Pulsar · Metrics playbook · No. 07 · Funnel leak diagnosis. Real funnel from the test pixel, 30-day window.
The Leak Map
Your money leaks at one step more than the others. Find that step, fix that step — before you spend a cent more on traffic.
Companion to No. 02 — Quantity vs. Quality (which introduced this funnel), No. 01 — Efficiency & Fatigue, No. 03 — Garbage In, Lookalike Out, No. 04 — The Invisible Sales, No. 05 — Who Really Made the Sale? and No. 06 — Your Ad Is Dying.
Contents
01 The widest crack · 02 The funnel, drop by drop · 03 Read the leak, not the total · 04 What each leak means · 05 One leak at a time · 06 Beats buying traffic · 07 The fix
Money leaks at the widest crack
A funnel loses people at every step — but not evenly. One step almost always leaks far more than the rest. Pouring more traffic in without fixing the biggest crack just spills faster. Find the widest crack first.
More traffic doesn’t fix a leak — it just makes you lose money faster at the same broken step.
This deck maps a real 30-day funnel from the test pixel (distinct visitors, counted as people — not sessions), shows how to find the step where the money actually escapes, and matches each kind of leak to its fix. The funnel itself was introduced in No. 02 — Quantity vs. Quality.
The real funnel, drop by drop — real data
The test pixel, 30-day window, distinct visitors. Read the step-to-step drop on every transition — one of them is nothing like the others.
Bar widths are on a square-root scale so the small stages stay visible — the numbers are the truth, the bars are the shape. Distinct people at each stage, 30-day window.
Read the leak, not the total — the leak map
Two shops can have the same 0.27% end-to-end conversion and leak at totally different steps — and need totally different fixes. The end-to-end number tells you THAT you’re leaking; only the step-to-step drops tell you WHERE.
The leak map — same total, different leak
Each transition is a pipe; the red wedge is the crack, sized to its drop %. The biggest crack pulses red. Hover a step for the counts; click a step (or use the buttons) for its likely cause + fix.
The leak map is interactive in the HTML version. Shop A is the real funnel (3,675 → 1,529 → 93 → 37 → 32 → 10; biggest crack: view → cart at −93.9%). Shop B is a constructed funnel with the identical 0.27% end-to-end (3,675 → 2,940 → 735 → 22 → 18 → 10) whose biggest crack is cart → checkout at −97.0%. Same total, opposite diagnosis — A needs product-page work, B needs checkout work. Clicking any step shows the cause-and-fix mapping from section 04.
The end-to-end rate says that you leak. Only the step-to-step drops say where — and “where” is the only thing you can fix.
What each leak means — the diagnosis table
Each step leaks for a different reason — so each drop points at a different part of your business. Match the step to the cause, and the cause to the fix.
| Step that drops hard | Here | Likely cause | The fix |
|---|---|---|---|
| VISIT → VIEW PRODUCT | −58.4% | Wrong AUDIENCE or ad-to-landing MISMATCH — the ad promised something the page doesn’t deliver, or you bought traffic that doesn’t want it. | Targeting + message-match + landing relevance. |
| VIEW → ADD TO CART | −93.9% · biggest | PRODUCT / PRICE / PAGE persuasion — they looked and didn’t want it at that price or presentation. | Offer, pricing, product page, social proof, urgency. |
| CART → CHECKOUT | −60.2% | FRICTION / TRUST / surprise costs — they wanted it, then the process got in the way. | Fewer steps, guest checkout, show shipping early, trust badges. |
| CHECKOUT → PURCHASE | −68.8% shared → purchased | PAYMENT / final-price shock / confidence at the last click. | More payment options, security cues, no last-second fees. |
Not every drop is a disease. The checkout → shared-details step here loses only −13.5% — that’s a healthy step. Leave healthy steps alone; every hour spent there is an hour not spent on the −93.9% crack.
Fix ONE step at a time — the method
Chasing every drop at once tells you nothing — if three things changed and the funnel moved, you don’t know which one did it. The method is deliberately slow, and that’s why it works.
Find the biggest step-to-step drop
Not the end-to-end rate — the single worst transition. In this funnel: view → cart at −93.9%.
Form ONE hypothesis for why
Use the diagnosis table. “The price feels high without social proof” is a hypothesis; “the page needs work” is not.
Change ONE thing
One change per test — a review block, a price anchor, a better photo. Two changes at once means zero learnings.
Watch that step’s rate for a week
Not the end-to-end rate — the one step you touched. A week smooths the day-to-day noise on small funnels like this one.
Only then move to the next-biggest leak
Plugged or not, write down what happened and re-rank the drops. Then repeat on the new biggest crack.
One leak, one fix, one week. Then the next.
Why this beats buying more traffic
Doubling traffic doubles the cost and keeps the same leak rate — twice the water through the same crack. Halving your biggest drop can double customers for FREE: same spend, same ads, more sales.
The worked line for this funnel: improving view → cart from 6.1% toward 12% roughly doubles carts — and everything downstream doubles with them — with zero extra ad spend. Drag the slider and watch it happen.
The fix calculator — plug the view → cart crack
Same 1,529 product viewers, same downstream rates as the real funnel — only the view → cart rate moves. Keyboard: Tab to the slider, ←/→ to adjust.
The calculator is interactive in the HTML version: a slider moves the view → cart rate from 3% to 15% (default: the real 6.1%) and recomputes the funnel with the real downstream rates. At 12%, the 1,529 viewers produce ≈183 carts → ≈73 checkouts → ≈20 purchases — roughly double the 10 baseline purchases at the SAME ad spend. The reference strip shows the alternative: doubling traffic to 7,350 visits also yields ≈20 purchases, but at 2× the cost, because the leak rate is unchanged.
The fix — four habits, in order
Map the drops, attack the worst one, test like a scientist — and re-map, because the biggest leak moves as you plug them.
Look at STEP-TO-STEP drops, not just the end-to-end rate
The end-to-end number hides the diagnosis — two funnels with the same 0.27% can need opposite fixes (section 03).
Attack the single biggest drop first
Here that’s view → cart at −93.9%. A gain at the widest crack multiplies through every step below it.
One hypothesis, one change, one week
Section 05 is the method. Slow is deliberate — it’s the only way to know what actually worked.
Re-map after each fix
The biggest leak moves as you plug them. Plug view → cart and cart → checkout may become the new widest crack — re-rank before you spend another week.
Where Pulsar fits: it draws this behavioral funnel for your pixel, counts real PEOPLE (not sessions) at each stage, and shows the step-to-step drops so the biggest leak is obvious.
Cross-reference — No. 02 — Quantity vs. Quality introduced this funnel and the quantity-vs-quality read of it; No. 01 — Efficiency & Fatigue shows what a leak costs you in CPA.