Dog Academy — research summary (Meta leads → purchasers)

2026-09-25 UTC · research phase only · no forms submitted, no Meta changes. Detail files are linked at the bottom. Labels: seen = observed on the live site/docs; inferred = our reading.

Bottom line

1. The product that makes the money is invisible on the site. Trained for Life (about $1,200) is sold on the phone. Online it exists only as legal terms and one unlisted, no-index page whose GET STARTED button is a dead link (seen).
2. The lead path: every "Book Training" button opens a 22-step quiz. It then sends the lead to self-book a "free 30-minute virtual meet & greet", and a sales call pitches TFL (seen / inferred). There is also a second intake: Heyflow city funnels on training.dogacademy.org. They show the phone number used for Google ads, so they are probably Google's landing pages (inferred).
3. Meta is blind past the page view. There is one browser-only pixel with no Conversions API and no event IDs. The program pages and the Heyflow funnels carry no pixel at all. Meta never learns who booked, showed up or bought (seen).
4. Meta only learns from website events within 7 days of the click. A phone sale that closes 10–20 days later never trains a website campaign.
   • The route that works for this funnel: Instant Forms on the "Maximize number of qualified leads" goal, with the CRM sending each stage back through the Conversions API. It learns from stages up to 28 days after the lead.
   • Optimize to consultation completed first, then to sale once the volume allows (Meta docs).
5. Blocker: our Meta login can't see Dog Academy's ad account, Page or pixel 481578137302100. We need access to the Business that owns them before anything else.

The offer

• Trained for Life terms (effective 2026-02-02, seen):
  • one dog, for the dog's life or 14 years;
  • up to 8 in-person sessions, at most 1 a week, plus virtual sessions (at most 2 a week);
  • online courses;
  • aggression cases start virtual-only;
  • a $300 transfer fee if the dog is re-homed;
  • refund 100% within 30 days or 50% within 6 months, and only if no replacement trainer exists.
• Matthew's version: up to 8 in-person sessions, unlimited virtual sessions and the courses. "Unlimited" fits the terms as a lifetime total, at up to 2 a week.
• Price anchor (inferred): the refund policy values an in-person session at $159.99, and 8 × $159.99 is about $1,280, close to the ~$1,200 AOV. No TFL price appears anywhere public.
• Variants on the call:
  • In-Person TFL and Live Virtual TFL;
  • Aggression In-Person and Aggression Virtual;
  • a separate psychiatric-service-dog (PSD) funnel and script.
• Public prices conflict:
  • private training is $75–$130/hr (homepage FAQ), $130–$180/hr + Klarna (program pages), and "from $95" (SEO pages);
  • group is $39–$75/hr, live online $35–$50 per 25 minutes, courses $6.99–$349.99.
• Proof on the site:
  • "1,000+ trainers" (other pages say "hundreds" or "a dozen AKC-certified");
  • static 5-star badges; live Trustpilot 4.4 from 1,583 reviews;
  • city pages carry an unsourced "Top 5 Trainer {City}" badge and the same six "local" trainers in every city.
• Not on the site: any "Voted #1" claim (which the Google ads use), a military discount (which the call script asks about) or an insurance offer.

The lead path in detail

Friction (inferred): 18 input screens, and 4 TFL, service-dog and insurance screens before contact details. A failed step shows no error. Adding ?zip_code= to the URL skips step 1.

Tracking today (seen)

• Meta pixel 481578137302100:
  • hard-coded, browser only;
  • PageView plus automatic button clicks;
  • quiz steps fire server-chosen fbq('track', name) with no event ID, value or user data;
  • automatic advanced matching is on;
  • no CAPI or Gateway.
• No pixel on the private, group and online program pages or on the Heyflow funnels.
• Call tracking by source:
  • utm_source=facebook → 605-377-6002;
  • Google cpc → 605-220-7609;
  • everything else → 605-223-1971.
  • Facebook calls can be counted at source level, not per click.
• Stack:
  • a Laravel/Livewire app on AWS;
  • its own attribution cookie (UTMs/gclid, not fbclid alone);
  • GA4 G-EFSX7XBE5Q through a legacy tag, with a city_lead key event and a Google Ads link;
  • Klaviyo, server-side Amplitude, TikTok, Bing, HelpScout.
• Not visible from outside: the CRM, dialer, scheduler and payment processor.
• Inferred: the Laravel backend already sees the lead's IP and can read _fbp/_fbc, so it is the natural Conversions API sender.

