HomeDeco BE — Campaign Segmentation Plan

5 campaigns. Conversion value = commission, so ROAS = POAS and breakeven = 100%. ← BE status page · NL plan →

Why discovery-by-margin, not by label. BE runs ~170–210 conversions/month across 111k products, so products convert very rarely. A time-aware backtest (labeling each conversion by what the product was when it converted) shows ~83% of conversions are cold starts — a dormant "Zombie" gets a click, converts, then graduates to Winners the next day. So the discovery pool, split by margin, is the engine. Winners and Villains are small but useful carve-outs (scale the proven; restrict the tested losers).

The 5 campaigns

Conversions = time-aware backtest (Apr 15 – Jul 15), per month. Only campaigns holding ≥50 conv/mo can run Target ROAS; the rest run Maximize Conversion Value (correct for their scale-proven / discover / restrict roles). Daily € are starting points from the ~€3.7k/mo BE spend — adjust after launch.

#	Campaign	Filter (custom labels)	Conv/mo	Daily €	Bid strategy
1	BE-PMX-Winners	CL3 ∈ {Heroes, Sidekicks}	31	18	Max Conv Value
2	BE-PMX-Villains	CL3 = Villains	4	3	Max Conv Value / restrict
3	BE-PMX-Zombies-Low Margin	CL3 = Zombies & CL1 ∈ {1-5, 5-10}	77	40	Target ROAS 110%
4	BE-PMX-Zombies-Medium Margin	CL3 = Zombies & CL1 ∈ {10-20, 20-30, 30-50}	69	38	Target ROAS 110%
5	BE-PMX-Zombies-High Margin	CL3 = Zombies & CL1 ∈ {50-70, 70-100, 100+}	31	18	Max Conv Value
	TOTAL		~213	117

Only Zombies-Low + Zombies-Medium clear the 50-conv/mo floor → real Target ROAS. Winners (31), Villains (4) and Zombies-High (31) sit below it → Maximize Conversion Value (no target). Villains is deliberately tiny — a restrict bucket, minimal budget.
Zombies-High (31) borderline: if you'd rather run it on tROAS, pool it into Medium (Med+High = 100/mo ✅) — that's 4 campaigns instead of 5.
Backtest spans the stronger May/June (~213/mo); the recent 30-day run-rate is softer (~168), so treat budgets a notch conservative and watch the Winners pool — if it dips under ~50 keep it on Max Conv Value.

Labelizer classification thresholds — how the daily label (custom_label_3) is set

3 segments by margin, mirroring the 3 Zombie campaigns (Winners and Villains span all margins). Producthero 2x2 per segment: clicks > threshold decides Hero/Villain vs Sidekick/Zombie; POAS ≥ target decides Hero/Sidekick vs Villain/Zombie. Threshold is kept below each segment's average clicks-per-conversion (≈ 0.6×), so a product is only branded a Villain after it has had a real shot. Target ROAS = 110% (10% above the 100% breakeven).

Segment (sets the threshold)	Products	Click threshold	Target ROAS	Avg clicks/conv
Low Margin	48,186	30	110%	54
Medium Margin	39,459	40	110%	78
High Margin	23,705	65	110%	118

Current label snapshot: Heroes 1 · Sidekicks 153 · Villains 9 · Zombies 111,187. Winners pool = 154 products.
The threshold only controls the Villain cut (which non-converters get restricted vs sent to discovery) and the Hero/Sidekick split inside Winners — it does not move products between converting and non-converting. Raise a segment's threshold to restrict fewer products as Villains; lower it to restrict more.
Live thresholds + clicks/conv always current on the status page.

Snapshot 2026-07-16 · Google Ads acct 6343955323 · conv value = commission (POAS, breakeven 100%) · backtest Apr 15 – Jul 15.