Hyrax — end-to-end marketing funnel

Spend → traffic → funnel → signups, by channel, source, country, device · Data pulled live from Google Ads, Vercel, PostHog, Reddit Ads, X Ads, GA4 · Aug 7, 2026

Dates: Source:

–

Google Ads spend in range (Reddit/X in §4)

–

site visitors in range/source (Vercel)

verified external signups since Aug 5

internal/test signups (team, filtered out)

1 · Inputs: what goes in the top

Google Ads spend per day split search vs Demand Gen (bars, left axis) and blog posts published per day (line, right axis). Reddit runs ~$110-140/day and X ~$40-50/day; their APIs report window totals, shown in §4. Responds to the date controls.

2 · The funnel — ads to first job, end to end

Starts where money is spent. Impressions and clicks are platform-reported; site visits are Vercel by UTM source; app stages are PostHog, external people only (8 team/test accounts removed). The funnel uses the three windows where every ad platform reports verified totals. Counts under 10 are directional.

Window: Channel:

3 · Inside the app

PostHog app events since Aug 5, split external vs internal (team). This is where the funnel hands off to product.

Event	External	Internal (team)	Read
Reached the app	4 people	22 people	4 real prospects crossed over in 3 days
Created an account	0	8	the acquisition funnel ends here today
Onboarding: connect-org screen shown	6 views (1 person)	19 views (8 people)	one external user hit the connect screen 6 times
Onboarding: sample repo opened	2 (1 person)	13 (8 people)	sample-repo path is being exercised in testing
GitHub org connected	1	3	the external connect is an existing customer
Jobs submitted / completed	2 / 2 (1 person)	0	an existing customer ran two jobs to completion
Plan changed	1	0	one external plan change in the window
Payment failed	1	0	a paying external user hit a payment failure; follow up
Rage clicks	2	5	5 team rage clicks during onboarding testing is a UX signal in itself

4 · Targets and kill rules

The bar this funnel is judged against, set Aug 7. Benchmark sources: GrowthSpree 2026 ($60M+ managed B2B SaaS spend), PipeRocket 2026 (dev-tools cut), Affect Group 2026 (Demand Gen), AdBacklog 2026 (Reddit), WordStream 2026, WebFX (X).

Metric	Today	Target	Rule
Site visitors / day	~600	2,000	scale winners only; no junk reach buys
Verified external signups / day	0	100 (leadership goal)	first measurable week starts Monday Aug 10
Visitor → signup rate	0%	2.5-4% (PLG benchmark)	pricing v2 + signup CTA + tagged landings shipped Aug 7
Cost per verified signup, per channel	not measurable until UTM handoff ships	<$150 (dev-tools CPL benchmark: $855 avg, $180-350 typical B2B)	any channel above $150 after one full week of attribution gets cut
Text-ad CTR floor	X at 0.2%	0.5%	X verdict Aug 12 with Reddit text; below floor = cut
Platform-pixel conversions	Reddit claims 111 in 3 days	—	never steer budget; PostHog-verified numbers only

5 · What's failing and the fix

Every red number on this page, its root cause, the fix, and where the fix stands. Dates are commitments already made, not aspirations.

Failing	vs benchmark	Root cause	Fix	Status
0 external signups	2.5-4% of visitors expected	Homepage CTA pointed at sign-in; 75% of clicks were mobile; ads landed on untagged pages; old pricing	Signup CTA fixed + mobile excluded + ads moved to /code-review + pricing v2 and onboarding rebuild	ALL SHIPPED Aug 5-7 · first measurable week starts Mon Aug 10
No channel attribution for signups	—	PostHog runs in the app only; UTMs die at the site→app boundary	Marketing-site snippet + cross-domain UTM handoff + identify-by-email at signup (GTM5-66)	v0 prompt built, awaiting paste · due this week
Internal/test filter relies on geography	—	App does not identify users by email at signup	Same GTM5-66 ship adds identify; filter switches from city-cluster to email domain	bundled with the row above
X CTR 0.17-0.21%	1-3% benchmark, 0.5% kill floor	Broad targeting, weak creative resonance; 226K impressions for 381 clicks in 3 days	Tagline reframe test running at $50/day; hard verdict with Reddit text	VERDICT Aug 12 · below floor = cut
Reddit clicks land untagged on the homepage	—	Reddit API partner tier blocks destination edits programmatically	Manual edit in Reddit Ads Manager: destination → /code-review, desktop-only, budget to ~$150/day	PENDING · 2-minute manual change
Reddit pixel claims 111 conversions in 3 days, reality 0	$50-150 per qualified B2B signup benchmark	Pixel counts view-through and page events, not accounts	Standing rule shipped to §4: platform pixels never steer budget; PostHog-verified only	RULE IN EFFECT Aug 7
hyrax.dev/signup returns 404	—	No redirect on the marketing site; buttons link to app.hyrax.dev/signup directly	One-line redirect via v0 prompt	PENDING · found Aug 7
GA4 dead since Jul 28	—	Site redeploy shipped without the tag	PostHog becomes the single behavioral source of truth once site snippet ships; GA4 stays for Google Ads conversion import only	decision folded into GTM5-66
External payment failure (1 user)	—	Unknown; single event in PostHog Aug 5-7	Identify the account and follow up directly	NEW · surfaced by this report
2 DG ads "approved (limited)" on YouTube	—	YouTube content requirements flag on 2 of 5 videos	Aug 7 URL change re-triggered full review of all 5 ads	re-review in progress

5b · Campaign verdicts by channel

Every campaign that spent money in the last 21 days, with its result and its status decision. Windows stated per row.

