A buyer's security questionnaire arrives, one line asks for a penetration test report, and you start collecting quotes. I'm Denys, a Forward Deployed Engineer. Written with AI assistance, checked by me.
Before the first scoping call I would prepare one page. A tester charges for the days the work takes, and a page that answers their first ten questions shortens the call. It also gives you something to hold two quotes against.
Why the same app gets such different quotes
On 6 November 2026 I read the cost pages of five sellers and took each range as written. Most start between $5,000 and $15,000 and end at $30,000 or $35,000, and one says anything listed below $4,000 is probably not a real test. I did not ask for quotes, and none of the sellers is mine. The ranges are in my write-up of what a penetration test costs in 2026.
The spread comes from scope. Who does the testing, how many roles and endpoints count, what the price includes and how it is billed all move the number. The pack below fixes the scope part before anyone names a price.
What goes on the page
- Roles. Every kind of user in your app and what each should be allowed to see. In a web app, the number of roles and of API endpoints is the scope.
- Endpoints. Every route the app exposes, with the public ones marked. Export it from your router instead of writing it from memory.
- Two test accounts in two different companies. Without them a tester cannot try tenant isolation, which is where small SaaS apps tend to fail first. Ask each seller whether they get such accounts. If they don't, you are paying for a test that skips the most likely hole.
- Your own cross-company result. Log in as user A in company 1 and attack company 2 by every path you can think of. Then read the database to see what changed. Reading the responses is not enough. In my own test on a deliberately broken schema, the four read checks most people try came back clean while 6 of 14 requests at a second company leaked.
- Where each key lives. A secret key skips every access rule, so one in the browser undoes the rest. List each key and where it is stored.
- Environments. Which ones are in scope. Testing production and testing staging are different jobs with different risks.
Put the date next to every result. A buyer can read a list of paths, what each returned and when. A tester can skip what you already covered.
What the page does not do
It does not replace a test when a buyer, an auditor or a rule asks for one. I did not verify what each framework or each buyer accepts, so ask them before you buy. The idea that a prepared app takes a tester less time is my own reasoning, and none of the sellers said it.
It also does not tell you if a cheap quote is good. For that, put the same questions to every seller. Is a person or an AI leading the test, and who reviews the findings. Are retests included, and how many rounds. Is the platform a separate line on top of the test price. Will my buyer accept the report you hand me.
If the answers differ a lot between two quotes, the price gap is probably explained there and not in the hourly rate.
A short test you can run today
Make two accounts in two different companies. From the first, request every record id you can find that belongs to the second, through the list pages, the detail pages, any export and any file link. Write down each request and the answer. Then open the database and check that no row changed. It takes an afternoon on a small app, and whatever you find, you have found it before a buyer's tester did.

