Feb 2 – Apr 19, 2026
Findings
Observations across Weeks 9–19. CPA throughout is signup-page visits, not confirmed new orgs — read directionally.
Channels we added or changed that punched above their weight
- Carbon Ads — Added W10. Became a top-3 efficiency channel almost immediately — CPA has hovered $6–$15 most weeks, with multiple 70–130 conversion weeks on sub-$1k spend. Note: a chunk of early volume was international AV traffic that inflated signup-page visits; we broke that out to a separate UTM in March, so post-W13 numbers are the cleaner read.
- SourceForge — Added W12. Quietly compounding — 14 → 35 → 37 → 47 → 280 conversions as we leaned in. CPA regularly under $15, dipped to $2.39 in W18. Clear scale signal.
- ClaudecodeMarketplace — Brand-new (W18). First two weeks: ~$250/wk → 20 and 15 conversions. Very early but CPA ~$12–$17 is on par with our best channels. Worth expanding the test.
Channels that got worse — and what happened when we moved spend
- MDN — Eroded steadily. $645 → $857 → $669 → $386 → $0. Conversions fell in lockstep (22 → 11 → 15 → 9 → 0). Cutting was the right call — efficiency was declining even before we pulled it.
- Product Hunt — Scaled hard: $1.5k → $2.8k → $4.4k → $7.8k. Conversions grew (19 → 66 → 85 → 104) but CPA worsened $41 → $43 → $51 → $75. Returns diminished — we are paying meaningfully more per conversion at the top of the curve.
- Reddit — Mostly flat spend ($1.0k–$2.4k) but CPA improved — $40 early to $12–$23 range after W13. The creative/targeting changes are holding.
Email placements — brand, not conversion
- ProductHunt Frontier Newsletter — $3,572 twice for effectively 0 tracked conversions. Treat as a brand/awareness buy — do not judge on CPA.
- Ruby Weekly — Similar pattern — low-volume, niche audience. Modest visits, few conversions. Value is reaching the Ruby community, not driving signup-page clicks.
- Indie Hackers — Newsletter/community placement running ~$1k/wk for 0–3 tracked conversions (CPA $181–$333). Treat as brand exposure to the indie dev audience, not a conversion channel.
- SED Podcast — One-off $1k buy with negligible tracked traffic. Brand placement.
Recommend reporting email/newsletter spend separately from performance channels so blended CPA on the rest of the mix stays honest.
Conversions here = visits to the signup page, not confirmed new orgs. A channel with high signup-page CPA may still convert well on the backend (and vice versa). Use this dashboard for directional read on channel efficiency; pair with downstream signup data before reallocating material budget.
Paid signup visits vs. new orgs created
What our ads drove to the signup page (paid only) against every new org AppSignal logged (paid + organic + direct). Weeks 9–19.
- W933 paid visits / 74 orgs (0.45×). Paid had barely ramped — only Ruby Weekly, Ethical, and SED were live. The orgs AppSignal logged this week were almost entirely organic/direct. Treat W9 as the pre-campaign baseline, not a conversion signal.
- W15276 paid visits / 58 orgs (4.8× — highest ratio). New orgs hit the low of the series despite healthy paid volume. W14 had a Carbon Ads spike (134 signup visits on $1.7k) that likely pulled forward the high-intent pool; W15 also saw Product Hunt spend collapse from $4.7k → $426. Read as a pull-forward effect plus a PH pullback, not a conversion failure.
- W18613 paid visits / 89 orgs (6.9× — biggest divergence). The paid-visit spike is real but low-quality: SourceForge drove 280 visits at $0.75 CPC with 38s time-on-site, and Product Hunt ran $7.7k with Frontier Newsletter in the mix. Neither cohort is a strong signup-completer, and the lag did not show up in W19 (orgs fell to 68). Best read: traffic-quality dilution, not timing lag.
- W19253 paid visits / 68 orgs. If W18's surge were going to lag-convert, we'd expect an org bump here — we didn't get one. Reinforces that W18 was volume-only, not deferred conversion. Also partial data (week ended 4/19).
Cross-week lag check: W13→W14 (265→358 visits) did not lift W14 orgs above W13; W17→W18 (304→613) did not lift W18 orgs above W17; W18→W19 dropped. Paid-visit spikes are not consistently predictive of next-week org volume in this window.
Reading the conversion picture
Across Weeks 9–19 we drove 2,979 paid visits to the signup page. In the same window, AppSignal logged 912 new orgs — but that count includes organic and direct signups, not just paid-sourced ones. This is not a conversion rate. Without backend attribution we can't say what share of those orgs came from paid; we can only compare the two volumes side-by-side.
Weekly new-org volume has been relatively steady (58–104) even when paid signup visits swing hard (33 in W9, up to 358 by W14, 613 in W18). That tells us organic traffic is doing meaningful work on the baseline, and that big spikes in paid visits — especially W18's SourceForge jump (280 conversions on $668 spend) — did not translate proportionally into new orgs. In other words: more signup-page clicks aren't linearly minting more orgs, which reinforces that paid CPA is directional, not predictive.
