Every ad dollar from Meta, Google, and the 2026 budget sheet, matched against Builder Prime leads and sold jobs — plus what the numbers say to do next as the marketing moves in-house.
| Source | Yesterday | Last 7 days | Value per lead | Cost per lead | Est. 7-day value |
|---|
Counts pull nightly from the shared lead tracker — aggregate numbers only, no customer details. "Value per lead" is what a lead from each source historically turned into in sold revenue over the trailing CRM year (revenue, not profit — send job-cost data to upgrade this to true profitability). Recommendations are generated daily from the last 30 days of flow plus last year's seasonal pattern. Today's bar keeps filling in until end of day.
Google Local Services closed 43% of its leads and sold $315K on under $6K of spend (Aug ’25–Jan). Since February it has logged zero leads — while spend nearly tripled to $2.8K/mo — $15.5K spent for nothing, and still burning at ~3× its $1K/mo plan per the August weekly sheet. The account is either paused, suspended, mis-targeted, or leads aren't reaching the CRM. Finding out is the single highest-value hour anyone can spend this week.
GLS leads per month, Sep → Aug. Bars stop dead in Feb.
Feb–Apr, a sold job cost ~$610 in ad spend. May–Jul, spend jumped from ~$47K to ~$72K/mo and cost per job hit $1,100–$1,600 — the extra dollars bought progressively worse leads. The last ~$25K/mo of spend was producing far below breakeven quality. Scaling should have gone into budget earlier in the year (see seasonality), not higher in the summer.
Meta delivered 4,621 leads (81% of all paid leads) but closed only 5%, vs Google's 19.5%. 96%+ are Instant Forms — people who never left Facebook. That's not a reason to kill Meta (5.7× ROAS); it's a reason to treat those leads differently: instant AI speed-to-lead, qualification questions on the form, and a nurture sequence, instead of the same follow-up as a high-intent Google search lead.
Since the Indy accounts launched (Apr 2026): $123K spend → 36 sold jobs → $303K. That's ≈$3,400 per sold job and ~2.5× ROAS, vs NWI's ~$600–1,100 and 6–7×. Worse: the Indy budget sheet shows Google ran $20.9K over plan and Meta $24.4K over plan — the weakest market got the biggest unplanned raise — and recent weeklies show Indy Google at ~$2.9K/week for 1–2 leads. New markets ramp slowly, but Indy needs its own creative, reviews, and a GLS/LSA presence — not a copy-paste of NWI campaigns at growing budgets.
ChatGPT Ads: $4,660, zero leads — and per the August weekly sheet it's still running at ~$700/week. Google (non-GLS) ran 11 straight weeks — Sep through mid-Nov 2025 — spending $24.8K with zero CRM-matched leads; tracking gap or waste, either way unacceptable, and nobody caught it for nearly three months. A TOFU test in IL/MI: $6,176 for 2 leads ($3,088 each). Bing and TikTok were rightly killed early. Rule going forward: any channel that can't show a CRM-matched sale inside 60 days gets cut.
Client referrals closed 51%, past clients 48%, organic 38% — together $2.1M sold at near-zero media cost. There's no line in the budget for a referral program. A systematic ask-for-referral + review engine (automated post-job, $250–500 reward) is likely the cheapest revenue available to the business.
Dollar value of jobs sold, from leads created in each month. August is faded — its deals are still closing.
All paid platforms combined.
Spend ÷ jobs sold that month. Lower is better. Jul–Aug faded: close-rate lag.
Share of CRM leads that became sold customers. Google leads are ~4× more likely to buy.
You're right that Northern Indiana winters hit the top line: January was the year's worst month ($290K sold, vs $807–823K in March–April). Lake-effect snow, frozen ground, and nobody thinking about their garage floor in single digits — the demand dip is real. But the data says something more useful than "Q1 is bad":
February–April is the most efficient window of the entire year. A sold job cost $567 in February, $626 in March, $642 in April — versus $2,000–2,500 in November–December and $1,100–1,600 in May–June. Google close rates peaked at 27–31% in Feb–Mar. The people who inquire in late winter are planners getting ready for spring; they're fewer, but they buy.
Part of the bad Q1 is self-inflicted. The 2026 budget cuts to $45K in Jan–Feb, then jumps to $75K in March — spend arrives after the highest-converting leads have already been bought cheap. Radio, events, and print were all dark in Jan–Feb too. Meanwhile the truly inefficient months (Nov–Dec: Meta closing ~3.3%, cost per job $2K+) kept full budgets.
Move ~$15–20K/mo out of Nov–Dec and mid-summer, into Feb–April. Same annual spend, pointed at the window where a job costs $600 instead of $2,000. Ramp Google Search from mid-January — that's when planner-intent shows up.
Garage epoxy is indoor work — capacity doesn't freeze, demand does. Run a "Book by Feb, installed before spring" offer with real scarcity (install-slot calendar), and push commercial/industrial floors (weather-independent decision makers) Nov–Feb.
