How to build a paid media strategy that survives contact with a budget.

What a paid media strategy actually is

Most documents called a paid media strategy are really a media schedule with a mission statement stapled to the front. They name the platforms, allocate a percentage to each, list some audiences, and set a target return on ad spend that nobody can trace back to a source. None of that is a strategy, because none of it can be wrong. A strategy makes claims specific enough to be falsified by the next quarter's numbers.

Strip it back and there are only four decisions worth arguing about:

1. What is a customer worth — in gross margin, over the horizon you are actually prepared to wait for.
2. What can you afford to pay to acquire one — the ceiling every channel is judged against.
3. Which channels can plausibly deliver at that price, and at what volume before they saturate.
4. How will you know — one measurement framework, agreed before launch, that every channel reports into.

Everything else, including the part most people think of as the work, is implementation. Ad copy, audience segments, bidding strategy and creative formats all matter enormously, but they are answers to questions the four decisions above have already framed. Get the frame wrong and excellent execution takes you somewhere you did not want to go, efficiently.

Start at gross margin, not at the platform

The single most common failure in paid media planning is importing a target from outside the business. A 4x return on ad spend sounds respectable and means nothing on its own. At a 70 percent gross margin it is comfortably profitable. At 22 percent it loses money on every order, and the harder the account works the faster the money goes.

So the first number to establish is not a marketing number at all. It is contribution margin per customer: revenue, less cost of goods, less the variable costs that only exist because the sale happened. Payment processing. Shipping and packaging. Returns and chargebacks at your actual rate rather than the rate you would prefer. For services, the delivery cost of fulfilling the work. For subscriptions, the servicing cost across the months you are willing to count.

Pick a horizon you will actually live with

Lifetime value is where planning goes to die. Twelve-month or twenty-four-month projections are usually built from a cohort that has not finished existing yet, and they conveniently justify whatever spend was already planned. They are not wrong in principle, but they are unfalsifiable in the timeframe that matters.

A more useful discipline is to pick a horizon that matches your cash position and stick to it. If you can fund acquisition for ninety days before the customer has to have paid for themselves, plan on ninety-day contribution margin. You can always run a second, longer model alongside it, but the ninety-day figure is the one that determines whether you can keep buying.

Set an allowable cost per acquisition

Allowable cost per acquisition is the ceiling. It is what you are willing to pay to acquire a customer while still hitting the profit you need from the programme. Once it exists, every channel, campaign and bid strategy has something concrete to be measured against, and most arguments about performance resolve themselves.

The calculation is not complicated. Getting the inputs honest is the hard part.

# Illustrative only. Substitute your own figures.
# All amounts AUD, per acquired customer, 90-day horizon.

Average order value                     $420
Orders per customer in 90 days          1.4
90-day revenue per customer             $588

Cost of goods            (54%)         -$318
Payment processing       (1.8%)         -$11
Shipping and packaging                  -$22
Returns allowance        (6%)           -$35
Contribution margin                     $202

# Decide what share of that margin the programme keeps
# as profit, and what share it may spend on acquisition.

Target profit share      (35%)          -$71
Allowable CAC                           $131

# Converting to a platform-facing target:
Break-even ROAS   = 588 / 202          = 2.91
Target ROAS       = 588 / 131          = 4.49

Two things fall out of that worked example immediately. The first is that break-even sits at 2.91, not at 1.0, which is the number most dashboards imply. The second is that the gap between break-even and target is the entire negotiating range for the programme. Anything above 4.49 is over-delivering and probably under-spending. Anything between 2.91 and 4.49 is profitable but below plan. Below 2.91 the account is selling money.

New customers versus everyone

If a meaningful share of your revenue comes from people who would have bought anyway, a blended allowable cost per acquisition will flatter you. Run the calculation a second time on new customers only, using new-customer revenue and new-customer contribution margin. The two numbers together tell you something neither does alone: how much of your reported efficiency is acquisition and how much is retention wearing an acquisition costume.

Separate demand capture from demand creation

Paid media does two fundamentally different jobs, and almost all of the confusion in cross-channel reporting comes from judging them with the same yardstick.

Demand capture meets people who have already decided they want something and are looking for it. Branded and non-branded search, shopping, comparison surfaces and retargeting all sit here. It is efficient, it is measurable, and it is finite. You cannot capture more demand than exists, and once you are taking most of the available impression share, additional budget buys progressively worse traffic at progressively higher prices.

Demand creation puts you in front of people who were not looking. Paid social prospecting, video, display and audio sit here. It is less efficient per click by design, its effects show up on a lag, and a meaningful portion of its value will land in channels it never gets credited for.

The practical consequence is that these two should never share one efficiency target. Hold demand creation to the same cost per acquisition as branded search and you will switch it off within a quarter, then watch capture volumes decline six months later without connecting the two events. The standard pattern in accounts that have done this is a search programme with excellent reported efficiency and flat total revenue for two years.

Choose channels against saturation, not fashion

Every additional channel carries a fixed cost that has nothing to do with media spend: creative production in a new format, a new measurement integration, a new set of platform quirks to learn, and a share of the attention of whoever is managing the programme. Below a certain spend, that fixed cost exceeds anything the channel can return, and the channel never gathers enough conversion data for automated bidding to leave the learning phase.

A reasonable sequence for most Australian businesses, in order of what has to be true before you add the next thing:

• Branded search first, and understood honestly. It is the cheapest revenue you will ever buy and the most misleading line in the account. Know what it costs and never let it sit inside a blended target.
• Non-brand demand capture next. Search and Shopping against the terms that describe what you sell. Take this to a genuine impression-share ceiling before adding anything else.
• Then one demand-creation channel, chosen for where your audience actually is rather than where the case studies are. One, funded properly, for long enough to produce an answer.
• Then a second, only once the first has either found its ceiling or failed for a reason you can articulate.

