Three and a half years of monthly Northbeam data, healthy 2023–2025 as the baseline against a struggling 2026. The goal isn't to confirm CAC is up — everyone feels that. It's to find which leading indicator moved first, so next time we see it coming.
In 2026 we are buying more, cheaper attention — click-through rate is +52% vs baseline and CPM is up only +9%. But that attention converts to new customers at half the old rate: new-customer conversion is −47%, new customers per month are −29%, and new customers now make up 29% fewer of our orders. The leak is in conversion and traffic quality, not in the cost of clicks. This is the same pattern your Instagram-traffic and zero-second-session diagnostics surfaced — now confirmed at the portfolio level across years.
If CAC were a cost problem, CPM and CTR would be hurting. They aren't — they improved. The damage is entirely downstream, in whether a click becomes a new customer.
Read it as a funnel: more clicks × far lower new-customer conversion = fewer new customers at a higher cost each. Spend was actually pulled back −15%, so this is not an over-spending problem either.
Every core indicator, healthy-years average vs 2026 average, with the change and whether that change is good or bad for the business.
| Indicator | 2023–25 avg | 2026 avg | Change | Verdict |
|---|
Each panel is one indicator over 42 months. Dashed line = 2023–25 baseline average. Shaded band = 2026. Watch new-customer conversion and new-customer share fall off a cliff while CTR climbs.
Cross-correlation of each indicator's monthly change against CAC's change, at 0–3 months of lead. A strong value at lag ≥1 means that metric tends to move before CAC — an early-warning candidate. Honest result: at monthly grain the signals are modest, and the strongest links are same-month — which is exactly why Phase 2 drops to weekly, ad-level data.
| Indicator | Same month | 1mo lead | 2mo lead | 3mo lead | Strongest |
|---|
Read: New-customer share and new-customer conversion track CAC almost one-for-one in the same month (ρ ≈ −0.66 / −0.48) — they are the live vital signs. CTR leads CAC by ~1 month (ρ = −0.30): a CTR move tends to show up in CAC the following month — the best monthly early-warning we have. Note the sign: when we win cheaper clicks (CTR up) without fixing conversion, CAC follows the conversion, not the click.
CTR is up 52% and it didn't save CAC — because the clicks don't convert. Chasing thumbstop/CTR in isolation is pouring more water into a leaking bucket. Fix the bucket first.
New-customer CVR (−47%) and new-customer share of orders (−29%) are the metrics that actually explain the CAC rise. Put a weekly target on them; they are the live vital signs.
More clicks converting worse points to audience, landing-page, or offer mismatch — not media price. Cross-reference the Sept 2025+ landing-page shifts and the begin-checkout tracking break already on file.
CTR shifts precede CAC by ~a month. But read it correctly: a CTR spike with flat conversion is a warning that CAC will rise next month, not a win.
This macro view proves the leak is conversion-side and that monthly data is too coarse for true early warning. Phase 2 builds the predictive layer you approved: pull the full daily lifecycle of every ad (winners and killed ones), define a winner as sustained scale at/under target CAC, and learn which day-1–3 signals predict it — a "Gem Score" any teammate reads on a brand-new ad, plus scale/kill rules with dollar impact.