Akka Performance

Window

Account

Daily spend and revenue

Every day since inception. Both series are dollars on one scale.

Meta spend SamCart revenue

Daily blended ROAS

Bars are each day's revenue ÷ spend. The dashed line is the 1.40 target.

Ad sets

Day by day, by ad set

Metric Days

Showing Spend. Today's column is partial — Meta spend is current but SamCart revenue lags by hours, so today reads low on any revenue-based metric.

How these numbers are built, and what to distrust

Revenue = PostHog / SamCart. It catches the upsells and bumps that never post back to RedTrack, so RedTrack reads Akka roughly 35–40% low. Account and campaign revenue here match the PostHog tiles exactly.
Spend = Meta, not PostHog. PostHog's spend is a Meta→warehouse copy that can run a full day behind intraday — it once showed $10 against Meta's real $1,674. Meta is used everywhere so today's row is honest.
Every rate is weighted. ROAS, CPA, AOV, CPM, CTR and CVR are computed from the window's summed numerator and denominator. Averaging daily ROAS would weight a $60 day the same as a $2,000 day.
Windows exclude today except the Today tab, because both spend and sales settle overnight. Lifetime means every settled day through 2026-10-03.
Killed and armed rows are merged in the performance tables. A kill strips the [ON] prefix, so PostHog reports spend on the untagged row and sales on the tagged one. Both spellings fold into one entity there. The names shown do the opposite on purpose — they keep Meta's exact spelling, because which side of that rename an entity is on is what the two groups in the ad-set table are built from.
The state chip needs two signals to agree before it says "no ads". Meta's ad-level effective_status disagrees with the ad set's own: on 09-23 fourteen ad sets came back ACTIVE while every ad under them reported CAMPAIGN_PAUSED — and they were spending that same day. So an ad set is only flagged as serving nothing when it has no active ad and no spend today.
7 ad sets were renamed in Meta after launch. PostHog stores the name as it was when the click happened, Meta returns today's name, so the plain join drops them. They are matched back on their HL#_AT# cell token among the ad sets that spent that day: HL1_AT1, HL1_AT2 United States Broad, HL1_AT3 United States Broad, HL1_AT4 United States Broad, HL2_AT1 United States Broad, HL2_AT2 United States Broad, HL2_AT4 United States Broad.
1 campaign was renamed after launch and is matched back on its date + angle + format tokens: 08.28.26 - Akka - MG1 - MG - Weight - VID - Test - Open - ABO.
35 ad-set names exist in two campaigns at once (both 09.03 campaigns ship identical ad-set names). PostHog joins sales by name and cannot tell them apart, so their revenue is split between the two campaigns in proportion to each one's Meta spend that day. Campaign totals are unaffected — those come from PostHog's own campaign tile.
Unattributed / Direct is $11,561 of lifetime revenue (12%, 67 orders) — but almost all of it is one early outage, not a live problem. From 08-26 to 09-01 100% of orders arrived with no tracking parameters at all; something was fixed around 09-02 and the rate has run about 9% since. Read the share on the row itself, which moves with the window you pick, not this lifetime figure. PostHog is not discarding attribution it has: it already carries a customer's campaign across their other orders the same day, and only buckets an order here when utm_campaign is missing, empty, or an unresolved {{…}} macro. On every recent day but one the campaign and ad-set tiles flag the exact same orders, so the whole parameter set is absent rather than one macro being malformed.
The ad-set table starts 2026-08-26. PostHog's ad-set tile is capped at 200 rows, about a week. This dashboard keeps its own append-only copy, so from now on nothing scrolls off — but ad-set spend before 2026-08-26 is Meta-only and carries no revenue. Campaign rows reach back to 2026-08-26; Meta spend to 2026-08-24.
IC underfires on Akka. Checkout sits on a separate domain, so the measured ratio is about 1.87 IC per sale, not the house 3. That is why the IC CPA target is $63 — one third of CPA would be a false-kill machine here.