AdRoastAdRoast/ TeardownsRoast your ads free
Ad teardown · AppSec / SAST

A page that sells accuracy.
And can’t keep its own numbers straight.

The Snyk comparison page has the right strategy and fights its own execution. The ad leads with a screenshotable stat, 17.5 hours vs 26 minutes, plus an elite logo wall (Dropbox, Snowflake, Figma, GitLab) and a real third-party benchmark. Then the landing page says 2.5 hours, not 26 minutes, so a page selling accuracy cannot keep its own numbers straight. And a $100 gift card pulls gift-card hunters, not AppSec leaders.

Semgrep
Semgrep
semgrep.dev · LinkedIn
6/10Ad-to-page
score

The teardown, on LinkedIn

The ad, torn apart
LinkedIn post
Open on LinkedIn →

What the teardown found · tap to open each one

What’s happening

The ad leads with a number you could screenshot, 17.5 hours vs 26 minutes, backs it with recognizable logos (Dropbox, Snowflake, Figma, GitLab), and cites a real third-party benchmark. For a competitive-displacement play against Snyk, that is the right opening.

What’s happening

The ad says 26 minutes. The landing page says 2.5 hours. A page whose entire argument is accuracy cannot keep its own headline number straight between the ad and the destination, and a skeptical AppSec buyer will notice.

The fix

Make every number match. Pick the real figure and use it identically in the ad, the page and the benchmark, so the accuracy claim survives the click.

What’s happening

The incentive attracts gift-card hunters, not AppSec leaders. It inflates form fills with people who will never buy and buries the real buyers in the noise.

The fix

Drop the gift card, or gate it behind a qualifying step, so the page measures intent instead of rewarding tourists.

Questions, answered

The ad is strong: a screenshotable stat (17.5 hours vs 26 minutes), an elite logo wall and a real third-party benchmark, which is exactly the right opening for a comparison play against Snyk. It scores 6 because the page it lands on undercuts the ad’s own argument.
On the comparison page. The ad says 26 minutes; the page says 2.5 hours. A page selling accuracy cannot keep its own number straight between the ad and the destination, and a $100 gift card pulls gift-card hunters instead of AppSec leaders.
Make the numbers match everywhere, pick the real figure and use it identically in the ad, the page and the benchmark, and drop or gate the gift card so the page attracts buyers, not tourists.
Put your ad and your landing page next to each other and check the headline number: is it identical? If a page built on a metric cites a different metric one click later, you are arguing against yourself. AdRoast scores this automatically, and the first roast is free.

This is Marina Kogan’s original teardown of Semgrep’s live ad. See the full infographic on LinkedIn →

More teardowns

Cycode
Four analysts sold the click. The page hid them.
Read the teardown →
Apiiro
A sharp hook that dies on a blog post.
Read the teardown →
Endor Labs
The best AppSec story of the quarter, buried.
Read the teardown →

Your ads, roasted the same way

Pull your live ads, score each against your buyer, see where the click leaks.

Roast your ads free →