
Headstart uses Machine Learning to help companies decide on who to…
Turn this teardown into a decision-ready prompt for ChatGPT, Claude, or your agent.
If you only have a few minutes to spare, here’s what investors, operators, and founders should know about Headstart (S17).
Headstart sold a difficult promise to large employers: screen high volumes of applicants faster while reducing, rather than automating, the bias already present in hiring. Its system ranked candidates using qualifications, psychometric inputs, and contextual data instead of relying only on CV pedigree. By 2019, the London company named Accenture, Lazard, and Smiths Group as customers and said its software had supported thousands of hires.[1] [2]
Headstart's independent chapter ended in an acquisition. Silverback United bought it for $14 million in stock in August 2022 and kept it as a wholly owned subsidiary.[3] The deal valued its data, customer relationships, and technical team, but the public record does not disclose revenue, investor returns, or the value ultimately realized from the buyer's shares. Headstart proved demand for fairness-aware screening inside enterprise recruiting, then exited before it established a durable independent category leader.
Nicholas Shekerdemian began with a personal frustration. After leaving Oxford and running an online business in China, he confronted the repetitive graduate-job application process and decided to build a better match between candidates and employers. He teamed with technical co-founder Jeremy Hindle, whose background included founding MENTIUM and working on machine learning.[4]
The first product was pitched as more than a CV filter. It created a candidate "fingerprint" from personality, interests, skills, qualifications, experience, and social context, then matched that profile against a role. A 2016 account says the system integrated with employers' recruiting software, used a chatbot to collect applicant information, and was being trialled by blue-chip companies including Vodafone.[5] Headstart joined Y Combinator's Summer 2017 batch.[6]
The founding insight was sound: recruiters must reduce a large pile of applications, yet credentials alone are a noisy proxy for potential. Headstart tried to make the ranking richer and more consistent. That created a second, harder problem. Any model trained on historical employment data can reproduce the very exclusions it claims to remove. The company therefore had to sell speed, fairness, explainability, and trust at the same time.
Headstart sat behind an employer's application flow. It combined the job description and information about successful employees with each applicant's CV, assessments, and questionnaire responses. It also used public contextual datasets, such as school performance and indicators of social disadvantage, to judge achievement relative to opportunity. The output was a percentage suitability score rather than a binary pass or fail.[1]
The product later split into two deployments. FairScreen added ranking and bulk eligibility checks to an existing applicant-tracking system. The broader Headstart platform could replace the ATS, adding candidate pipelines, automated messages, analytics, assessment scheduling, and reporting. Match Scores were configurable by role, while diversity analytics showed where different groups left the process.[7]
That architecture addressed a practical buying constraint. Large employers rarely replace a core HR system to test one new screening method. An overlay reduced switching risk, while the full suite created room for larger contracts. Human recruiters retained the final decision. Headstart's software narrowed and ordered the pool; it did not formally hire people on its own.
Read the complete post-mortem, the rebuild playbook, and the exact reasons Headstart is still worth studying now.