Bottom line. Blended customer-acquisition cost rose ~44% from April to June (~$94 → ~$135) on flat ad spend, and June new-customers/day are down ~34% year-over-year. The genuine anomaly is new-customer acquisition — not overall conversion, and not a traffic surge.
⚠ Root cause found: ~1,003,991 zero-duration ("instant-bounce") Instagram sessions (+926%) — bot / invalid traffic — are flooding the funnel. They load and leave in 0 seconds and never convert, which is what craters Instagram's conversion rate and drags down new customers & CAC. This is the engine beneath the numbers below. Full breakdown, tactics, diagnostic queries, and the fix are in the next section.
What's true about Instagram: traffic is up — IG-referred sessions run well above 2025 across the half (+84% in April, the most of any channel) — and its on-site conversion collapsed, from ~2.2% to under 1% (−50% YoY). More IG visitors converting far worse. (Two myths to retire: overall site conversion is not down — blended order-CVR is flat YoY, the spring→summer dip is seasonal; and we are not spending more on IG — Meta spend is flat-to-down.)
What actually broke — and where. Because total order-conversion held YoY while new customers collapsed, this is an acquisition problem, not a checkout break (a site/checkout regression would have dragged all orders down — it didn't). It's concentrated in paid prospecting, above all Instagram — more, cheaper, Reels-heavy IG traffic converting at half its old rate. Returning-customer demand (Email, SMS, Direct) held up, which is why blended conversion still looks normal. Two levers: (a) Instagram / paid-prospecting quality, and (b) a likely change in the new-customer / first-order offer vs last June.
Instagram is the worst-hit channel, and there's real waste to cut. IG conversion more than halved year-over-year and fell under 1%. Two things stack: IG inherits the site-wide drop, and IG-specific quality worsened — Meta is delivering cheaper, Reels-heavy inventory (Reels now 54% of IG impressions, our lowest-converting placement); 71% of IG impressions come from one broad "sandbox" campaign; we cut our best-converting scaled campaign; and a batch of influencer/partner videos convert at 0.4–1.2%. Net: more, cheaper IG clicks — far fewer sales.
Two measurement cautions. (1) Meta's reported CAC looks better than reality because attribution moved to a 7-day-click window — judge on blended/Northbeam CAC, not Meta's dashboard. (2) The account-level "ad fatigue" signal (rising frequency, falling CTR) is a Facebook effect, not Instagram — don't chase it on the wrong platform.
Validated against Meta Ads API, Shopify (funnel source of truth) and the Daily Stand sheet (blended-CAC source of truth); seasonality controlled year-over-year. Caveat: a site tracking gap Jun 12–17 understates the last 1–2 weeks; trends hold.
| Zero-duration IG sessions | ~1,003,991 (+926%) |
| Device | 99.9% mobile |
| Geography | ~98% US (proxied) |
| Desktop IG (real users) CVR | 4–6% |
| Mobile IG CVR | ~1.5% |
| Conversion of zero-dur sessions | ≈ 0% |
99.9% mobile + 98% US + instant-bounce + scaling with Reels/Advantage+ = invalid traffic (bot / incentivized / accidental-tap), not a foreign desktop click-farm. Note the step-up around Dec 2025–Jan 2026 — pinpoint what changed then.
| Vector | How it works · signal |
|---|---|
| Ad-delivery fraud | Junk Audience-Network / Reels inventory + mobile emulator farms on US residential proxies tap ads and bounce. Broad Advantage+ leans into this cheap inventory. |
| Affiliate / influencer click fraud | Partners paid per click/traffic bot their IG links. Ties to our 0.4–1.2% CVR partner ads (partner-SHA, Sophia, hindzsight). |
| Competitor sabotage | A rival bots your ads to drain budget and poison your conversion signal so Meta optimizes toward junk. |
| Referrer spoofing | Bots hit the store directly with a forged instagram referrer — pollutes analytics, may not cost ad spend. |
| Accidental Reels taps | Full-screen Reels fat-finger taps open & instantly close = zero-duration. Not malicious, same damage; scales with Reels. |
The vicious cycle: these convert at ~0%, and Meta optimizes against your pixel — so the algorithm learns to chase more of the same cheap junk. 7-day-click attribution + broad Advantage+ accelerate it.