What we can say confidently:
- Paid is consistently producing signup-page volume that's several multiples of weekly new-org count — so the ceiling on paid's contribution is high, the conversion step (signup page → created org) is where the leakage is.
- Channels with low signup-visit-to-org conversion likely exist but we can't identify them without backend attribution. A paid-vs-organic attribution breakout of organization.create events is the next step.
- Until we have that, treat paid signup CPA as a traffic-quality proxy — useful for ranking channels relative to each other, not for forecasting org volume.
Reddit — ad group breakdown
Weekly signup-page visits by Reddit ad group, plus per-group spend efficiency. Weeks 9–19.
| Ad group | Spend | Clicks | CPC | Signup visits | CPA | Trend (W9–13 → W15–19) |
|---|---|---|---|---|---|---|
| General Dev Communities (R1.3A) | $5,289 | 21,029 | $0.25 | 361 | $15 | 110 → 213 |
| Ruby Communities (R1.1A) | $2,632 | 7,572 | $0.35 | 176 | $15 | 73 → 86 |
| Ruby Keywords (R1.2A) | $2,640 | 8,490 | $0.31 | 160 | $17 | 59 → 72 |
| Sentry Keywords (S1.1A+B) | $1,572 | 3,149 | $0.50 | 41 | $38 | 25 → 12 |
| SaaS Communities (S1.2A) | $1,844 | 4,918 | $0.37 | 48 | $38 | 8 → 36 |
| Retargeting (ART.1–6) | $1,023 | 750 | $1.36 | 4 | $256 | 2 → 1 |
Reddit read & recommendations
Reddit is ~$15,000 spent across six ad groups for 790 signup-page visits (blended ~$19 CPA). Efficiency is bimodal: three groups are in the $14–$17 range, two more than double that, and retargeting is a firehose pointed at the ground.
- Scale: General Dev Communities (R1.3A). Best CPA at ~$14.65, most consistent weekly volume (20–65 visits), and the curve trended up into W17–W18. This is the horse — push more budget here.
- Hold: Ruby Communities (R1.1A) + Ruby Keywords (R1.2A). Both sit at ~$15–$17 CPA and together produce ~40% of visits. Ruby Communities peaked W13 (35 visits) and has softened; Ruby Keywords is steady. Keep running at current levels — solid efficient tail.
- Question: Sentry Keywords (S1.1A+B). CPA ~$38 and trending toward zero (9 → 0 visits across the window). Competitor keywords aren't converting — rework creative or pause and redirect spend into R1.3A.
- Question: SaaS Communities (S1.2A). CPA ~$38 but volume actually grew late (peaked W16 at 15 visits). Worth one more cycle of creative iteration before cutting — if it holds late-window gains, it may be a second scaler.
- Kill: Retargeting (ART.1–6). ~$1,023 spent for 4 signup-page visits (CPA ~$256). Audience is too small or creative is wrong. Pause and reclaim budget.
Proposed reallocation: pull retargeting (~$1k) and Sentry Keywords (~$1.5k) and reinvest in R1.3A General Dev. That would push it from ~35% to ~50%+ of Reddit spend while we test fresh creative on the S-variants.
$60k allocation — May / June / July
~80% into evergreen channels we know work, ~20% into new-channel exploration to identify opportunities before evergreen fatigues. $20k/month for three months.
We expect the upcoming website redesign to drive two motions: a redesigned creative strategy and a new conversion path with a different visit → signup conversion rate. We will emphasize new visuals on Reddit and test new messaging on niche networks like Carbon, Ethical, and Product Hunt.
| Channel | Monthly | Quarter | Share | Rationale |
|---|---|---|---|---|
| $5,000 | $15,000 | 25% | Scale R1.3A General Dev; hold Ruby variants; cut Sentry + Retargeting. | |
| Product Hunt | $3,000 | $9,000 | 15% | Cap — diminishing returns above $4k/wk. |
| SourceForge | $2,500 | $7,500 | 13% | Scale signal is clear; CPA regularly sub-$15. |
| Carbon Ads | $1,500 | $4,500 | 8% | Efficient; hold at current weekly cadence. |
| Ethical Ads | $1,500 | $4,500 | 8% | Steady dev-focused volume at reasonable CPA. |
| ClaudecodeMarketplace | $1,000 | $3,000 | 5% | Expand the W18 test — early CPA on par with best channels. |
| Email / brand (Ruby Weekly) | $1,500 | $4,500 | 8% | Brand floor in the Ruby community; report separately from performance. |
| New channels (to be identified) | $4,000 | $12,000 | 20% | Discovery budget to hedge against evergreen fatigue. |
| Total | $20,000 | $60,000 | 100% |