GO TO WAR
Military.com — Appendix C Dataset Capi
Source: appendix_c_dataset_capi.md

Appendix C — Dataset / CAPI / Signals

Live API pull, June 4, 2026.

Datasets attached to the account

Three unique datasets (one shown twice via duplicate business associations):

ID Name Created Last Web Fired Last Server Fired Business Status
1536649764252698 MA Military Insurance Dec 1, 2025 Jun 4 14:42 ✅ NEVER (1969 epoch) Military.com (10153481725199270) ACTIVE
587241028099131 Military.com's Pixel Feb 2016 (legacy) Jun 4 14:39 ✅ Mar 2023 (3+ years silent) Military.com (10153481725199270) ACTIVE
332555186604235 Military Insurance May 21, 2024 Jun 4 14:39 ✅ Jun 4 14:34 ✅ 551525685236215 (QW/Delty integration?) ACTIVE

The decode:

Event Match Quality (EMQ) by dataset

Dataset 332555186604235 (the lead-gen workhorse)

Event EMQ Composite Email % Phone % fbc % fbp % zip % Upload Frequency
Purchase 9.1 85.2% 97.1% 87.0% 100% 99.9% hourly
CompleteRegistration 8.3 100% 97.8% 100% (not shown) 97.8% hourly
PageView 6.9 12.7% 11.3% 82.7% 99.2% 80.3% real-time

Key validations:

Dataset 587241028099131 (Military.com's Pixel — content/site)

Event EMQ
PageView not surfaced (no match keys)
SubscribedButtonClick not surfaced
Lead not surfaced

No CAPI events, no EMQ scoring. This dataset is browser-only. Maps to Mo's "Dataset C content pixel CAPI Gateway 0% set up" finding — CONFIRMED. Adding a Signals Gateway / CAPI for this pixel is the +23% cost/result opportunity Mo flagged.

Dataset 1536649764252698 (MA Military Insurance)

Event EMQ
Purchase not surfaced
PageView not surfaced

Browser-only, no EMQ data. Dormant from a server-side perspective.

Event volume (last 7 days, Dataset 332555186604235)

Rough totals across the 7-day window May 28 14:00 – Jun 4 13:59:

Server vs Browser split (across same window): SERVER ~52%, BROWSER ~48%. CAPI is doing roughly equal volume to the Pixel. Healthy dual-channel setup.

Server uplift over browser: ~5–8% per hour on average. Mo's "17.5% more Purchase conversions via CAPI last 7d" was likely from a different reference window or measurement methodology (Meta's UI sometimes shows attribution-window-adjusted lift). The current dataset health is good but the 17.5% number may not be re-derivable today.

The Purchase value malformation issue (the root-cause)

This API cannot validate or refute Mo's "98% of Purchase price data is malformed" claim directly. That signal lives in Meta Events Manager's Diagnostics UI, which surfaces parameter-validation issues that are invisible to the dataset_quality / dataset_stats endpoints (those only check presence, not validation).

What this API DOES confirm: - Purchase events are firing at healthy volume (~755/day) - EMQ is excellent (9.1) - All match keys have high coverage

What this API CANNOT confirm: - Whether the value parameter being sent is a real QW payout amount, or a flat placeholder ($1, $0, etc.) - Whether the value parameter format is well-formed (numeric, decimal-correct, etc.)

Mo's audit conclusion stands and is not refuted: Purchase events fire, EMQ is great, but if the value parameter is flat/malformed, Max Value campaigns optimizing on RETURN_ON_AD_SPEND are training against noise. This is consistent with purchase_roas: "Not available" at every level of the account (Appendix A account-level, Appendix B every ad). If Meta had usable Purchase value data, it would compute ROAS — it doesn't, because the value is unusable.

Mapping to Mo's 5-workstream task list

Workstream Item Live Verification
W1-1 Reconcile Browser Purchase value vs FunnelFlux API confirms Purchase fires at high EMQ; value validity needs Events Manager check (not API-accessible)
W1-2 EMQ optimization CompleteRegistration EMQ 8.3 / Purchase EMQ 9.1 — strong. Marginal room: PageView EMQ 6.9 (low email/phone).
W1-3 fbc/fbclid integrity API: fbc 87–100% present. Truncation (value malformation) is a manual Events Manager validation.
W1-4 PageView match quality EMQ 6.9 — improvable. Email coverage 12.7%, phone 11.3% on PageView. Most-impact fix: capture email/phone at PageView when known (post-login pageview, etc.)
W1-5 (new) Dataset C CAPI Gateway 0% set up CONFIRMED — 587241028099131 "Military.com's Pixel" has zero server activity since Mar 2023. Adding a Signals Gateway is the +23% cost/result lever Mo identified.

The MA Military Insurance dataset puzzle

A dataset created Dec 2025 with server never fired is suspicious. Two hypotheses:

  1. MediaAlpha integration that was set up but never activated. Iron Corp had plans to route through MediaAlpha but the server-side hookup was never completed. If true, this is a quietly-broken integration that should be either activated or removed.
  2. A test/staging dataset left behind. Less concerning but should be archived.

Anthony's "ignore exchange.qmp.ai and MediaAlpha, focus on Delty + Meta" framing earlier suggests MediaAlpha is intentionally not active. This dataset's state is consistent with that.

Open items for Tuesday

  1. Manually open Meta Events Manager → Dataset 332555186604235 → Diagnostics tab to verify the Purchase value-malformation claim (98% per Mo). The API confirms events fire and EMQ is high; only Events Manager Diagnostics shows value validity.
  2. Confirm with dev: is Dataset 587241028099131 (Military.com's Pixel) intended to have CAPI? This is the +23% cost/result lever — if there's no policy reason not to, adding the Signals Gateway is a 24-48 hour engineering task.
  3. Confirm with Anthony/dev: what is the MA Military Insurance dataset for? Activate or archive.
  4. Verify the Dataset 332555186604235 business_id (551525685236215) — confirm this is the QW/Delty integration entity.
← Back