Everything we actually do to run performance advertising without wasting budget: how auctions work, the CAC math that keeps campaigns viable, channel selection, account structure, creative testing, landing pages, attribution in a post-cookie world, measurement, scaling, and the mistakes that burn the most money. No fluff, no withheld methodology.
This guide is long on purpose. Paid media is sold as a lever you pull, and the complexity is obscured so agencies can charge for managing it without explaining it. What follows is the opposite: the actual mechanics, the order we work through them, and the numbers we hold ourselves to. Read it end to end and you will understand the engagement before we ever speak.
Every ad platform runs an auction. When a user performs a search, opens an app, or loads a page, the platform runs a real-time calculation to decide which ads to show and in what order. Understanding how that calculation works determines whether you win cheaply or overpay for clicks that do not convert.
Google Ads does not sell positions to the highest bidder. It ranks ads by a combination of your bid and your Quality Score, which reflects expected click-through rate, ad relevance to the query, and landing page experience. The formula matters because it means a tightly written ad that earns clicks cheapens your effective cost per click. A well-matched ad group can outrank a higher bid from a poorly organised competitor.
Ad Rank is calculated as: bid multiplied by Quality Score multiplied by expected impact of extensions and other ad formats. Your actual cost per click is set just above what the next advertiser would need to pay to rank above you, not the maximum you bid. This is why two advertisers can have the same bid but pay very different CPCs: their Quality Scores differ.
Meta, LinkedIn, and similar platforms run relevance-based auctions too, but the signal set is different. Instead of keyword matching, they use audience signals: demographics, interests, behavioural data, and, critically, how your ad compares to others competing for the same person's attention at the same moment. Meta calls this "estimated action rate" and it is baked into the auction outcome invisibly.
On Meta, an ad that earns engagement (clicks, saves, comments that indicate genuine interest) gets cheaper delivery over time. An ad that earns low engagement or high negative feedback (hidden, reported) gets more expensive. The algorithm is trying to show people things they will not resent. Your creative quality is not just a branding consideration: it is a cost lever.
The core implication. Both auction types reward relevance. On search, relevance is keyword and landing page alignment. On social, relevance is creative quality and audience fit. Getting those right lowers your CPCs without changing your bids.
Cost per acquisition is the number that determines whether paid media is worth running at all. If you spend more to acquire a customer than that customer returns over their lifetime with you, the channel is destroying value regardless of how good the click-through rate looks.
The basic model is straightforward: CAC = total ad spend divided by total customers acquired. The useful version goes further and breaks that into the cost chain:
| Metric | What it measures | Where it is lost |
|---|---|---|
| Cost per click (CPC) | What you pay for each visit | Weak Quality Score, broad match, low relevance |
| Landing page conversion rate | Visits that become leads or trials | Mismatched message, slow load, poor form UX |
| Cost per lead (CPL) | CPC divided by conversion rate | Either of the above |
| Lead to customer rate | Leads that close to paying customers | Audience quality, offer fit, sales process |
| CAC | CPL divided by close rate | Any stage above |
| LTV to CAC ratio | Whether the channel is profitable | Pricing, churn, product |
Working backwards from your target CAC tells you what conversion rate you need at each stage to justify the channel. If your target CAC is 200 and your average lead closes at 20 percent, your target CPL is 40. If your current CPC is 4, you need a landing page converting at 10 percent or better. If it converts at 3 percent, you are already 3x over your target cost per lead before a single close rate assumption.
Most campaigns that underperform do so because this chain was never calculated before the campaigns went live. Budget is allocated, campaigns run, and the post-mortem asks why CAC is too high rather than what conversion rate was assumed at each stage. We build the target model before touching a campaign.
Every engagement opens with an audit of the existing account, and we give it away because it is the only honest way to scope the work. We pull 90 days of spend data and answer a specific question: where is money going that it should not, and what is each leak worth in monthly wasted spend?
The audit covers five areas:
The output is a ranked list of issues. Every item has the spend attached to it, an estimate of the monthly budget that would be freed or reallocated by fixing it, and the specific action required. You see it before any money changes hands, and you keep it either way.
Channel selection is a business decision, not a media planning one. The right channel is determined by where your buyers are in the decision process when your ad reaches them, what they are doing when it appears, and what the CAC math supports at the conversion rates realistically achievable on that channel.