-- 1 · Is Shopify flagging it as bot? FROM sessions SHOW sessions, online_store_visitors WHERE referrer_name = 'instagram' AND session_duration <= 0 GROUP BY human_or_bot_session SINCE 2026-01-01 UNTIL 2026-06-25 -- 2 · WHICH ads / partners it concentrates in FROM sessions SHOW sessions WHERE referrer_name = 'instagram' AND session_duration <= 0 GROUP BY utm_campaign, utm_content SINCE 2026-01-01 UNTIL 2026-06-25 ORDER BY sessions DESC LIMIT 50 -- 3 · Device / OS / browser fingerprint (bots cluster) FROM sessions SHOW sessions WHERE referrer_name = 'instagram' AND session_duration <= 0 GROUP BY session_device_os, session_device_browser, session_device_browser_version SINCE 2026-01-01 UNTIL 2026-06-25 ORDER BY sessions DESC LIMIT 50 -- 4 · Geo concentration + exact referrer FROM sessions SHOW sessions WHERE referrer_name = 'instagram' AND session_duration <= 0 GROUP BY session_city, referrer_site SINCE 2026-01-01 UNTIL 2026-06-25 ORDER BY sessions DESC LIMIT 50 -- 5 · Sanity: cart-add & conversion should be ~0 FROM sessions SHOW sessions, added_to_cart_rate, conversion_rate WHERE referrer_name = 'instagram' AND session_duration <= 0 SINCE 2026-01-01 UNTIL 2026-06-25
(For reference, the within-2026 April→June moves: CAC +44%, new customers −36% — but the like-for-like YoY comparison above is what isolates the real anomaly from normal seasonality.)
| Window | 2025 | 2026 | YoY |
|---|---|---|---|
| April order-CVR | 2.31% | 2.28% | −1% |
| June order-CVR | 1.84% | 1.90% | +3% |
| Instagram CVR (Jun) | 1.96% | 0.97% | −50% |
Total order-conversion is normal YoY — a checkout/site break would drag this down. It didn't. Instagram is the exception.
| Window | NC/day | Blended CAC |
|---|---|---|
| April 2025 | 550 | $88 |
| June 2025 | 528 | $98 (+11%) |
| April 2026 | 511 | $94 |
| June 2026 | 348 | $135 (+44%) |
2025: CAC +11%, volume held (−7% YoY in June). 2026: CAC +39% YoY, new customers −34% YoY. The break is acquisition, not conversion.
On attribution (important): "Instagram" here = referring-platform Instagram in Shopify — the cut that splits IG from Facebook — not the default channel grouping that buckets Meta as "facebook / paid social." It includes organic + paid IG. The on-site Meta Facebook↔Instagram split is imperfect (we cross-checked: Shopify attributes ~47% of FB+IG sessions to IG vs Meta's own ~25% paid delivery), so treat the split as directional; Meta's publisher-platform data is the ground truth for paid placement, and that's what the placement/spend figures use.
| Campaign | IG share | CVR now | CVR Apr | CPO |
|---|
| Ad | Impr | CVR | CPO |
|---|
| Hypothesis | Likelihood | Evidence |
|---|---|---|
| Invalid / bot traffic flooding Instagram | Confirmed | ~1M zero-duration IG sessions (+926%), 99.9% mobile, 98% US, ~0% conversion. Directly craters IG CVR and new customers — the root cause beneath the symptoms below. See the invalid-traffic section. |
| Instagram / paid-prospecting conversion collapse | High (symptom) | IG CVR −50% YoY (June, now <1%) — largely caused by the invalid traffic above + Reels/broad delivery. IG is a core prospecting channel, so it hits new customers directly. |
| New-customer / first-order offer change vs 2025 | High | Order-CVR is flat YoY but new customers −34% — consistent with a first-order/welcome offer last June that this June lacks. Untested in-data — confirm first. |
| Broad/Advantage+ floods cheap IG Reels with low-intent traffic | High | Sandbox 44%→66% of Meta spend; Reels 49%→54% of IG; IG CPM holds while FB CPM falls. |
| IG media mix-shift — cut high-CVR scaler, added low-CVR tests + influencer ads | High | Verified at campaign/ad level: partner-SHA ASC cut −64%; new tests at 0.96–1.5%; influencer videos 0.4–1.2%. |
| New AppLovin channel adding low-intent sessions to the blend | Medium | Appears as a new band in 2026 share-of-traffic; magnitude/CVR not yet isolated — pull its own CAC. |
| Facebook saturation (reach down, frequency up) | Medium (FB only) | FB frequency 2.91→3.88, CTR −18%. Real, but Facebook — not Instagram (IG frequency fell 2.27→2.03). |
| Site-wide checkout / conversion regression | Low | Argued against: blended order-CVR is flat YoY — a checkout break would drag all orders down. (GA4 checkout events did break ~May 18, but that's a tracking issue, not necessarily a real funnel break.) |
| "Meta reallocated budget to Instagram" | Rejected | IG spend share fell 42%→37%; IG spend −24%/day. Dollars moved away from IG. |