Channel	Campaign	Spend / result (window)	Verdict
Google Search	CodeReview_NonBrand	$73/day recent · 25 start-free clicks at $7.81 (Jul 30–Aug 6) · tracked signup Aug 6 · CPC $8.15 vs $8.50-14 dev-tools benchmark	WINNER — budget raised to $150/day Aug 7
Google Search	DevSecOps_NonBrand	$75-118/day · tracked signup Aug 7	WINNER — $120/day + tCPA $35 set Aug 7
Google Search	Brand-Defense	$10-24/day · protects the brand query	KEEP — $10/day
Google Search	CodeSecurity_NonBrand	$34/day avg · 0 signups (Aug 1–7)	PAUSED Aug 7 — winners-only consolidation
Google Search	CompetitorSwitchers_NonBrand	$46/day avg · 0 signups (Aug 1–7)	PAUSED Aug 7
Google Search	SonarQubeSwitchers_NonBrand	$50/day avg · 0 signups (Aug 1–7)	PAUSED Aug 7
Google DG	Developer Review Aug2026 (old)	$388 on Aug 5 alone, 1,177 clicks — 2-3 AM international mobile YouTube	PAUSED Aug 5 — replaced, never re-enable
Google DG	Developer Review US (new)	735 clicks / $118 today at $0.16 vs $0.40-1.20 optimal benchmark · CTR 3.3% vs 1.5-2.5% optimal · desktop-only, US, daytime, lands on /code-review	SCALED — $200/day, Max Clicks
Reddit	TextOnly_Traffic	$647 / 2,732 clicks at $0.24 vs $0.60-2.50 benchmark · CTR 3.4% vs 0.2-0.5% benchmark (21 days)	WINNER on clicks — verdict on signups Aug 12
Reddit	Q3_Connect (video)	$1,739 / 21 days · 289 pixel conversions at $6.02, unverified	PAUSED — spend concentrated on text + ConvAds
Reddit	3× ConvAd Signups (VibeCoding / Security / ReviewFatigue)	$182 / 21 days · 210 pixel conversions at $0.69-1.08, unverified	ACTIVE — verify against PostHog when UTM handoff ships
X	Text Ads GitHub App v2	$360 / 391 clicks · 141K impressions Aug 5 alone · 0 signups	STOPPED Aug 6
X	Reframe Text Ads Aug2026	$42 yesterday / 84 clicks · CTR 0.21% vs 1-3% benchmark, below the 0.5% kill floor · 0 signups	TEST at $50/day — Tristan's tagline test, runs to verdict
LinkedIn	All campaigns (July relaunch)	$1,599 Jul 21-27 · 2 real leads	CUT Jul 27 — only Tyler thought-leader boosts remain ($60+$60/day nominal)
Meta	All campaigns	—	CUT Aug 1 (GTM reset)

6 · Spend: how much, over time, and the split

Left: total spend per week by channel. Right: percent distribution for the last full 7 days (Google Jul 31–Aug 6 computed from daily data; Reddit and X Jul 30–Aug 6 as reported by their APIs).

Google spend by week (computed from daily API data): $984 → $1,338 → $1,803 in the current 4-day week, with Demand Gen going from zero to over half the mix after the Aug 4 launch. Add LinkedIn's final week ($1,599, Jul 21-27, cut Jul 27) and Reddit's $2,568 over the trailing 21 days for the full picture. Last 7 days across all live channels: $2,906, split Google 67% / Reddit 26% / X 7%. The $600/day cap took effect Aug 6; Aug 6 ran $761 because two budget changes crossed mid-day, and Aug 7 is pacing inside the cap.

7 · Daily visitors, annotated with every change

Vercel unique visitors per day for the selected source; the selected date range is shaded. Aug 7 is partial.

July's 500-800/day baseline was paid social: mobile, short sessions, high bounce. The Aug 1 reset cut it on purpose and traffic fell to ~120/day. The Aug 5 spike (1,235) was the old Demand Gen campaign spending overnight on international mobile YouTube; it was found that morning and the campaign was rebuilt the same day. Aug 6 onward is the clean baseline: desktop, US, business hours, tagged landing pages.

8 · Traffic by source

Vercel unique visitors by UTM source within the selected dates.

9 · Traffic by country

Vercel unique visitors by country within the selected dates. Product scope is US-only at launch; all Google campaigns are now US-only.

10 · Traffic by device

Vercel unique visitors by device within the selected dates.

All 8 raw signups came from desktop; zero from mobile. GA4's last full baseline (Jul 8-27, tag died Jul 28) measured mobile paid sessions at 30-52 seconds with 80%+ bounce, versus 75-459 seconds on desktop. Mobile is excluded on every non-brand Google campaign as of Aug 7. This split shifts toward desktop from today.

11 · What changed, in order

Date	Change	Effect
Jul 27	LinkedIn campaigns cut ($1,599 for 2 real leads)	Paid social spend to zero
Aug 1	GTM reset: Meta cut, budget concentrated on Google/Reddit/X	Traffic -85%, junk removed
Aug 4	4 Demand Gen YouTube videos live	New reach channel
Aug 5	3 AM anomaly diagnosed and campaign rebuilt US-only, 7 AM-11 PM ET; PostHog live in app	Overnight junk eliminated; signups measurable
Aug 6	Maximize Clicks bidding; $600/day all-channel cap	DG cost per click $0.91 → $0.16
Aug 7	Mobile excluded on all non-brand; ads land on tagged /code-review; homepage CTA routed to signup (was signin); 3 losing search campaigns paused; winners' budgets raised	Every Google dollar on desktop, US, business hours, converting pages