“Then we also want to target mid-sized companies through the Figure Eight platform, so we’re setting up teams to go after those segments. Then, [also] investing in the UK and Europe and in the Asia Pacific outside of China.”
For the 2019 full-year to December 31, Appen’s results were buoyed by demand from the tech giants for both speech and image data, as well as content relevance information.
Underlying earnings before interest, tax, depreciation and amortisation came in ahead of Appen’s upgraded guidance, growing 42 per cent to $101 million, or 29 per cent to $87.9 million on a statutory basis, assuming an Australian dollar conversion of $1 to US74¢. This was ahead of guidance of $96 million to $99 million.
The company also recorded revenue growth of 47 per cent to $536 million and a 32 per cent jump in underlying net profit to $64.7 million. On a statutory basis, its profit was flat at $41.6 million.
Appen’s underlying figures exclude costs related to its acquisitions, share-based payment expenses and a lower milestone payment related to its Figure Eight acquisition.
The $3 billion business provides the world’s biggest technology companies with the crowd-sourced data needed to train the AI algorithms that power everything from search engines to voice assistants and image recognition technology.
A highlight of Appen’s result was the performance of its speech and image data division, with revenue leaping 32 per cent on the previous corresponding period, compared to a historical three-year growth rate of 17.9 per cent.
Mr Brayan said this pick up in growth was driven by the company’s decision to forego some gross margin in order to generate growth.
“A few years ago we were quite reliant on some government clients that were very high margin. But those projects came to an end and we had to chase business elsewhere,” he said.
“We made the decision to strategically back ourselves that we could grow margin over time, so we could chase businesses in the lower margin tech sector.”
Appen’s larger content relevance division, which provides data to tech companies to help them improve the relevance of results surfaced by their search engines, also remained strong, with organic revenue up 37 per cent to $430 million, as a result of its greater scale and technology improvements.
A second highlight was the return to form of its acquired Figure Eight business, which underperformed in the first half of 2019 following the acquisition, but rebounded in the second half, thanks to a focused effort on sales and customer support from Appen, which led to new business and fewer lost deals.
From the start of January, Figure Eight has started to be integrated into the core Appen business.
Appen’s so-called crowd is composed of more than 1 million remote workers who annotate its data.
A typical task for Appen’s crowd might be sorting through a data dump of 1 million images of streets, with a request from a company to go through and tag every street sign in the photos. This would then be used to train the algorithms that will one day help driverless cars to function.
While annotating data has historically been a labour-intensive task, the business has been investing in technology – a major driver of the Figure Eight acquisition – and has started deploying machine learning algorithms to speed up the data annotation process.
It has also created an automated crowd applicant checking algorithm, which reduces the onboarding time of a new crowd member from up to 20 hours to a matter of minutes.
Mr Brayan said the efficiency gains from its adoption of machine learning were starting to materialise, but it was still in the very early stages.
“We’re piloting machine learning assisted transcription using speech recognition and we’re seeing efficiency gains in the 33-35 per cent range,” he said. “This is a pilot project, but if we can play that out over time into production work, it will be meaningful.”
Appen has also launched a “LiDAR” tool which lets its crowd annotate 3D data required for autonomous vehicles.
On the back of the result, Appen declared a 5¢ per share interim dividend, up 25 per cent on the previous corresponding period.
For the 2020 full-year to December 31, the company forecasts underlying EBITDA to be in the range of $125 to 130 million, at a $1 to US70¢ conversion rate. If the Australian dollar stays low, the business will have a currency tailwind.
Appen was one of an increasingly smaller pool of companies to report no expected impact from the outbreak of the coronavirus, with its China operations still in its infancy.