By Jaya Kumari
Jobtech founders need to be aware of how this common issue can affect their startups and understand strategies to mitigate this issue.
The “winner-takes-all” phenomenon exists in multiple spaces ranging from business markets to the sports arena. In sports, the phrase refers to the winner of a competition getting it all–endorsements, fame, etc—leaving second place with not very much. In computing, winner-take-all networks refer to competitive learning via output nodes in networks inhibiting each other until only one output node is active. In the mountains of East Africa, alpha silverback gorillas are able to mate with 90%+ of the females in their family, leaving the rest to be shared amongst the other silverbacks. In business, it refers to markets where the best performers are able to capture a large share of the available rewards. Such scenarios are good for the winners, but generally bad for the markets they’re operating in.
The “winner-takes-all” issue can especially plague jobtech marketplaces. Many marketplaces utilize algorithms that create a reputation score based on a variety of inputs. User feedback is an important input here—it allows platforms to ensure that high-quality buyers are paired with high-quality sellers. The issue occurs when the earliest sellers are able to establish a high reputation score, and the algorithm doesn’t allow for newer sellers to attract high-quality buyers. In turn, the newer sellers fall off the platform after a frustrating lack of sales, which in turn hurts the overall platform offering. And while this is fine up to a certain point, it limits platforms’ capacity to scale.
This winner-takes-all model also masks hidden hierarchies where one account wins the work and then divides out the work to others, which can lead to quality risks and diminished network effects. Platforms like Upwork have identified this issue and actively warn against it. Ultimately, if your platform gets built around a small number of sellers, your platform is at a greater risk of them leaving and taking customers with them, or disintermediating you.
The challenge is that this phenomenon can happen really easily in marketplaces, even if it is not deliberate. Chris Maclay, the former COO of Lynk – the gigmatching marketplace for blue collar workers in Kenya – explained, “Our model hypothesized that customer ratings would enable informal workers who otherwise had no provable reputation to get more work. The hypothesis was so correct that it almost became a problem – with general distrust of informal workers so high, at one point, all the work was aggregating around a few users who had the most ratings. Even when we knew, for example, that John wasn’t a great painter compared a newly onboarded painter, John would continue to get the jobs because he was onboarded first and had the most ratings.”
If you find that your jobtech platform faces this all-too-common issue, it’s not hopeless from here. There here are ways of identifying and mitigating this issue:
Decide what desired utilization looks like for your users: Is it 1 job in a day? An hour? A week? Your platform needs to optimize utilization by deciding what “good” looks like and building operations around it. For example, many ride hailing platforms use ‘floors’ and ‘ceilings’ – adding more people to their platform when they pass a certain surge rate (floor) and removing drivers when workers have less than a certain amount of jobs per hour (ceiling). Think about what good looks like for your platform.
Pay attention to the issue through accessible data: The most important thing is to recognise the issue itself. Is the growth coming from top sellers? Are a few sellers capturing the majority of buyers on your platform? Is your ‘average sales per user’ figure functionally a vanity metric, as it is completely unevenly distributed? In an ideal world, you have this seller data accessible via dashboards you can easily analyze. For example, it could be very helpful to track seller utilization more granularly than ‘average platform sales’ (which could be misleading) or analyze annual seller revenue compared to seller tenure on your platform. Ensuring you have easy access to this internal data will help your team accurately pinpoint where and how these issues are occurring on your platform.
Invest in new user conversion and skill-enhancement through supporting users: Provide adequate instruction and training for new sellers on your platform to make it easy for them to understand your platform and get their first jobs. Training can also encourage lower-skilled sellers to join the platform. This instruction can be given in different formats, depending on your user demographics. You can try video instruction, online primers, in-person training, or even paper pamphlets. Examples of this that other marketplaces have employed include:
- In order to aid new seller conversion, Lynk used to give really strong guidance on ensuring that they asked ratings from buyers in their first gigs.
- Uber hosted in-person tutorial sessions in Uganda during the driver onboarding process in order to attract drivers that may be less familiar with smartphones. Not only did this serve as an attractive selling point to attract drivers, but it also enhanced driver quality in this emerging market.
- Beyond buyer and seller training, Glovo has invested in developing a state-of-the-art training platform for their customer service agents. They utilize personalized images, descriptions, and online courses to deliver a customized training experience. After the course, Glovo also provides certification and asks for feedback.
Level the playing field at platform-level: There’s a wide variety of strategies that you can employ here, but they all revolve around either supporting the newest sellers on your platform to convert into reliable users, or working on your platform’s subsequent distribution of work. You can offer discounts to buyers for trying and reviewing newer sellers. To be clear, the goal isn’t to lower the quality of the sellers on your platform but rather give newer sellers the opportunity to thrive on your jobtech platform. Examples of this that other marketplaces have employed include:
- Encourage and reward new sellers: One strategy your startup can employ is encouraging and rewarding newer sellers through competitions–take a look at Fiverr’s e-commerce contest here as an example. Showcasing them on your website might encourage a user to take a chance on a promising new seller that has been validated by a competition.
- Don’t let customers choose: This may sound counterintuitive as you may assume that seller ratings drive buyer choice, but once your platform has established a reputation, your buyers will trust you as much as the sellers.
- Guarantee a job within a week of sign-up: Combined with the above strategy, you can force early conversion using this strategy.
- Ensure wider utilization: On an ongoing basis, change your platform’s algorithm itself to ensure wider utilization. For example, don’t always show the users with the most reviews, or deprioritize them in the listings.
- Change your algorithm’s new seller default settings: Automatically score sellers 5 stars when they’re being onboarded until they receive their first jobs. Additionally, you can use proxy data to deal with a seller’s cold start. At Lynk, new sellers were asked to provide 3 references, and these references’ ratings and reviews were used to get the buyers’ first 3 ratings on the platform.
Overall, the winners-take-all issue isn’t the simplest to identify and fix, but being aware of and fixing this critical issue can help accelerate your platform’s growth.
Jaya Kumari is an Associate Consultant with Bain & Co, working with the Jobtech Alliance as a Venture Building Extern.
Read more here:
Thank You. This is really an eye opener. It gave me a new perspective.