Sep–Oct leads are plentiful but close soft. Capture them into an email/SMS nurture and re-offer in January with winter pricing, instead of buying brand-new expensive leads in the trough.
| Paid channel | Spend | Leads | Sold | Close | Cost/lead | Cost/sale | Sold $ | ROAS |
|---|---|---|---|---|---|---|---|---|
| Google LSA (GLS) | $21,398 | 88 | 38 | 43.2% | $243 | $563 | $315,120 | 14.7× |
| Google (Search/PMax/YT) | $250,660 | 1,089 | 212 | 19.5% | $230 | $1,182 | $1,801,610 | 7.2× |
| Meta Ads | $318,115 | 4,621 | 232 | 5.0% | $69 | $1,371 | $1,817,380 | 5.7× |
| ChatGPT Ads | $4,660 | 0 | 0 | — | — | — | $0 | 0× |
GLS spend split out of the Google account total using campaign names. Google leads exclude GLS (separate CRM source). Jul–Aug close rates still maturing, so full-year ROAS is slightly understated.
| Non-paid source | Leads | Sold | Close | Sold $ | Avg job |
|---|---|---|---|---|---|
| Organic | 264 | 100 | 37.9% | $969,807 | $9,698 |
| Client referral | 166 | 84 | 50.6% | $682,602 | $8,126 |
| Past client | 83 | 40 | 48.2% | $445,100 | $11,128 |
| Misc / other | 235 | 33 | 14.0% | $260,341 | $7,889 |
| Radio | 52 | 19 | 36.5% | $135,425 | $7,128 |
| Referral partner | 30 | 12 | 40.0% | $122,486 | $10,207 |
| Website lead form | 63 | 13 | 20.6% | $117,075 | $9,006 |
| Direct mail | 48 | 23 | 47.9% | $116,800 | $5,078 |
| Google Business Profile | 7 | 4 | 57.1% | $49,860 | $12,465 |
| Website (generic tag) | 47 | 3 | 6.4% | $20,800 | $6,933 |
Radio spend was ~$24K (2026 budget) against $135K sold — roughly 5.6× on the matched portion. The generic "Website" tag closing at 6% vs "Website Lead Form" at 21% suggests a CRM tagging or form-routing problem worth cleaning up.
77% of all Meta spend ($243K) sits in one campaign ("201 – Mega Campaign"), 98% in two. That's fine while it works — and CPLs of $35–75 on the best ad sets are genuinely good — but it means one fatigued campaign or one account issue takes down 81% of your lead flow. Going in-house, keep 2–3 always-on campaigns with fresh flake/metallic transformation creative rotating monthly, and kill any ad set above ~$100 CPL (several ran $99–$125+, and the IL/MI test hit $3,088).
Budgeted amount per the master sheet vs. delivered media spend. August is partial (through 8/29).
NWI: Jan–Aug budget was $540K; delivered media was $437K — about $103K under, concentrated in Apr–Aug. Meanwhile Indy — the market closing at 2.5× — ran $45K over its plan (Google +$20.9K, Meta +$24.4K). Dollars flowed away from the proven market and window, toward the unproven one. The August weekly sheet also shows a week-one pacing stall ($2.5K delivered against ~$12K planned), then a scramble to catch up.
NWI: management fee $55.7K + retainer $21.0K. Indy: management fee $19.6K (and it ran $6.6K over its own planned fee). Total: $96.3K for 8 months — a ~$144K/yr pace, ≈16% on top of media. That's the going-in-house budget: it funds serious AI tooling, tracking infrastructure, creative production, and a part-time specialist, with money left over.
A %-of-spend management fee gets paid more for spending more — the May–Jul scale-up that doubled your cost per job also raised the fee. Whoever manages this next (in-house or contractor), tie compensation to cost per sold job and CRM-verified revenue, never to spend.
Supermetrics ad-platform spend + Builder Prime CRM export, Aug 29 2025 – Aug 29 2026, matched by month + source (no lead-level key exists yet — that's the UTM fix). Master budget workbook (NWI + Indy year tabs, plus monthly weekly-pacing tabs) for planned vs. actual and agency fees. One year of data, not two — if Builder Prime holds an earlier year, exporting it would let us verify the seasonality pattern across two winters.
· Jul–Aug close rates and cost-per-sale are artificially poor (deals still in the sales cycle) — faded in every chart.
· Aug 2026 shows $585K sold-amount against 3 closed customers: the CRM's Sold Amount and Customer-status fields disagree and need an audit.
· Sep–Oct 2025 Google shows spend with zero matched leads — tracking gap or waste, unresolved.
· Indy had no separate ad accounts before Apr 2026.
· ~3% of leads have no region tag.
1. Day-by-day spend export (past month) → daily pacing panel above.
2. Prior-year Builder Prime export if it exists → two-winter seasonality proof.
3. GLS account status screenshot → resolve the #1 finding.
4. Job-cost / margin data → shift every target from revenue-ROAS to profit-ROAS.
| Month | Ad spend | Paid leads | Paid sold | Paid sold $ | Other sold $ | Total sold $ | Cost/sold job |
|---|