Four accounts.
Real numbers.

The client

Demanding, and reasonably so. They wanted results from day one and were benchmarking against their pre-COVID numbers — a bar set in a different market, with different competition and different costs. For the first stretch we were fielding complaints roughly every week.

What we did not do

We did not chase the complaint. Loosening match types and buying cheaper clicks would have made the lead count jump and the cost per lead fall on paper, and it would have filled their phones with people who were never going to book. Instead the account was rebuilt around one question: what does a real first-time customer look like, and how do we buy more of exactly that?

How the leads are counted — this is the part that matters

Every lead is tracked in CallRail. A conversion is only counted for a first-time caller or a form submission. Repeat customers and abandoned calls are excluded, and only calls a human has marked as qualified get imported back into Google Ads as conversions.

That is why the numbers below do not agree, and why they should not: CallRail logged 539 contacts from Google Ads over the period — 480 calls and 59 forms — while Google Ads counts 467 conversions. Thirteen percent of the raw contact volume is deliberately not counted. Most reporting works the other way.

Where it stands

Cost per lead has halved. The weekly complaints stopped. The account now runs on a mix of Performance Max and Search, with 89% of all leads arriving as phone calls — which is what a bee removal business actually looks like, and why the whole account is built around the phone rather than the form.

What is next

Offline conversion imports are being implemented now, so the account will optimise against booked revenue rather than lead count. That is the step that changes the question from "what did a lead cost?" to "what did a customer earn?", and it is where this account goes from working to compounding.

The situation

A Portuguese port house with real heritage at home and effectively zero brand awareness in the United States, entering the market recently on about $2,300 a month across two campaigns. No US search history to bid against, no existing customer base to remarket to, and a product most American buyers think of once a year.

Why this category is harder than it looks

Alcohol is restricted content on Google. Under Google's alcohol policy, ads for the online sale of alcohol may only run in approved locations, targeting anyone below the legal drinking age is prohibited, and several formats are simply unavailable — dynamic search ads, Gmail ads, Discovery ads, app install ads and others are all off the table. On top of the platform rules sit the American state-by-state direct-to-consumer shipping laws, which decide where a bottle can legally be sent at all.

The practical consequence is that a lot of the usual growth levers do not exist here. What is left is search intent, shopping feed quality, and knowing exactly which geography can actually receive an order.

Read the dates before the numbers

The earlier window runs 3 November to 31 March. It contains Christmas — and port is a gifting product whose year is weighted heavily toward the festive season. The later window, 1 April to 27 August, is the off-season.

So the comparison is not flattering to the newer period; it is the opposite. Revenue more than doubled in port's quietest five months, measured against a period containing its busiest. A like-for-like seasonal comparison would show a wider gap, not a narrower one.

Where it stands

The account crossed 1.0×. At 0.64× every dollar of ad spend returned 64 cents and paid traffic was subsidised by the rest of the business. At 1.49× it pays for itself and contributes. That is the threshold that matters far more than the percentage improvement either side of it.

A genuine before-and-after, matched window to window

The account compares 14 May – 10 September 2026 against the preceding window of equal length. Leads (tracked conversions) rose from about 21 to 55.25 — up 34.26 — while the cost per lead fell R22.21, from about R64.48 to R42.27, even though spend went up. More leads and a lower cost per lead at the same time is the combination that matters. The tiles reconcile: R2,340 of spend over 55.25 leads is R42.35 each, against the R42.27 shown.

How the cost came down

The same cleanup discipline we apply to every account: conversion tracking confirmed and trustworthy first, then tighter campaign settings and targeting, a full search-terms and keyword clean-up to cut the searches that were never going to buy, and reworked ad copy on the terms that were. Nothing exotic — just the account being run properly, week to week, instead of left on Google’s defaults.

The honest caveat

This is a small budget — about R2,340 over roughly four months — so it is a cost-efficiency and lead-volume result, not a scale-of-spend one, and we present it that way. There is no revenue tracking on this account, so we report leads and cost per lead, not a return.

What the curve shows

This is the South African arm of a practice with offices in both the UK and South Africa. Over 1 March – 10 September 2026 the account produced 71 leads at a blended R278 on R19,747 of spend. The line to watch is the cost-per-lead curve: it starts near R600, spikes to roughly R1,200, then falls to a sustained band in the low hundreds across the back half of the window — more than halved from its peak.

Being straight about this account: it shows the whole period, not a matched before-and-after, so we quote the blended R278 and point you at the curve rather than claim a single before/after percentage. A matched-window view would pin the exact figure — and we would rather under-claim than publish a number this view cannot back.

How the cost came down

Conversion tracking was sorted out first — nothing else can be trusted until it is. Then the structural fixes: Search Partners switched off, location targeting set to people regularly in the area rather than merely interested in it, and the campaign settings tidied. After that, a full search-terms clean-up and keyword pruning, a set of higher-intent keywords added, and the ad copy reworked to match. That sequence is what turned a rising cost per lead into a falling one.

The honest caveat

A modest budget — about R3,300 a month — and no revenue tracking on this account, so we report leads and cost per lead, not a return. The before/after is read from the trend line rather than a matched-window tile, as said above.