What does $10,000 in advertising actually buy an Inner West business?

The short answer: not all of it is media

When an owner says “I spent $10,000 on ads”, the number usually covers more than media. It can include GST, the cost of making the ads, the cost of making sure the website and tracking work, and a management fee. What is left reaches the auction, and part of that first buys information, not customers.

This guide walks through that split with three modelled allocations. Every figure in them is illustrative: it shows the mechanism with arithmetic you can check, and it is not a benchmark, a forecast or a promise. Where we need a cost per click or a conversion rate, we state it as an assumption, because we have no verified Inner West figure to use. Replace our assumptions with your own numbers.

Spend buys access to attention. It does not guarantee memory, demand or sales.

Where $10,000 goes in practice

Treat $10,000 as the total cash leaving the bank over three months, including GST. Six lines take a share of it.

• GST. Google says all Google Ads sales in Australia are subject to 10% GST for accounts with Australian business addresses. If we assume every line of the $10,000 is GST-inclusive, GST is one-eleventh of the total: $10,000 ÷ 11 = $909, leaving $9,091 to allocate. If you are registered for GST and meet the ATO’s conditions, you may be able to claim that $909 back as a GST credit, but it still leaves the bank first. Whether it returns depends on your registration and what you sell, so ask your accountant. See the budgeting guide for the ex or inc decision.
• Creative. Ad copy, photography, short video, and the words on the landing page. This is the part a person sees, so it is rarely the place to cut.
• Tracking and landing page. Conversion tracking for calls, forms and bookings, plus a page built for the service being advertised. If Google Ads cannot see which clicks became enquiries, its automated bidding has nothing to learn from.
• Management. An agency fee, or the owner’s own hours if you run it yourself. Owner hours are a real cost even when no invoice arrives.
• Platform media. What you pay Google, Meta or another platform for clicks and impressions.
• Learning spend. Part of the media pays for tests that do not work: keywords that attract the wrong searches, audiences that click and never enquire. We show it as an assumed share of media.

Three illustrative allocations

The three businesses below are composites we made up to show the arithmetic. They are not clients. Each starts with $10,000 including GST: $909 GST and $9,091 to allocate. Media is the remainder after the other lines.

The point is the last row. In each case between a third and a half of the $10,000 reaches the auction as media that is working, and the rest pays for GST, preparation, the tests, and the people doing the work. That is not a scandal. Tracking and landing pages are what make the media readable. The question is whether anyone told you.

What the working media might produce (assumptions stated)

Marrickville Road dentist. Assume clicks cost $5, 1 visitor in 12 enquires, 1 enquiry in 3 becomes a new patient, and a new patient is worth $700 in first-year gross profit. Working media of $3,704 buys 3,704 ÷ $5 = 741 clicks, which gives 741 ÷ 12 = about 62 enquiries and about 21 new patients, or 21 × $700 = $14,700 of first-year gross profit against a $10,000 outlay. Now halve the click-to-enquiry rate to 1 in 24, which is an easy thing for a slow landing page to do. That gives about 31 enquiries, about 10 patients and $7,000. The same spend moves from a gain to a loss on one assumption. For Ahpra rules that shape what the ads can say, see the dentist guide.

Leichhardt renovation builder. Assume clicks cost $9, 1 visitor in 40 enquires, 1 enquiry in 10 becomes a signed job, and a job returns $18,000 gross profit. Working media of $3,309 buys 3,309 ÷ $9 = about 368 clicks, which gives about 9 enquiries and about 0.9 signed jobs. In real life that means zero or one. One job returns $18,000 against $10,000. Zero returns nothing. Nine enquiries are too few to tell a good campaign from a lucky one, which is why a builder should read enquiry quality, not signed jobs, at three months. See the renovation advertising guide.

Newtown boutique. Assume clicks cost $1.50, 1 visitor in 60 buys online, the average order is $85 and gross margin is 50%. Working media of $5,104 buys 5,104 ÷ $1.50 = about 3,403 clicks, which gives about 57 orders, $4,845 of tracked online sales (57 × $85) and $2,423 of gross profit (57 × $42.50). Against $10,000, that is a tracked return of $0.48 in sales per dollar. It does not mean the boutique failed: the ads may bring shoppers into the shop and repeat purchases that the click data cannot see. It does mean the case rests on something other than the tracked sales. See why a good return on ad spend can sit beside bad growth.

