The Novee Gym is open for business at Black Hat Europe

Book a meeting

The Novee Gym is open for business at Black Hat Europe

Book a meeting

What a Bug Bounty Program Really Costs: Triage Is the Line Item Nobody Budgets For

Key Takeaways Before a single bounty is paid, HackerOne’s federal price list charges $128,362.72 a year for triage alone. That figure is one triage tier on the GSA schedule, priced at nearly the cost of the $154,035.26 subscription it supports. If a platform charges that much to read reports, the work has a real price, […]

Novee Team

10 mins

Explore Article +

Key Takeaways

  • Subscriptions and reward rates are visible lines: Both show up on price lists and reward tables before launch, though total payouts still depend on how many valid reports arrive.
  • Triage is its own line item: On the GSA schedule, one triage tier for HackerOne’s FedRAMP edition lists at nearly the price of the subscription it supports. Triage done in-house costs reviewer hours instead.
  • Unpaid reports still cost review time: ISC’s bug bounty for BIND drew 150 reports and found 8 valid, and its team still had to assess the rest.
  • Pre-launch steps cut payouts and triage: Pentesting known bug classes can cut early payouts and duplicates, scoping production assets avoids paying for bugs that don’t exist in production, and launching privately holds down report volume.

Before a single bounty is paid, HackerOne’s federal price list charges $128,362.72 a year for triage alone.

That figure is one triage tier on the GSA schedule, priced at nearly the cost of the $154,035.26 subscription it supports. If a platform charges that much to read reports, the work has a real price, and a team that triages in-house pays it in reviewer hours instead. Those hours rarely make it into a bug bounty cost estimate built from the platform quote and the reward pool. 

Simply put, triage is labor, and it runs on every report, paid or not. A team that buys it from the platform sees the price on the order form.

AI-assisted reporting widens that gap. Reports now cost less to write, so more arrive, and someone still needs to read each one, reproduce it, and decide whether it counts.

Teams that budget well price those hours alongside remediation and retesting, then cut payouts and duplicates by testing before the first submission arrives.

You’ll be able to build a full-cost budget for a bug bounty program and decide which costs to cut before launch.

The Visible Costs: Platform Fees and Bounty Payouts

Six lines make up the full bug bounty cost. Price lists and reward tables show the platform subscription and bounty pool before launch, plus outsourced triage when a platform offers it. In-house triage, internal review, remediation and retest, and policy upkeep are labor. Triage grows with every report that arrives, valid or not, and review and fix work grow with every valid one.

Bug bounty platforms sell subscriptions in tiers. HackerOne’s FedRAMP edition appears on GSA Advantage at several subscription levels, with triage sold as separate tiers that scale with the bounty budget. Those are ceiling prices for a government edition, and a commercial quote may differ. The structure is the useful part, because the platform fee and the triage fee arrive as different lines.

Bug bounty payouts follow the program’s reward table. GitHub’s guideline for a critical bug is $10,000 in its public program and $30,000 or more in its private one, and GitHub says the amounts are not guaranteed. Google’s reward table lists $101,010 for remote code execution on its most sensitive domains, then multiplies each reward by 0.8x, 1x, or 1.2x for report quality because, in Google’s words, quality “facilitates the triage, analysis, and remediation process.”

Triage Hours as a Hidden Program Expense

Triage covers every step between a report landing and a decision on it. Someone reads the submission, reproduces the issue, scores its severity, checks for duplicates, and answers the researcher. 

In its first three years, through late 2020, the bounty run by GSA’s Technology Transformation Services drew an average of 21 submissions a month, and about 68% of reports were invalid. GSA used third-party triage, so its team saw only verified reports, and moved program oversight from a rotating internal cohort to dedicated staff.

The curl project’s mid-2025 figures show what that work costs a small team in hours. Its lead maintainer described the load on the project’s seven-person security team in three numbers, with a fourth that follows.

  • Reports per week: About two, on average, in the first half of 2025.
  • People per report: Three to four team members engage with each one.
  • Time per person: 30 minutes per report, sometimes up to about three hours.
  • Weekly load: About 3 to 4 person-hours in a typical week, by our arithmetic from those figures, and more when a report takes hours, before any fix work starts.

Run your own inputs with a loaded hourly rate, and the triage line has a number before launch.

How Invalid and Duplicate Submissions Inflate Spend

A bounty pays only for valid, unique findings. Review time runs on everything else too, and in the queues below, everything else is the majority.

Invalid Reports Still Need a Full Read

ISC’s bug bounty for BIND drew 150 reports between its December 2025 launch and April 2026. Of those, 8 were valid, 8 were still under investigation, and the rest were dismissed. ISC has since paused the program and wrote that “it is now easier to generate a report than it is to read and assess it.”

Disclosure channels show a similar pattern. CISA’s Vulnerability Disclosure Policy Platform, which receives and routes reports for federal agencies, took in over 12,800 reports from researchers in fiscal 2025, of which over 1,200 were valid.

Each Duplicate Report Still Needs a Read and a Match

Programs pay the first valid reporter of a bug. Every later report of the same bug still has to be read, reproduced far enough to match, and closed with an explanation. After a rise in duplicate reports early in 2026, Mozilla extended its automated tools’ head start to seven days, so its bounty skips issues those tools find first.

