass the privacy problem so that digital marketers can simulate campaigns and results.

“Synthetic data has some limitations in terms of real customers and their real behavior,” a researcher and machine learning specialist at Unity Group, a digital trading company in Wroclaw, Poland. In situations where the ability to obtain regular data is limited (e.g. GDPR compliance or limited dataset size), synthetic data can be a good representation.

“In most cases, the anonymity of the actual data looks better. Anonymous data includes all models taken directly from reality,” “However if there are sporadic or anomalous cases in our data. Traditional methods of anonymity fail, even if anonymity is formally compatible with the GDPR, companies that don’t use synthetic data to protect outliers could lose their images. Anonymous leaks.

Reality is being confused in a positive sense

After all, working with the right thing has advantages. “With the right data, an analyst can ‘discover hidden nuances and patterns that are not revealed by other techniques,'”, CEO of Beyond the Arc, a San Francisco Bay Area company specializing in strategic management algorithms. . the data could also cause a fatal error in identifying these business models, he said.

Predictive models depend on different data sources and groups of models, “There is the potential to use synthetic data to extend datasets and deliver more data where it is scarce.” It is the analyst’s job to understand the integrity of any data source.

“Synthetic data can never be as accurate as real data,”l. Although based on real standards, synthetic data always lack the essential “reality factor” and can therefore only be used in a limited number of cases.

“It is more difficult to distance yourself from the source data,”  Most algorithms repeat the distribution in the source data. “The errors can be repeated even in the synthetic data”.

Notice the plastic opening

Machine learning requires a lot of data, Dilmegani notes. Data marketers may need to purchase synthetic data to get the right data training for an AI application. “The demand for synthetic data is increasing”,

One application for synthetic data could, for example, be AI training to drive a self-driving car. Synthetic data was also used for the deep learning programs needed for image processing, says Dilmegani, a technique that has been around for nearly a decade.

‘I’m skeptical about using synthetic data. : “Building a machine learning / artificial intelligence model is not enough.” This is the heart of machine learning when it comes to artificial intelligence. About 60-80% of the work on creating an AI model is devoted to collecting and preparing data,  explains. This process is “the job”.

“The request is to apply an algorithm or process to create new data points,”. Synthetic data is created through a process that is also subject to disruption. We often see data as the ultimate source of truth … We often talk about how to make the data talk: “What do they say when the data is produced?

 ‘In terms of accuracy, summary data can in some cases be an opinion with factual data.

Synthetic forecast

Applying synthetic data to digital marketing is an evolution, not a revolution. Applications are limited and demand-driven. This will be another feature in the toolbox. “

Machine learning consumes a lot of data, so the demand for data can increase the use of synthetic data, Dilmegani adds. Like many things in machine learning and artificial intelligence, synthetic data will evolve. Because the use cases are limited, digital marketers will have a better idea of when the synthetic data matches well and when it doesn’t work, he said.

Translate »