Forensic Analysis

February Conversion
Data Anomaly

A day-by-day investigation of conversion data in February 2026 revealing a batch data import that inflated reported ROAS from 273% to 351%.

01

The Signal

Something happened Feb 9–13

vega Sales conversions (confirmed Shopify purchases) jumped from a normal ~190/day to over 2,600/day — while ad spend stayed flat. This is physically impossible from advertising alone.

Feb 1

198

$8,567

Feb 2

233

$7,291

Feb 3

195

$7,753

Feb 4

170

$5,654

Feb 5

212

$5,768

Feb 6

143

$6,359

Feb 7

199

$5,854

Feb 8

234

$6,451

Feb 9

871

$5,962

Feb 10

1,790

$6,723

Feb 11

2,521

$8,009

Feb 12

2,664

$8,098

Feb 13

918

$7,553

Feb 14

207

$6,853

Red bars = vega Sales conversions. Right column = daily ad spend. Notice how spend barely changed ($5.9K–$8.1K range) while conversions exploded 786%. Real ad purchases don't behave this way.

02

The Proof

Spend +8%. Conversions +786%.

Comparing the 8 normal days (Feb 1–8) to the 5 anomaly days (Feb 9–13):

Normal Days (Feb 1–8)

198

avg conversions / day

Anomaly Days (Feb 9–13)

1,753

avg conversions / day

Normal Spend / Day

$6,706

average daily ad spend

Anomaly Spend / Day

$7,269

only 8% higher

Normal Conv / $1K Spend

29.6

conversions per $1K

Anomaly Conv / $1K Spend

241.1

8.1x higher efficiency — impossible from ads

The math doesn't work. If these were real ad-driven purchases, $7,269 in ad spend would need to generate 1,753 Shopify transactions at $41.48 average order value. That's a $4.15 cost per purchase — 12x better than any other day in the account's history.

03

Corroborating Evidence

Two conversion actions spiked simultaneously

Both vega Sales (one-time purchases) and Purchase–Subscription (recurring) experienced identical spike patterns on the exact same 5 days. This confirms a system-level data event, not organic sales.

Date	vega Sales Conv	vs Normal	Subscription all_conv	vs Normal	Combined Anomaly
Feb 8 (baseline)	234	1.0x	144	1.0x	—
Feb 9	871	4.4x	792	5.5x	1,663
Feb 10	1,790	9.0x	1,666	11.6x	3,456
Feb 11	2,521	12.7x	2,357	16.4x	4,878
Feb 12	2,664	13.5x	2,569	17.8x	5,233
Feb 13	918	4.6x	805	5.6x	1,723
Feb 14 (return to normal)	207	1.0x	133	0.9x	—

16,953

Total excess conversions across both actions during the 5-day anomaly window

$161K

Excess vega Sales value attributed to ad campaigns during the spike vs normal daily average

5 days

The anomaly started Feb 9, peaked Feb 12, and fully resolved by Feb 14 — a classic batch import curve

8%

Ad spend only increased 8% during the spike — confirming these were not ad-driven purchases

04

Full February — Day by Day

Every day, every number

Complete daily breakdown showing spend, vega Sales conversions, revenue, and single-day ROAS. Red rows are the anomaly days.

Date	Spend	vega Sales Conv	vega Sales Rev	Day ROAS	Flag
Feb 1	$8,567	198	$21,729	254%
Feb 2	$7,291	233	$20,055	275%
Feb 3	$7,753	195	$20,127	260%
Feb 4	$5,654	170	$16,680	295%
Feb 5	$5,768	212	$16,788	291%
Feb 6	$6,359	143	$13,921	219%
Feb 7	$5,854	199	$16,876	288%
Feb 8	$6,451	234	$21,690	336%
Feb 9	$5,962	871	$34,774	583%	ANOMALY
Feb 10	$6,723	1,790	$50,002	744%	ANOMALY
Feb 11	$8,009	2,521	$61,723	771%	ANOMALY
Feb 12	$8,098	2,664	$63,632	786%	ANOMALY
Feb 13	$7,553	918	$42,724	566%	ANOMALY
Feb 14	$6,853	207	$20,025	292%

Pattern: Normal days (Feb 1–8, Feb 14+) show 143–234 conversions/day. The spike starts Feb 9, peaks Feb 12 at 2,664 conversions, and drops back to normal by Feb 14. This ramp-up → peak → tail-off is the signature pattern of a batch data import processing over several days.

05

The Impact

February reported ROAS was inflated by $161K

Removing the excess conversion value from the 5 anomaly days reveals the true February performance:

Reported February

351%

$730,092 conv. value
$208,119 spend

Includes ~$161K in phantom
revenue from batch import

Corrected February

273%

~$569,000 conv. value
$208,119 spend

Based on normal daily average
of ~192 conv / $18,326 per day

Phantom Revenue

$161K

excess vega Sales value from batch import

ROAS Inflation

+78pts

351% reported vs 273% actual

Excess Conversions

7,804

vega Sales conversions above normal baseline

True Feb Performance

273%

still strong — real purchases, real ROAS

06

Visual Evidence

The spike is visible in the Google Ads dashboard

This chart from the Google Ads account shows cost (blue) staying flat while conversion value (green) and conversions (orange) spike dramatically in mid-February — the visual fingerprint of a data import.

Google Ads dashboard showing February conversion spike

Google Ads Account Overview — Jan 21 to Apr 20, 2026. The mid-February spike is clearly visible.

What you're seeing: The green line (Conv. value) and orange area (Conversions) both spike dramatically while the blue line (Cost) remains completely flat. This pattern — unchanged spend with an explosion in conversions — is the definitive indicator of historical data being imported into the conversion tracking system.

07

What This Means

The data is clear. Here's the summary.

Key findings from the forensic analysis of February 2026 conversion data:

1

A batch data import occurred Feb 9–13. Both vega Sales and Subscription conversions spiked simultaneously while ad spend stayed flat (+8%). This is consistent with historical Shopify transaction data being backfilled into Google Ads conversion tracking.
2

February reported ROAS of 351% is inflated. Approximately $161K in phantom revenue was attributed to ad campaigns during the import window. The corrected February ROAS is approximately 273%.
3

This affected the bidding algorithm. Google's automated bidding saw 2,664 conversions on a $8,098 day and interpreted it as genuine ad performance. This trained the algorithm to expect unrealistic conversion rates, potentially impacting bid calibration for weeks afterward.
4

The underlying February performance was still solid. At 273% corrected ROAS on ~$7,400/day spend, February was performing well on real purchases. The issue is purely data integrity, not campaign performance.
5

This has been identified and isolated. The anomaly is confined to these 5 days. January and March data do not show similar batch import patterns in the vega Sales conversion action.

08

February ConversionData Anomaly

Reported February

Corrected February

February Conversion
Data Anomaly