Paid Ads

Paid Ads

Paid acquisition is renting attention. It works when the math works and bleeds money when it doesn't — so set up measurement before you spend, not after. The platform's job is to spend your budget; your job is to make sure it spends on people who convert profitably.

Pick the channel by intent and audience

Rule of thumb: Search captures demand, social creates it. If people are already searching for what you sell, start on Search — it's the warmest traffic you'll buy.

Campaign structure

Keep it tight. Over-segmentation starves the algorithm of conversion data; one giant blob gives you nothing to optimize. Modern platforms (Performance Max, Advantage+) want consolidation — fewer campaigns, more budget per ad set so the optimizer gets enough signal (~50 conversions/ad set/week is the rough learning threshold).

• Google Search: Campaign = budget + goal. Ad group = one tight theme of keywords + matching ads. Don't dump 200 keywords in one ad group; the ad can't be relevant to all of them.
• Meta: Campaign = objective. Ad set = audience + budget + placement. Ad = creative. Test creative within an ad set, test audiences across ad sets.
• One conversion objective per campaign. "Awareness + conversions" in one campaign optimizes for neither.

Targeting

• Search: target keywords with intent, not people. Use match types deliberately — phrase/exact for control, broad only with a good conversion signal and a tight negative-keyword list. Negative keywords are half the job on Search; review the search-terms report weekly and prune junk ("free," "jobs," "diy" if you're not those).
• Social: start broader than feels comfortable. The algorithm finds your buyers faster than your manual interest-stack does. Layer interests/demographics only when you have a reason. Always exclude existing customers from prospecting.
• Retargeting: highest ROAS, smallest reach. Segment by depth (visited pricing > visited homepage > bounced). Cap frequency so you don't burn them.
• Lookalikes/Advantage: feed them a clean, high-value seed (purchasers, not all signups) and they get better.

Budget & bidding

• Start with enough budget to exit the learning phase (need ~50 conversions/ad set/week or the algorithm flails). If you can't afford that on a $50 CPA goal, narrow to one campaign.
• Bidding ladder: start on max-conversions / max-clicks to gather data → switch to target CPA or target ROAS once you have ~30–50 conversions. Don't set a tCPA on day one with no data; you'll choke delivery.
• Change budgets ±20% at a time. Big swings reset the learning phase.
• Set a tCPA from your real unit economics: max CPA = LTV × target margin. If you don't know LTV, you're guessing.

Creative & ad testing

Creative is the biggest lever on social — it beats targeting tweaks. On Search, the offer and landing page are the lever.

Test in this order (biggest impact first):

1. Offer/angle — what you promise and to whom
2. Hook — first 3 seconds of video / the headline
3. Format — video vs static vs carousel
4. Then details: colors, button copy, music

• Test one variable at a time or you can't attribute the win.
• Run 3–5 distinct creatives per ad set; let the algorithm allocate. Kill the bottom performers, not after 100 impressions — wait for ~1,000+ impressions or a few days.
• Creative fatigue on social is real: when frequency climbs and CTR drops ~30%+ from peak, refresh the creative. Budget for a steady creative pipeline, not one hero ad.
• Match the ad to the landing page (message match). A great ad to a mismatched page wastes the click.

Tracking — do this before you spend a dollar

If you can't measure a conversion, you can't optimize toward it. Set this up first.

Core metrics — what they mean and what's "good"

Reading them together (the funnel diagnoses itself):

• Low CTR → ad/creative or targeting problem (top of funnel).
• Good CTR, low CVR → landing page / offer / message-match problem (bottom of funnel).
• Good CVR, bad CAC/ROAS → CPCs too high or LTV too low — economics, not execution.
• Profitability check: ROAS must clear breakeven 1 ÷ margin (50% margin → need ROAS ≥ 2.0 just to break even). Or judge by CAC < LTV — payback under ~12 months for most businesses.

Launch checklist

[ ] Conversion tracking live + test-fired (pixel + server-side)
[ ] UTM convention documented and applied to every link
[ ] Target CPA/ROAS derived from real LTV and margin
[ ] One objective per campaign; tight ad groups / clean audiences
[ ] 3–5 creatives per ad set; message-matched to landing page
[ ] Negative keywords loaded (Search); customer-exclusion list (social)
[ ] Budget high enough to exit learning phase
[ ] Daily/total budget caps and a kill date set
[ ] Landing page fast, mobile-clean, single CTA

Diagnose a failing campaign

Spending but ZERO conversions?
├─ Is tracking actually firing? (test the event FIRST — most "failures" are broken pixels)
├─ Tracking OK → is the landing page converting non-paid traffic? If no → fix the page, not the ad
└─ Page OK → wrong audience/intent → tighten targeting / keywords

Getting clicks but no conversions? (good CTR, low CVR)
└─ Landing page, offer, or message-match. The ad wrote a check the page didn't cash.

No clicks? (low CTR)
└─ Creative or targeting. Refresh hook/angle; check you're not showing to the wrong people.

Converting but unprofitable? (good CVR, bad ROAS/CAC)
├─ CPCs too high → improve Quality Score / relevance, or shift channel
└─ LTV too low → it may just not be a viable paid channel for this product

Was working, now decaying?
├─ Frequency up + CTR down → creative fatigue → new creative
└─ CPMs up seasonally (Q4/holidays) → expected; reassess targets

When to cut: an ad set that's spent ~3× your target CPA with zero conversions is dead — pause it. When to scale: when CAC is comfortably under LTV and stable for a week+, raise budget ~20% at a time and watch CPA hold. Scaling too fast inflates CPA and resets learning.