Channel selection questions worth answering in writing, before any budget is committed:

• Is the audience reachable here at scale? A B2B product with 4,000 possible buyers in Australia does not need a channel optimised for reach; it needs precision, and probably LinkedIn.
• Can we make the creative this channel demands, repeatedly? TikTok and YouTube consume creative. A single hero asset is not a channel strategy, it is one week of one.
• Will this channel clear the minimum viable spend? If the honest budget will not sustain roughly 30 to 50 conversions a month, automated bidding will not stabilise and you will be reading noise.
• How will we attribute it, given it will be undercredited? Decide this before launch, not during the review where someone proposes cutting it.

The channel pages cover what each one is genuinely good at: Google Ads for capture, Meta for scaled prospecting, YouTube for demand creation with measurable lift, ChatGPT Ads as an early-stage surface worth a test budget rather than a plan.

Splitting the budget

Once the channels are chosen, the split follows a simple principle: fund what is proven to its ceiling, fund what is promising to a level that can produce evidence, and reserve a small amount for things that will mostly fail.

In practice that lands close to a 70/20/10 shape. Seventy percent to channels and campaign types with a track record at or inside your allowable cost per acquisition. Twenty percent to things working but not yet at scale, or at scale but not yet efficient. Ten percent to genuine experiments, with the expectation that most produce nothing except an answer.

Two rules make the shape hold. First, the ten percent is spent whether or not anyone feels like it; an experimental budget that is quietly absorbed into the proven channel every month is not a budget, it is a rounding error. Second, no channel receives less than its minimum viable spend, even if that means running fewer channels than the plan called for. Three channels funded properly beat six funded symbolically, every time.

The mechanics of that calculation, including how to work out a floor for each channel and when to move money rather than add it, are covered in full in the paid media budget guide.

Agree the measurement framework before launch

The measurement plan is part of the strategy, not a reporting task to be sorted out later. Agreeing it after launch guarantees that the first performance conversation becomes an argument about whose numbers are right, and those arguments are never won, only survived.

Three layers, each answering a different question:

• Platform reporting answers "what is happening inside this channel". It is the right tool for optimisation decisions and the wrong tool for allocation decisions, because every platform claims the same conversions.
• Blended reporting answers "is the total working". Total media spend against total new-customer revenue, from your own systems, on a fixed cadence. Crude, unglamorous, and the only number that cannot be gamed by an attribution setting.
• Incrementality testing answers "would this have happened anyway". Geo holdouts, conversion lift studies and clean on/off tests, run on the channels where the question actually matters.

Write down, in the strategy document, which number decides what. A useful default: platform numbers for within-channel optimisation, blended numbers for between-channel allocation, and incrementality for the periodic decision about whether a channel deserves to exist. The measurement guide sets out how to build that stack and reconcile it to what the business actually banked.

Build the testing cadence in

A strategy without a testing programme decays. Creative fatigues, auction dynamics shift, competitors enter, and the thing that worked last year becomes the thing everyone is doing. The account needs a steady supply of new attempts, and a rule for how each one is judged.

Three practical constraints that stop testing programmes from producing noise:

• One variable at a time, at sufficient volume. If a test cannot reach roughly 100 conversions per arm, the result is a coin flip with a chart attached.
• A pre-registered decision rule. Write down before launch what result means adopt, what means kill, and how long you will wait. Deciding afterwards is how every test comes back positive.
• A log that survives staff turnover. Date, hypothesis, what changed, what happened, what was decided. Most accounts re-run the same failed test every eighteen months because nobody wrote the first one down.

The one-page plan

If the strategy does not fit on a page, it is a wish list. Here is the shape that survives a board meeting, because every line is either a number or a named decision.

OBJECTIVE
  New customers per month          [ target ]
  At or below allowable CAC of     [ $ ]
  Horizon used for margin          [ 90 days ]

ECONOMICS
  Contribution margin per customer [ $ ]
  Break-even ROAS                  [ x ]
  Target ROAS                      [ x ]

CHANNELS          role         monthly     target CAC   floor
  Search, brand    capture      [ $ ]      [ $ ]        [ $ ]
  Search, non-brand capture     [ $ ]      [ $ ]        [ $ ]
  Paid social      creation     [ $ ]      [ $ ]        [ $ ]
  Experiment       test         [ $ ]      n/a          [ $ ]

MEASUREMENT
  Optimisation decisions by        [ platform reporting ]
  Allocation decisions by          [ blended new-customer CAC ]
  Existence decisions by           [ incrementality test ]
  Reconciled against               [ source of truth ]

REVIEW
  Weekly    pacing and anomalies
  Monthly   allocation between channels
  Quarterly strategy, allowable CAC, channel set

THIS IS WRONG IF
  [ the condition that would make you abandon it ]

The last block is the one people skip and the one that makes the document a strategy. Naming the failure condition in advance is what separates a plan from a hope, and it is remarkable how often writing that line changes the plan above it.

Frequently asked questions

Where this fits

A strategy is a set of decisions you can be held to, written down so that next quarter's numbers can prove you wrong. Four decisions, one page, one failure condition. Everything else is execution, and execution is much easier to fix than a plan that was never really made.

If you would rather have a senior set of eyes on it, paid media consulting engagements open with exactly this process, and the free audit gives you the findings and the three highest-impact fixes whether or not we end up working together.