apphouse-marketing

name

apphouse-marketing

description

Evidence-based marketing for AppHouse iOS apps — turn real user pain into angles, hooks, and short-form content (TikTok/Reels/Shorts videos and slideshows). Covers marketing fundamentals (pain → desire → angle → hook), engagement design (comments/shares/saves), TikTok-native captions and safe zones, trending sound selection, per-platform publish kits (title/description/hashtags), and a spec-driven production workflow (AI generation + FFmpeg assembly). Use when the user mentions marketing, ads, hooks, content ideas, video scripts, UGC, slideshows, TikTok/Reels/Shorts, captions, trending sounds, "make a marketing video", "content for my app", or "why isn't my app getting downloads".

AppHouse Marketing

Marketing here means one thing: find the real pain, promise the real transformation, prove it fast. No growth-hack mysticism, no guaranteed-viral promises.

Core rules

1. Evidence before angles. Never invent a pain point. Pull it from: App Store reviews (yours and competitors' — use app-store-reviews), support emails, comments on competitor content, search terms. A hook built on a quoted review is aimed; a guessed one is a coin flip.
2. One creative = one pain, one desire, one message, one engagement trigger. Never stack.
3. Claims come from the manifest. Read apphouse.project.yml → product.approved_claims / prohibited_claims. Nothing outside approved claims goes into any script or caption. If missing, establish them with the user first.
4. Hook decides everything. ~80% of effort goes into the first 1–3 seconds. Always produce multiple hook variants per concept.
5. Provoke engagement, never request it. Comments/shares/saves are what algorithms pay for — design them in with the tactics in references/engagement-playbook.md. Literal "like and share!" begging is demoted by platforms.
6. Safe zones and native captions always. All text/faces/CTAs inside the universal safe rect; captions in the TikTok-native style (white bold, black outline). Exact numbers and the ready-made ASS style: references/platform-specs.md.
7. Never promise virality. Volume + iteration wins. Judge by hold/completion and comments/shares per reach — never raw views.

Folder structure

Each app keeps a marketing/ folder:

marketing/
├── strategy/
│   ├── brief.md            # strategy brief (pain, desire, proof, claims)
│   └── angle-bank.md       # angles: untested / testing / winner / dead
├── creatives/
│   └── <id>/               # e.g. 2026-07-ugc-01  (one folder per creative, video OR slideshow)
│       ├── spec.yml        # from templates/video-spec.yml — approved BEFORE production
│       ├── assets/         # recordings, generated clips/images, VO, captions.ass
│       ├── output/         # final exports (master + platform variants)
│       └── publish.md      # copy-paste kit: sound, captions, titles, hashtags, pinned comment
└── learnings.md            # running log: what held, what died, why

Creative IDs: <yyyy-mm>-<format>-<nn> (e.g. 2026-07-slideshow-03). Everything needed to reproduce or vary a creative lives in its folder.

Workflows

A. Strategy — build the angle bank (once per app, refresh quarterly)

1. Read apphouse.project.yml (product section) for positioning.
2. Gather evidence: mine reviews of your app + top 3 competitors for repeated complaints, desires, exact phrasing (route to app-store-reviews). Verbatim quotes become hook copy.
3. Fill the strategy brief (references/marketing-101.md §Brief) → marketing/strategy/brief.md.
4. Generate 5–10 angles (pain × desire × format), score candidate hooks with the rubric → angle-bank.md.

B. Produce a video

1. Pick one angle; create marketing/creatives/<id>/.
2. Write the spec (templates/video-spec.yml): hook + 3–5 hook variants (references/hooks-library.md), scene list, one planned engagement trigger (comment/share/save tactic from references/engagement-playbook.md), CTA. The spec gets user approval before any generation — text is cheap, video is not.
3. Produce assets per scene: screen recordings (give the user an exact shot list), generated visuals (Higgsfield MCP: generate_image/generate_video, workflows via get_workflow_instructions), voiceover per spec.
4. Pick the sound: trending-sound workflow + licensing rules in references/publish-kit.md. Standard mix: trending sound at ~1% volume under VO at 100%.
5. Assemble with FFmpeg: trim/concat, 1080×1920, loudness-normalize, burn captions using the ASS style in references/platform-specs.md (1–4 word chunks, white/black-outline, safe-zone positioned). Cut hook-variant versions from the same body footage.
6. QA gate: resolution/duration per platform targets; all text inside the universal safe rect (x 60–940, y 250–1420); captions present and synced; first frame works as a thumbnail; audio level; no prohibited claims; no placeholder frames.
7. Write publish.md using the template in references/publish-kit.md: chosen sound, per-platform captions/titles/hashtags (copy-paste ready), pinned comment, first-hour reply plan.

C. Produce a slideshow (TikTok photo mode / IG carousel)

Slideshows cost ~10% of a video and often win for app marketing — run them as a first-class format at volume (4–5/week is realistic).

1. Same spec file, format: slideshow; 6–8 slides: slide 1 = outcome hook with implied speed ("How I ___ in 3 minutes"), slides 2–7 = one step + one screenshot + one line each, last slide = CTA.
2. Slide 1 follows hook rules (~1.5s to survive): bold text, clean background, safe zones.
3. Export: 1080×1920 for TikTok (up to 35 slides), separate 1080×1350 set for Instagram (9:16 gets cropped there). Specs in references/platform-specs.md.
4. Attach a trending sound (slideshows distribute via audio too) + full publish.md like any video.

D. Iterate — learn from performance

• One variable per test (hook OR angle OR format). Metrics map: 3-sec hold → hook; completion → story; profile taps/installs → offer + audience. Diagnose the right stage.
• Post-publish is part of the job: pin the seed comment, reply to everything in hour one, turn the best objection comments into video replies (self-feeding content loop).
• Log results in the spec's results: block and marketing/learnings.md. Dead angles are information — record why.

Guardrails

• Truthful claims only; show the real app, never mocked features that don't exist.
• No fake testimonials presented as real users. UGC-style acted content is fine; fabricated "reviews" with invented names are not.
• No rage bait — it's demoted and poisons the trust an app purchase needs.
• Business-account music licensing is a real legal risk — follow the rules in references/publish-kit.md before using any trending sound.
• Platform fit over polish: native-feeling, slightly raw content outperforms produced ads.

Files

• references/marketing-101.md — fundamentals: brief template, desire taxonomy, angle construction, hook scoring rubric
• references/hooks-library.md — hook patterns and video formats with structure + examples
• references/engagement-playbook.md — comment/share/save triggers, slideshow tactics, anti-patterns, platform algorithm facts
• references/platform-specs.md — safe zones (exact px), TikTok-native caption ASS style + FFmpeg commands, export settings, durations, carousel specs
• references/publish-kit.md — trending sound workflow + licensing rules, per-platform title/description/hashtag rules, publish.md template
• templates/video-spec.yml — per-creative specification (videos and slideshows)