For the last 2.5 years, we crowdsourced sectors from startups profiles to accommodate all the different sectors on our platform. Fast forward to early 2016, we had 1300+ sectors/sub-sectors for startups. As you might predict, this made rendering an accurate search with all those sectors and matching them with investors a very cumbersome and tedious task. So in short; we needed a structure to streamline all the sectors/sub-sectors for a more efficient and effective search/pairing workflow.
How we rearranged 1300+ startup sectors to refine search.
At a glance
450+ startups create their profiles on LV every month, adding their sectors along with other relevant profile information. Fun fact: for the last 2.5 years, we crowdsourced sectors from startups profiles to accommodate all the different sectors on our platform. Fast forward to early 2016, we had 1300+ sectors/sub-sectors for startups. As you might predict, this made rendering an accurate search with all those sectors and matching them with investors a very cumbersome and tedious task. So in short; we needed a structure to streamline all the sectors/sub-sectors for a more efficient and effective search/pairing workflow.
Research. Analysis. Implementation
Two things we had to keep in mind as we ventured in pursuit of the right solution was to be able to do it with the same team size and work with our database of 12000+ startup profiles. After much research we got our answer from Quora; DAG (Directed Acyclic Graphs). This database hierarchy solution was used to ensure Quora’s topic parent/child relations can exist in multiple locations without the need for set parent and child classes. Here is an answer from Adam D’Angelo, Founder of Quora:
DAGs are like forests with each tree being considered as a parent and each branch a child. The most interesting part is that all these trees may also have inter-connected branches. For example - Health Food being a child of both parents: Food & Beverages and Health Care.
And then there are sectors like ‘Internet’, ‘Consumer Internet’, ‘Mobile’, ‘Hyperlocal’, etc that can encompass the whole forest. For example, an enterprise mobility startup can be in ‘mobile’ just like a mobile based quizzing startup in ‘mobile’. Two very different verticals. So note that some sectors are identified by tags that can be attached to any tree. The same goes for other tags like B2B/B2C, stage of the startup, etc. and we had to combine it all for a better search. In DAGs, the key innovation is that no child can become parent of the same parent by any permutation.
Our Solution
We defined horizontal sectors like Consumer Internet as well as vertical sectors like Healthcare, Fintech, etc differently. Our team went through a database of 35000+ Indian startups, both from LetsVenture and other external sources. We consolidated 1300+ sectors/sub-sectors (fortunately, it didn’t get more exhaustive than that) into 55 parent verticals, 150 children and 15 parent horizontals. Then, we rerouted our database to adopt the DAGs.
To avoid omission of new sector trends, we have made an allowance for ‘other’ child categories under each parent. So, if a startup needs to add a child/sub-sector not in the list, it is still possible without taxing the search and pairing process. Our team will then validate the new addition before we add it to the model.
In short, DAGs took our categorization game to the next level. Today, investors can search sector-wise, sub-sector-wise and horizontal-wise. Same goes for startups as well. It is so much easier to search for Investors interested in their sectors.
We thought this new step was important to share! In case you have any questions or are doing a similar database migration at your startup, reach us at startups@letsventure.com
.png)
.png)
