On November 18, 2019, data annotation provider Appen, which is listed on the Australian Securities Exchange (under the ticker APX:AX), announced an upgrade to its full-year earnings for 2019.
Appen said in the earnings upgrade announcement that its underlying EBITDA forecast for the year had increased to a range of AUD 96-99m (USD 65-67m). The company had previously guided the markets that underlying EBITDA was trending close to AUD 90m (USD 61m) for 2019. Actual full-year earnings should come in around AUD 1-1.5m above the revised forecast if there are no changes to the current US-Australian dollar exchange rate.
Appen explained that it was prompted to upgrade its earnings because of an increase in revenues and improved margins stemming from demand from existing customers.
The company also restated its confidence in Silicon Valley-based Figure Eight, a technology-first data annotation platform that Appen acquired in early 2019. Guidance for Figure Eight remains unchanged from earlier forecasts — the company is expected to generate AUD 30-35m in annual recurring revenues in 2019.
A market leader in so-called “AI support services,” Appen specializes in collecting and curating human-annotated datasets for use in machine learning and AI. Recent years have seen demand for language data skyrocket. In turn, providers who operate in this niche of the language industry have experienced substantial growth, as big tech companies seek to improve the performance of their data-hungry AI applications.
Language data is indeed an attractive niche; Appen has grown significantly since going public in 2015 and new entrants to the space, such as Scale AI, have attracted much interest from investors. The world’s second largest language service provider by revenue, Lionbridge, also competes convincingly in the AI-support services space.
Speaking at SlatorCon San Francisco in September 2019, Appen CEO Mark Brayan explained the importance of high-quality underlying data in ensuring the performance of AI, saying that without good quality data, you run the risk of getting “garbage in, garbage out.”
It’s this need for quality coupled with scale that make AI-support services a complex operation. “Data is not only expensive to collect; it is also complicated to collect and it’s complicated to work with,” Brayan said. To account for this, Appen has over a million crowd workers and pays around 50,000 people to work for them each month, he told the audience at SlatorCon.
Appen closed trading with shares up 13% on market open on the day the earnings upgrade was announced. Appen’s market cap stands at USD 2.2bn, with shares trading 16% shy of the summer highs.