It's possible that how much effort a candidate puts into demonstrating token emotional attachment to a job posting is not a great predictor of job performance.
Your strategy is yours to choose.
This seems like a reasonable deal to me.
PS: I had someone insist on the opposite process where they only looked at people who HR rejected, their team was awesome if a little odd.
Clearly this is not going to scale, but it really depends more on other companies being defective not a poor culture at your company.
PS: I am more of the opinion that focusing on great people is a mistake. Most competent people can be great, when in a role and team that fits them.
You do need to demonstrate that you can solve the problems of the potential employer.
People advertising for a job have a business problem to solve. If you can demonstrate that not only can you solve that problem but that you have solved similar problems for other businesses before, then you stand a greater chance of making it through to an interview.
However, software engineering is not such a job. A cover letter showcasing a software engineer's ability to author vaguely reality-flavored prose and get past the HR drone doing the reading is not showcasing their ability to perform critical job functions to address business needs. Such an engineer is spending significant amounts of their time to tick a prediction-value-free box on someone else's ill-informed list.
I've literally never had a good experience with a company that wanted a cover letter. I've done it dozens of times. In every single case, with precisely zero exceptions, it's produced no value for me. I've been on the other end of hiring processes as well, and there I've similarly found precisely zero value in cover letters across dozens of cases. Fact-dense and fluff-light resumes are much more valuable.
Code samples? Great! Toy problems? Great! Prose? Not so great.
Do you need help solving the problem you appear to have where you may be misevaluating candidates by measuring them wrong? I have some ideas! It could be costing you the best candidates!
> I've literally never had a good experience with a company that wanted a cover letter.
To provide a counter point, for a recent contracting job, I got an interview with a cover letter only (no resume), and then the position.
I had responded to a 'we're hiring' link at the end of a technical blog post (linked to from HN). I wrote a brief email introducing myself and talking about the value I thought could bring to the company (with links to examples of prior, related work), and that led to follow up emails and an interview. I don't think it would have gone any further if I had just sent an email saying 'Please find my resume attached' and then sent a generic resume.
When I'm applying to a company of greater than fifty or so people, I tend to assume that any general-purpose resume-inbox will be read primarily by an HR person or recruiter. Not personally by a hiring manager. That would almost certainly be a poor use of time.
What, precisely, does a cover letter show that a well-composed resume doesn't for an in-house full-time engineer? A willingness to accept the received wisdom that authoring prose on demand demonstrates critical core competencies for software engineering job performance, perhaps?
To cap it all off, I've found that at scale the improved response rate from submitting applications with cover letters doesn't justify the increased time spent applying. Doing the reading and writing to create a halfway decent cover letter takes at a minimum 10x the time it takes to click "Apply with LinkedIn".