AI-Generated Reports Raise the Volume

AI tools cut the effort bug bounty hunters spend writing a report, part of the falling cost of offense, and the review on the receiving end does not shrink with it. Google temporarily stopped accepting product vulnerabilities to its open source VRP on October 1, 2026, citing a rise in automated submissions, “the vast majority of which are not valid.”

Building a Full-Cost Budget for a Bug Bounty Program

Start from a published range, then add the labor. The US Department of Defense’s 2022 guidance tells system owners to expect $250K to $800K per bounty, depending on public or private access, complexity, and scope. 

DoD’s typical bounty runs live for about four weeks, so the range covers a single event. A full bug bounty cost forecast also counts the internal review and fix work behind each finding.

The table lists the six line items to forecast for a year, what drives each one, and how to estimate it.

Line itemWhat drives itHow to estimate it
Platform subscriptionTier, assets in scope, and editionVendor quote for the tier you need
Bounty poolReward table and number of valid findingsExpected valid reports by severity times the reward band, with a cap
TriageReport volume, valid rate, and hours per reportOutsourced triage tier, or reports times hours times hourly rate
Internal reviewOwner-team time to confirm impact and respondHours per valid report times hourly rate
Remediation and retestFix and verification work per valid findingEngineering hours per fix, plus retest time
Policy and scope upkeepSafe harbor terms, scope changes, and researcher communicationFixed hours per quarter


Add the six lines and divide by valid unique findings to get cost per valid finding, a figure that puts a bounty, a pentest, and an internal hire on the same basis, if each option’s figure includes its own review and fix time. The fix line can run large on its own. DISA obligated $977,844 for a task order in 2020 and 2021 described only as “XTS bug bounty remediation.”

Reducing Total Program Cost Before Opening Submissions

Penetration testing cost is easier to forecast, because scope and time set it. GSA schedule ceiling rates for penetration tester labor categories average about $158 an hour, based on our calculation from GSA’s CALC+ data in October 2026. 

At that ceiling rate, a 40-hour pretest by one tester comes to about $6,300, a fixed figure to set against an open-ended payout line. NIST’s testing guide pairs regularly scheduled scanning with periodic penetration testing, and running that work before launch can remove findings a bounty would otherwise pay for.

Five steps, most of them taken before the program opens, cut payouts, duplicates, or report volume. These include:

  • Testing known bug classes first: New programs see an early spike in reports, which Walshe and Simpson attribute to automated testing for cross-site scripting and SQL injection. In their 2019 data, Postmates spent $26,124 in its first week for 75 vulnerabilities.
  • Scoping production assets: DoD warns that testing staging instead of production can lead system owners to pay for vulnerabilities that do not exist in production.
  • Running a disclosure policy before a paid bounty: OWASP says bounties suit only organizations with a mature disclosure process, warns of junk reports and false positives, and notes that managed triage comes “at a cost.”
  • Starting private and widening later: GSA’s TTS program throttled report volume by starting each new system in a private bounty open to a small number of researchers.
  • Validating between releases: Continuous validation between releases can catch exploitable flaws before a researcher reports them, which matters more as AI speeds up vulnerability discovery.

Put Triage on the Budget Before Launch Day

The cheapest report to triage is the one a researcher never has to file because you found, confirmed, and fixed the bug first. A bug fixed before launch can’t come back as a payout or a pile of duplicates, though invalid reports will still arrive and still need a read.

Novee is a continuous offensive security platform whose AI agents pentest web applications and APIs, mobile apps, LLM-powered applications, and external attack surfaces. Its Validation Agent confirms exploitability with non-destructive testing before a finding is reported. Each finding carries proof-of-concept evidence, and one-click Retest replays the original exploit path after a fix ships.

List the applications you plan to put in scope first, with the bug classes you expect researchers to report. Book a demo today to see how Novee tests those applications before the program opens, so fewer reports reach your triage queue.

FAQs

Plan on weeks to a few months, depending on scope and procurement. The US Department of Defense runs its bounties on about four weeks of planning, four weeks live, and four weeks of wrap-up after a contract is awarded. Writing the policy, safe harbor terms, and scope, and lining up triage staff, comes first.


A private program is often cheaper to triage. Fewer researchers can submit, which keeps the triage load lighter, though some programs pay higher rewards for private findings. Scope and the reward table decide the overall cost, so compare full-cost forecasts for both.


Yes, with limits. A bounty demands real time and resources, which weighs more on a small team. Start with a vulnerability disclosure policy, test the main applications first, cap the reward pool, keep the scope narrow, and budget triage hours before inviting researchers.


The platform, reward pool, and triage usually sit with security, which runs the program. Fix and retest time lands on engineering. Splitting the forecast that way shows each owner its share and keeps remediation hours from disappearing into sprint work. Some programs have system owners fund their shares.


Expect a spike of reports in the first weeks, often in easy bug classes, then a steadier rate. Judge the return over several quarters by counting valid unique findings, total cost per valid finding including triage hours, and time to fix. A program that finds few valid bugs after pretesting may be doing its job.

Stay updated

Get the latest insights on AI, cybersecurity, and continuous pentesting delivered to your inbox