Many quantitative and fundamentals asset managers continue to find incorporating alternative data into their investment decisions and hypotheses a significant challenge, noted panelists during the Beryl Elites Alternative Investments Conference in New York City.

Even the definition of what alternative data remains somewhat hazy for the industry, according to various speakers.

For Olga Kokareva, head of data sourcing and strategy at Quantsellation, she separates alternative data into three distinct buckets: non-traditional unstructured data that did not exist a few years ago, such as web traffic and credit cards; traditional unstructured data that hedge funds have been using for years, but is extracted from different sources using different methods; and structured numerical data that is used in non-traditional ways.

“If you used options market data to trade equities or if you use market structure data to extract alpha, that is pretty alternative,” she added.

Once asset managers manage to source alternative data, many often will find the quality of the data is poor at best.

“About 80% of a data scientist’s job is janitorial,” said Joe Rothermich, senior director, quantitative research and data science at Refinitiv.

“It is a huge pain for us and takes a lot of time to clean up alternative data and make it usable,” agreed Quantsellation’s Kokareva. “If we see value in a dataset, we will work with the vendor an explain what we need. Many new vendors how have not worked with hedge funds do not know about timestamps or that their data needs to be mapped.”

Before asset-manager Point72 integrates a new alternative dataset into its processes, the dataset goes through rigorous compliance filters and must be able to simplify a complex issue, according to Sameer Gupta, director, market intelligence at Point72.

Once firms source, clean, and incorporate alternative data into their processes, they also need to pay attention to alpha decay, which may occur when data vendors start changing how they acquire their data.

“It takes time to work with the vendor and make sure it is consistent,” said Kokareva. “One way to maintain return-on-investment is recycling the dataset when you bring new datasets onboard. Audit your data library and find new applications for the data in-house.”