The data backs up that it's not hollow marketing-speak to call this past quarter our most productive on record (thanks, data!). Through a few months of long nights and busy weekends, we have chased down a functional implementation of pull request code review on GitClear.

GitClear's Diff Delta per month over the past 3 years. Q2 2024 was the highest three-month average in this range, thanks to the new pull request implementation

In this blog update, we breeze through a summary of the new PR tool. Plus the results of the expensive research we funded to evaluate the extent to which we need to compromise review "accuracy" on behalf of speed.

This post also marks the debut of our "Data Insights Program." That's the fancy name for a new subscription option that lets you delegate busywork tasks like "configuring charts," "parsing through notifications," and "translating GitClear data into your KPIs" to us. We can also work 1-on-1 with developers from your team to help create awareness around how your devs can use data to prove how smoothly the team is running.

If you've got 5 minutes to spare, we have pictures and videos to tell the story:

link🗜️ Pull requests: 30% less code to review than GitHub, plus newly published research

After 6 months, our pull request tool has finally exited beta and is ready to be put through its paces by any team that wants to spend less time on code review. I made a 7 minute video showing how GitClear's "commit cruncher" (our diff processing engine) manages to trim so much reviewable content:

Click the pic to watch our latest Youtube video explaining how it's possible to reduce so many lines to review

The feature release also comes with the release of our latest long-term, high-investment research: 30% Less is More: Exploring Strategies to Cut Pull Request Review Time, a shareable PDF version is also available.

You can also check out our 7 minute video on the pull request review features we launched beyond the diff viewer itself.

link👥 Data Insights Program: ACT on your data for measurable growth

GitClear's data can give you mountains of interesting insights about the way your teams function, their strengths and weaknesses, and key areas for improvement. Now, you can combine your mountain of data with a GitClear Data Specialist. That means you'll get hyper-personalized recommendations based on your KPIs and company goals.

Some benefits of the Data Insights Program, from the Data Insights landing page

It's one thing to gather data about your code quality, change velocity, and pull request efficiency. It's another to actively synthesize that data into specific plans that change the trajectory of your company.

We've seen hundreds of companies, and learned several common patterns that break down the rate at which software gets delivered. We can bring that experience and insight to your organization. The payoff: more time to focus on your core mission, and visible proof to executives that substantiates your management strategies lead to measurable improvement.

Here are a few specific examples of how you can leverage the Data Insights Program to drive results:

Configuration assistance to ensure that your settings maximize what you want to improve going forward.

Metrics breakdown helps pinpoint, configure, and monitor which charts among the 30+ that GitClear offers are best aligned with your existing KPIs.

Ongoing coaching with routine video check-ins to pick out the most significant and unusual trends in your data. We work with managers and developers to keep everyone in sync.

For teams that want to extract the maximum possible potential of their developers, we offer our Advanced Data Insights program:

Sharable analysis reports means you get pre-packaged reports, easily sharable with your management or developers. We guide you through creating, managing, and sending these.

Prioritized feature requests so we can tailor our development to your particular needs, team-by-team.

With the Data Insights Program, you can leave the ins and outs of understanding GitClear's metrics to us.

At $9 per month per contributor, you can add Data Insights to your subscription and still end up paying less than any of the sales-centric developer analytics companies. Signing up is easy - just visit the overview page and select the option to add it to your account. We'll be in touch with you within a business day.

link📣 New actionable notifications

This quarter harkened the debut of three new notifications for users to opt into.

linkHigh bug work

To a Lead Developer or Manager, a high rate of work on bugs can be a red flag that their team is getting bogged down in fixing what's already created rather than adding new features or improving existing systems. To a CTO or VP, it can indicate that their existing code quality measures are insufficient for delivering software as reliable as they'd hope. These can be strong lagging indicators of problems further up in the development process, and they warrant serious attention.

GitClear now highlights when Diff Delta focused on bugs is higher than recent weeks:

Currently, this is set to detect weeks that have a high share of work in bugs (in the 85th percentile or above, among weeks in the past year, for bug-fix percentage). This way, you can have an eye for when your development focus is potentially trending in the wrong direction.

linkHigh pull request post-merge work

As part of ensuring a high standard for quality, development teams frequently have a keen interest in minimizing the amount of work that occurs to revise work from a pull request after it's been merged. Such changes often imply hot-fixes and work on bugs that could have been caught during the review process.

GitClear now highlights when teams exceed their set threshold for post-merge work. When a week where a repo's post-merge work exceeds this threshold, you'll now receive this notification either through email or on your Highlights dashboard:

You can adjust this threshold by going into your Settings -> Code quality targets page and adjusting "Max % desirable post-merge Diff Delta"

linkHigh pull request under review work

One of the clearest standards of a team that can deliver work quickly and effectively is a smooth PR review process. A high amount of accumulated Diff Delta (ie: high change volume) during review implies that much is changing as a result of feedback given during the review process. This can lead to slower deliverables and a team busy giving, and evaluating changes to, feedback, rather than writing code and delivering features.

To alert you to this, GitClear now highlights when teams exceed their set threshold for PR review work. When a week where a repo's post-merge work exceeds this threshold, you'll now receive this notification either through email or on your Highlights dashboard:

You can adjust this threshold by going into your Settings -> Code quality targets page and adjusting "Max % desirable review Diff Delta"

link🎫 ISO 27001 Audited and Certified

GitClear completed a thorough pentest and ISO 27001 audit this quarter. You can read more at our security page. Next up: SOC2 accreditation, which is on track to be completed by our next blog update.

link🔮 What's next

Our pull request tool has evolved rapidly to get to this point, and it figures to continue evolving rapidly as we fine-tune the LLM that generates file summaries and suggested PR changes. If you like GitClear's PR stats and review tool today, we think you'll like it 1.8x more by the next blog update. 😅 In addition to improving its ability to predict which comments will be left, we believe it should be possible to label code based on the likelihood a comment will be left.

Our SOC2 certification is already more than 75% complete, so passing the final audit for that ought to be complete in advance of our next update.

We also plan to continue refining and adding to the notification system, so that you can get more data from GitClear without needing to manually visit 20 different reports. If you have an idea for a type of code or PR signal that should be instrumented, drop us a line and there's a very good chance we could begin to detect when its value strays away from the expected norms.