Hunt not perfect?
Sharpen spear.
First hunt never perfect. Nobody's is. Optimization is look at bones, find dull part of spear, sharpen it, hunt again. Small fix. Many time. Spear get very sharp.
Sharpen loop
Four step. Go round and round. Forever. This is whole secret of optimization.
Look
Count bones. Break apart. Find winner and loser.
Think
Why loser lose? Guess one reason. Write it down.
Change one thing
One. Not five. Then you know what work.
Wait
Let bones pile up. 3–7 day. Then go back to 1.
↻ Look → Think → Change → Wait → Look again
Ten way sharpen spear
Levers to pull. Start at top — top levers usually biggest win for least work.
- Learning Phase & Statistical Significance
Wait for enough bone
No change hunt every day. Each big change reset platform learning. Wait until each party/painting have enough data: rough rule 100+ click to judge CTR, 30–50 conversion to judge CPA. Small pile lie. Use A/B checker below.
- Budget Reallocation
Move rock to winner
Hunting party with best CPA/ROAS get more rock. Worst party get less, or pause. Move in steps — 20–30% at a time — so platform no freak out. Remember rock you saved in pouch? Throw here.
- Creative Rotation & Fatigue
Kill weak painting. Refresh tired painting.
Painting that lose test: pause. Painting that WAS good but CTR fall week after week: beast tired of it (fatigue). Make new version. Most painting get tired in 2–6 week.
- Frequency Capping
Stop show same beast too much
Set max times one beast see ad: e.g. 3 per day, or 5–10 per week. Too much = rock wasted on same beast, and beast start hate you. Retargeting need cap most.
- Bid Optimization
Change how much pay
Not spending? Bid maybe too low — raise 10–20%. Spending fast but CPA bad? Lower bid or set cost cap. Auto bid: change target CPA/ROAS slowly. Bid more for high-value beast (retarget, best land, best device).
- Placement Exclusions / Block Lists
Cut bad hunting spot
Look at site/app report. Some spot eat rock, give nothing — or have weird high clicks (bots, kid games, fat finger). Add to block list. Good spot go on allow list. Also set brand safety: no ad next to scary or bad content.
- Dayparting / Ad Scheduling
Hunt when beast awake
Look at bones by hour and day. If beast buy at night and weekend, put more rock there. If 3am catch nothing — turn off or lower bid. Need enough data first.
- Audience & Geo Refinement
Tighten or loosen herd
Some audience or land catch cheap beast? Split into own party, give more rock. Herd too small, no spend? Loosen: remove layer, widen lookalike 1% → 3%, bigger circle. Add converters to exclusion.
- Landing Page / CRO
Fix cave entrance
Many poke but no catch? Problem maybe not ad — maybe cave. Page slow? Message no match? Form too long? Button hidden? Fix page = every ad work better. Called conversion rate optimization (CRO).
- A/B Testing
Always have one test running
Split herd in two. Show A to half, B to half. Change only one thing. Winner become new normal. Then test new thing against winner. Never stop. Tribe that test, tribe that win.
What Wrong With Hunt?
Pick what you see in bones. Grog tell you what probably wrong and what to fix first.
Pick symptom above
Grog wait. Grog patient. Grog mammoth.
Which Painting Win? · A/B Checker
Put eyeballs and pokes (or pokes and catches) for two painting. Checker say if winner REAL or just luck.
Painting A
Painting B
Uses two-proportion z-test. 95% = strong. Under 90% = maybe luck, keep hunting.
Smart hunter do
- Change one thing at a time
- Wait for enough bones before judge
- Move budget in 20–30% steps
- Refresh painting before it tired
- Write down every change + date
Silly hunter do
- Change bid, audience, and painting same day
- Kill party after 1 day and 4 click
- Double budget overnight
- Run same painting 6 month
- Chase cheap click, forget real goal
Before you leave cave
You finish six step. You real hunter now. But beast change, season change, other tribe change. Go back to Big Rock. Start new hunt. Use what you learn. Each hunt sharper than last. You smash. Grog proud.