An Era for Startups:
While success for startups is far from assured, resources are widely available and barriers to entry have never been lower. Whether that be access to capital through crowdfunding and VCs, distributed talent pools, or the proliferation of startup incubators, new businesses are forming at exponential rates (and not just in Silicon Valley!).
Finding the right product/market fit is just the first step. Startups need to establish a plan to scale or they risk going the way of others who didn’t cross the proverbial chasm. For many startups, scaling will include data or back office work that fuels software features, business processes, or tasks that can’t be accomplished by technology alone.
Contrary to the recent focus on robots/AI replacing humans in the workplace, there will always be a gap between human intelligence and technology. These gaps involve transforming unstructured/semi-structured data into being usable, searchable, and actionable. Some examples include image recognition/photo tagging, image moderation, webscraping/lead generation, document digitization, audio/video transcription, etc.
When a startup begins to grow, these workflows become tedious, hard to manage and often unscalable. How an early-stage business handles these challenges will have a major impact on their success or failure.
So, what are the options for getting data work done?
- Hiring In-House
The first instinct of many is to consider hiring a team in-house, especially when the work requires little skill and therefore labor costs are relatively low. Hiring an in-house team has some inherent advantages. You can train the team directly and iterate quickly because there’s an instant feedback loop. It also provides real-time visibility into the process so that challenges can be met head on through direct management and communication.
However, hiring a team in-house for this kind of routine, repetitive work presents challenges. There are scalability concerns, as the volume of work grows, so does the number of people you need. This can get expensive quickly. It can also distract from the core focus of the business, for instance, a tech company needs to focus on innovation and growth, not on managing a growing workforce of low-skill workers. An expanding team creates management headaches as well. Employee churn presents a major barrier to scaling an internal operation as well.
Another path to consider is traditional Business Process Outsourcing (BPO). It’s an attractive option with some pros and cons. Typically, going the BPO route is less expensive than hiring an in-house team. BPO companies provide a dedicated workforce in countries like India and the Philippines. Those teams can be trained to work according to your business rules and they’re typically quick to get things rolling.
However, oftentimes they lack the quality and flexibility that fast moving companies need to compete. BPOs often play a game of bait and switch, using their best workers in the initial ramp up, then switching to lower quality workers to boost margins. In addition, while BPOs have improved their use of technology to accomplish work, they’re not innovators and they struggle to provide real-time visibility into worker performance as well as quality when it comes to high-volume data work.
Crowdsourcing work through platforms like Amazon’s Mechanical Turk is increasingly popular. This is a great platform for the Do it Yourself (DIY) crowd when fairly simple or project based data work needs to be done. Businesses can fill out a form, provide business rules, set the price they’re willing to pay for each task and assuming the task and price are attractive enough there is a workforce willing to accomplish that work. The process is straightforward which allows businesses to get up and running fast.
When considering an option like Mechanical Turk or similar platforms it’s important to know that the crowd is anonymous, therefore there is little to no ability to train or obtain any guarantees of service quality/accuracy. There is also no customer support, forcing businesses to deal directly with workers or resubmit tasks until they’re done correctly. As the volume of data increases, businesses are forced to deal with a larger group of unknown workers causing management headaches.
There is another model that combines elements of BPO and crowdsourcing, referred to as Business Processing as a Service (BPaaS). BPaaS is a managed services model that provides a full-stack approach to accomplishing data work by managing the process, technology and workforce. BPaaS represents a Do It For Me (DIFM) approach vs. the DIY approach of crowdsourcing, while also leveraging technology to facilitate and optimize the work, unlike traditional BPO companies. BPaaS solutions, like CloudFactory, offer the benefits of advanced technology to efficiently process data and a closely managed, elastic workforce that can be trained for specific business processes.
Depending on the nature of the use case or workflow, BPaaS isn’t for every situation (like 1 off projects that involve low-volume datasets). At CloudFactory we work closely with our customers to understand their needs and business rules so we build a solid foundation that can scale as our customers grow. This process may be a little slower out of the gate (2-4 week ramp), but it assures a smooth operation that meets our customer’s needs over the long-term.
Depending on business needs and the nature of the work being accomplished, each of these options carry pros and cons. Finding the right solution starts with being informed so that early-stage businesses don’t make a costly mistake that creates a barrier to long-term success.
What are some of the ways you are handling data work?