What media buys, and what it does not

Revenue distance: how far each number sits from money

Every number in an ad report sits at some distance from money in the bank. The further away it is, the easier it is to look good while the phone stays quiet. We call this revenue distance, and it is a useful way to read any dashboard.

Platform settings can move a metric closer to flattering. Google Ads separates primary conversion actions, which are reported in the Conversions column and used for bidding, from secondary ones, which are for observation only. It also lets you count one conversion per click or every conversion, and Google recommends counting one for leads. If your account counts every phone call, including a ten-second wrong number, as a conversion, the report moves closer to flattering and further from money. See how to check your tracking.

The practical rule: ask for the metric nearest money that is measurable for your business, and ask who measures it. For a dentist that is new patients who attended. For a builder it is qualified quotes, then signed jobs. For a boutique it is orders, then repeat customers.

The three-month learning view

We frame a first $10,000 as a three-month test, not a one-month campaign. The costs are front-loaded: tracking, landing page and creative are paid in the first weeks, while media ramps up. Judging after four days, or judging month one against the whole $10,000, is the most common way to stop a campaign before it has a fair test. Google’s automated bidding needs data to learn from, so allow a few weeks before drawing conclusions.

1. Month one: set up and prove the tracking. Place a test call and a test form, and confirm each one appears in the reports. Review the search terms weekly and add negatives. Do not judge performance yet.
2. Month two: prune and focus. Remove what attracted the wrong people, move money to what produced real enquiries, and read the enquiries themselves. Were they the customers you want?
3. Month three: read the result. Compare cost per enquiry and cost per booked job or paid customer against your ceiling.

What a readable result looks like

• The tracking was tested, and the reported enquiries match the calls and forms you received.
• Enough enquiries to read, roughly 10 a month or more. Fewer than that, and one week can swing the picture.
• Each enquiry is logged with what happened next: booked, quoted, paid or lost.
• Cost per enquiry has settled within a range for several weeks, rather than swinging wildly.
• You can name two changes you would make with the next dollar and the evidence for each.

The ten-thousand-dollar test

Run these six questions on any proposal an agency hands you, including ours. A good proposal answers them in writing without hesitation.

1. Where does each dollar go? Ask for the full $10,000 split line by line: media, fees, creative, tracking, GST. If they cannot split it, they do not know.
2. What share reaches the auction? Divide media by the total. In our three examples it is 51% to 73%. A share well below half needs a reason, and so does a share close to 100% (who sets up the tracking?).
3. What will we know in 90 days that we do not know today? The answer should be a question about your customers, not “we will optimise”.
4. Which metric is nearest money, and who measures it? If the headline number is impressions or clicks, ask for the next step along the chain.
5. What is the stop or continue rule? Agree in advance what result would make you spend more, the same or less, and the date you will decide.
6. Who owns the accounts and the data? The ad accounts, the tracking and the audiences should sit in your name, with the agency as a user.

A proposal that passes five or six of these is worth a conversation. One that passes two or fewer is asking you to trust it.

How AI-assisted checking changes this: an agent can match a proposal’s line items against an invoice and the platform’s own spend export in minutes and flag anything that does not reconcile. A person decides whether the explanation is reasonable.

When this analysis is wrong

• The assumptions are ours. We have no verified Inner West cost per click or conversion rate, so the ones above are placeholders. Real figures can be several times better or worse, and the answer moves with them.
• The splits are not typical. The fee, creative and learning shares were chosen to show the mechanism. They are not an industry average.
• Some businesses spend much less on fees. An owner-run account removes the fee but not the time it takes.
• Three months is a convention. A builder with a six-month sales cycle may not see paid jobs by month three.
• Ads can have effects the reports miss. Walk-in sales, word of mouth and later purchases are hard to attribute, and our revenue-distance view undercounts them.

What we would do first this week

• Write your own $10,000 split on one page using the six lines above, including GST.
• Work out media as a share of the total and the working media after an assumed learning share.
• Pick the metric nearest money that you can measure, and find out where it is recorded today.
• Place a test call and a test form enquiry and check where each one shows up.
• Write the stop or continue rule and a review date three months out.
• Run the six-question test on the proposal you have, or the one you plan to request.

Our research and sample report show what a read-only review looks like, and you can request an audit here if that is useful. The sample uses invented figures.