Conversions API plan (research recommendation, not built)

• Stages to send, from the CRM:
  • Lead received
  • Consultation scheduled
  • Consultation completed (first optimization target)
  • Sale closed (value about $1,200)
• Meta's gates (docs):
  • 200 leads a month;
  • at least 60% of leads matched with stage events, and at least 7 days of data;
  • the target stage converts 1–40% of leads within 28 days;
  • daily uploads, each event within 7 days of its time;
  • the CRM must store Meta's lead_id;
  • training takes about 1–2 months, and switching datasets restarts it.
• Why not optimize to Purchase through the pixel: 50 purchases a week per ad set at a 5.5% close rate needs about 909 leads a week, which is $36k a week at a $40 CPL.
• Where the website form still works: if most consultations happen within 7 days of the click, a website "consultation completed" event sent through the Conversions API can also be optimized. This is worth a website-form test cell once the pixel is fixed. Meta's Help Center and developer docs disagree on whether qualified leads works for website forms, so check in Ads Manager.
• Economics: revenue per lead is about $66 ($1,200 × 5.5%). Cost per sale is 18 × CPL, so a $30 CPL means about $545 per sale. We need the margin to set a real CPL target.

Pain-point bank

The table and caveats are in brands/dog-academy/NOTES.md. The top three are ignoring commands (16.7%), pulling on leash (15.3%) and jumping (12.7%), 45% of all selections. The counts show how often each pain is reported, not which one buys.

Reputation and customer voice (public reviews, seen)

• The headline rating hides the in-person problem.
  • The Trustpilot 4.4 (1,583 reviews) comes mostly from course buyers and from reviews written right after the sales call.
  • Of the last 200 reviews, 40 are about in-person training, and 17 of those 40 (43%) are 1–2★. Course reviews are about 8% 1–2★.
  • BBB rates it B, with 15 complaints, 14 of them in the last 12 months and almost all about in-person or lifetime training.
• In-person complaints (68 unique):
  • refund denied or reduced: 57%
  • trainer quality: 38%
  • sales promises or pressure: 32%
  • no local trainer: 31%
  • no-shows: 28%
  • pushed to virtual: 15%
• Prices customers report:
  • packages of $1,095–$2,345, including $1,595 for 11 sessions, which DA's own BBB reply confirms;
  • the $495 "Forever Framework" first session is the most disputed charge.
• Customers never say "Trained for Life". They say "lifetime", "unlimited" or "Forever Framework", and those words now show up mostly in broken-promise complaints.
• The sales call is the strongest part of the funnel: callbacks "within 10 minutes" and consultations of 45 minutes to 2 hours. The follow-up is heavy. The first review Trustpilot shows as "relevant" today is a 1★ about texts from someone who says they never filled out a form.
• Buyers' own words:
  • Pains: "he didnt listen to commands", "just about pull my arm out of socket", "first time dog owner… overwhelmed".
  • Failed alternatives: other trainers, YouTube.
  • Fear: re-homing the dog.
  • Wins: "I can take them anywhere without worrying", "a different dog on walks", "Cujo to a sweetheart".
• Reviews ≠ form (seen / inferred):
  • The reasons to buy that reviews mention skew to not listening, puppies, reactivity and pulling. Jumping and digging never appear.
  • Proof that lands is specific: a named local trainer, change within 1–3 sessions, and an owner who now knows what to do.
• For the CAPI plan (inferred): refunds are the top complaint. Before optimizing to "sale", consider counting only sales that are kept.

Offer facts to confirm before any ad copy

• The unlisted TFL page says "Unlimited Training… as many sessions as it requires" and "No Hidden Fees. Ever." The terms say 8 in-person sessions, a $300 transfer fee and possible relocation fees. What do ads say?
• Is the next step a concierge call or a self-booked meet & greet, and who runs it: a sales agent or a trainer?
• Can "Voted #1" and "Top 5 Trainer" be used, and what are they sourced from?
• Does a veteran discount exist? The script asks every caller.

Questions for Matthew

1. Meta access: which Business owns pixel 481578137302100 and the Dog Academy Page? Add Brett as admin or give partner access.
2. Which system holds scheduled, completed and sold? Can it send webhooks or API calls? Does it keep a lead ID?
3. The lead-to-sale lag: median, 90th percentile, and the share closed within 7 and within 28 days.
4. Monthly volumes and rates for lead → scheduled → completed → sold.
5. TFL price per variant, and how payment is taken (on the call? a link?).
6. The margin after trainer payout, to set CPL and CPA targets.
7. Service areas: nationwide, or the metros with trainer coverage?
8. The source, period and lead count behind the issue-count table.

Detail files

• OPERATOR-BRIEF.md: the brief, offer facts, call-script summary
• site/SITE-AND-OFFER.md: site map (25,523 URLs), offer, landing pages, contradictions
• flow/LEAD-FLOW.md: the 22-step quiz, Heyflow, post-submit, 43 screenshots
• tracking/TRACKING-STACK.md: pixel, tags, call tracking, stack, what's missing for CAPI
• capi/META-LEAD-TO-PURCHASE-OPTIMIZATION.md + capi/event-spec.json: the CAPI plan with Meta sources
• meta-account/META-ACCOUNT-READBACK.md: what our Meta login can and can't see
• voc/VOICE-OF-CUSTOMER.md: reviews and customer language
• prior/PRIOR-WORK-DIGEST.md: the 2026-09-22 Codex work, never reviewed or launched