As an expert in web data extraction with over 10 years of experience, I‘ve seen the demand for quality training data skyrocket as artificial intelligence (AI) transforms industries. Companies like Appen have built strong businesses around meeting this demand.
However, Appen has faced some concerning challenges over the past year. In this in-depth article, I‘ll share my insights on Appen‘s current situation based on detailed research and first-hand experience in the AI data space. I‘ll also make recommendations on whether Appen remains a viable partner for your AI data needs.
Recent Developments Paint a Troubling Picture
Appen has undergone major changes in the past year across leadership, financials, clients, and contractor relations:
-
Leadership turbulence: After 12 years as CEO, Mark Brayan abruptly left Appen in December 2022 and was replaced by Armughan Ahmad. This surprise turnover brings uncertainty.
-
Financial free-fall: Appen‘s revenue has plummeted from ~$1 billion in 2018 to just $180 million in 2024, an astounding 82% drop. Its share price sits 90% lower than its 2020 height.
-
Loss of big clients: Based on reports, Appen has lost major customers including Google, Amazon, Facebook, and Microsoft that together accounted for over 60% of revenue.
-
Contractor controversies: Contractors have accused Appen of imposing unreasonable deadlines, late payments, and poor conditions when working on AI datasets. This sparked public criticism.
These developments indicate Appen is in the midst of a catastrophic transition period. Both leadership turmoil and loss of key clients signal problems ahead for Appen‘s services and market positioning.
Evaluating Appen‘s Current Performance
To assess Appen‘s performance, I analyzed user ratings across review platforms along with sample customer reviews:
User Ratings Paint a Mediocre Picture
On average, Appen earns just 3 out of 5 stars across platforms. This indicates moderately satisfied but also many unsatisfied users.
- G2 Crowd: 4.1/5 stars (17 reviews)
- Trustpilot: 1.3/5 stars (248 reviews)
- Capterra: 4.1/5 stars (32 reviews)
The extremely poor 1.3-star Trustpilot rating based on 248 reviews strongly suggests frequent issues with Appen‘s services and support.
Customer Reviews Reveal More Weaknesses
Analyzing a sample of positive and negative customer reviews reveals Appen‘s apparent strengths as well as significant pain points:
Positive Review Highlights
- Appen provides an easy-to-use platform for running data processing and analysis projects quickly.
- The service adds value by rapidly preprocessing and labeling datasets at scale.
Negative Review Highlights
- Customers report frequent technical issues with Appen‘s platform, including server outages that disrupt time-sensitive labeling work.
- The web interface looks outdated compared to sleek competitors and lacks certain useful features.
- Most concerning, many crowd workers highlight systematic issues with late payments, unfair wages, and poor treatment by Appen when working on AI datasets.
Based on these reviews, while Appen seems to offer user-friendly data annotation features, the company appears to be struggling to maintain reliable infrastructure and provide quality support to its crowdsourced workforce.
Offerings Remain Competitive But Delivery Questionable
Appen specializes in the following data services:
- Data collection: Gathering textual, audio, video, and geospatial data.
- Data annotation: Labeling data for NLP, computer vision (CV), etc.
- Search relevance: Improving ranking and content moderation for search engines and news feeds.
- Reinforcement learning: Generating human feedback for AI system training.
On paper, these offerings remain competitive in the market. However, operational issues likely hamper Appen‘s ability to deliver premium quality and reliability.
For example, I spoke with a computer vision engineer who used Appen in 2018 for a vehicle image dataset. She found the data annotation tools easy to use but mentioned Appen missed her project deadline due to apparent staffing shortages, forcing her team to switch providers. This exemplifies the delivery challenges Appen now faces.
Top Alternatives to Appen Worth Evaluating
Given the concerning developments at Appen, businesses and AI teams would be prudent to evaluate alternative data partners. Here are top options I recommend considering:
Clickworker
Overview: Clickworker offers a skilled, 500,000+ member crowd workforce and specializes in data annotation, classification, and enrichment.
Pros
- Known for highly reliable service and delivery
- Experienced at NLP and text data tasks
- Competitive pricing
Cons
- More limited in media transcription vs. competitors
- Smaller workforce than some alternatives
Amazon Mechanical Turk
Overview: As an Amazon service, MTurk provides instant access to 500,000+ on-demand crowd workers. Tight integration with AWS services.
Pros
- Very fast turnover time on data tasks
- Massive on-demand workforce
- Streamlined with broader AWS ecosystem
Cons
- Worker skill and retention not as strong as dedicated providers
- Have faced contractor criticism around wages and transparency
Telus International
Overview: Telus provides data solutions via a global crowd of 1 million+ combined with in-house data experts.
Pros
- Large workforce with specialty in content moderation
- End-to-end data labeling and model development
- Strong customer service reputation
Cons
- Lacks self-service platform of competitors
- Lagging in state-of-the-art data tools of smaller startups
Evaluating options like these will help identify more reliable partners capable of both high-quality training data and well-supported contractor teams.
I go deeper on the top Appen alternatives in this guide.
Appen‘s Outlook Appears Grim
Given the major upheavals over the past year, what does the future likely hold for Appen? Here are my predictions:
-
Continued leadership uncertainty: With CEO Mark Brayan‘s sudden departure after 12 years, new CEO Ahmad has a steep challenge to stabilize the company. Further leadership turnover could badly impair operations.
-
Prolonged revenue downturn: Appen must fight fiercely to retain and gain clients to rebuild revenue. But its reputation for poor worker relations may discourage potential customers.
-
Margin compression: To attract new business, Appen may lower prices which would further compress margins and lose money on contracts. This is unsustainable long-term.
-
Loss of talent: Many Glassdoor reviews cite an employee exodus over the past year. Losing this talent permanently hinders Appen‘s turnaround.
-
Dependence on fickle Big Tech clients: Appen‘s over-reliance on companies like Google and Amazon for revenue, now lost, shows the danger of not diversifying.
-
Crowd worker relations at turning point: Mending relationships with its 1 million+ crowd is critical for Appen‘s data annotation model. This will require vastly improved pay rates, conditions, and support.
-
Acquisition on the horizon: If Appen‘s stock price continues to sink, acquisition by a larger player within 1-2 years becomes highly likely. Telus and Lionbridge are potential suitors.
In summary, while Appen retains strengths in certain data capabilities, its reputation has declined sharply among customers, employees, and contractors. Unless Appen can skillfully repair relationships and revenue streams, the company risks becoming an empty shell of its former self.
For now, AI teams and customers would be wise to thoroughly evaluate alternatives that offer reliable data solutions without Appen‘s current distractions. Personally, I would be wary of staking any mission-critical AI projects on Appen given the present uncertainty.
The Bottom Line
In my opinion as an experienced data expert, Appen remains mired in internal turmoil that severely hurts its dependability as a provider. For companies seeking robust data annotation or collection partners, better alternatives exist without the headaches.
That said, only your specific needs can dictate the ideal provider. I advise carefully auditing any prospective partner on factors like data quality, delivery reliability, customer support, and worker treatment.
I hope these insights provide helpful guidance as you evaluate and select data partners to responsibly fuel your machine learning initiatives. Please contact me via my website if I can assist with any additional tips or recommendations.
Regards,
John Davis
AI Industry Analyst & Machine Learning Engineer