Search captures in-market demand: people actively looking for a solution right now. The conversion rates are higher than social channels because the intent is explicit. The downside is that search volume is finite, you can only capture demand that exists, you cannot create it. For B2B products with long sales cycles and low monthly search volume, the ceiling on search spend is low. For e-commerce, SaaS tools with clear search queries, and services with defined buying intent, it is typically the first channel to prove.
Meta reaches people who are not actively searching but can be interrupted. The format works best for products with visual appeal, clear value propositions that land in a few seconds, and audiences that can be defined by demographic or behavioural signals. B2C conversion rates from Meta cold audiences are typically lower than search, but the reach is vastly larger. The key is matching creative to the stage of awareness: cold prospecting creative needs to explain why anyone should care; warm retargeting creative can assume awareness and push toward conversion.
LinkedIn is expensive per click by design. The platform charges a premium for professional audience targeting because the signals are more accurate than Meta's inferred demographics. For B2B products where the buyer is a director or above in a specific function or industry, LinkedIn can produce qualified pipeline at a CAC that makes sense even at 3 to 5x the CPC of Google. For anything outside that use case, the premium is rarely justified. We will tell you plainly if LinkedIn is not the right fit for your ICP and budget.
Retargeting is not a channel in the same sense as search or social: it is a layer that runs on top of them. The audience is people who have already visited the site, viewed a product, started a checkout, or engaged with an ad. These audiences convert at 3 to 8x the rate of cold prospecting audiences. Retargeting budgets are typically small relative to prospecting but deliver disproportionate returns. Every account should have retargeting running before increasing prospecting spend.
| Channel | Intent signal | Best for | Common CAC range |
|---|---|---|---|
| Google Search | Active buying intent (keyword) | In-market demand capture, SaaS, services, e-commerce | Low to medium if Quality Score is strong |
| Google Performance Max | Mixed (search, display, YouTube, shopping) | E-commerce, broad reach at scale | Varies widely; needs tight conversion tracking to optimise |
| Meta prospecting | Interruption (audience signal) | B2C products, visual brands, broad consumer offers | Medium to high; lower with strong creative |
| Professional context | B2B targeting by role, function, or seniority | High CPC; justified for high-LTV B2B products | |
| Retargeting (all channels) | Prior site or ad engagement | Every account, layered on top of prospecting | Lowest CAC across all channels when audience is warm |
Account structure is not a housekeeping detail. It is the mechanism by which you control where your budget goes, how accurately the algorithms can optimise, and how clearly you can read the data. A poorly structured account produces data that cannot be trusted and campaigns that cannot be improved.
Brand keywords (searches for your company name, your product name, your branded terms) and non-brand keywords (searches for the category, the problem, the competitor) should never share a campaign. They have different economics, different Quality Scores, different competitive dynamics, and different strategic purposes. Mixing them inflates the apparent performance of non-brand by averaging in the high conversion rates of brand queries. We separate them on day one.
The right level of granularity is determined by how different the landing pages and ads need to be. Ad groups should contain keywords that justify the same ad copy and the same landing page. If two keywords need different messages to be relevant, they belong in different ad groups. Too many ad groups and the data fragments below statistical significance. Too few and the ads are irrelevant to half the queries in the group.
On search, a negative keyword list is not optional. Without one, broad and phrase match keywords match queries that have nothing to do with your product. We build a negative keyword list from the search terms report on day one, add a shared library of brand-safety negatives (free, jobs, careers, DIY, tutorial, how to), and review the search terms report weekly during the first month to catch new leakage as it appears.
The structure principle. Budget should follow performance, not distribute evenly. Structure the account so you can see which campaign, which ad group, and which ad is driving results, and move budget toward the winner. If you cannot isolate performance to a level where you can act on it, the structure is too flat.
On search, targeting is implicit: the keyword defines who sees the ad. On social, targeting is explicit: you choose who the platform shows the ad to. The precision of that choice determines how much of your budget reaches people likely to buy, and how much reaches people who will never convert.
First-party data, your own customer list, CRM contacts, and site visitors, produces the most reliable audiences because the signal comes from actual behaviour with your product. Upload customer lists to Meta and Google to create matched audiences, then use them as seeds for lookalike modeling. Site visitor segments (all visitors, product page visitors, checkout starters, past purchasers) form the foundation of any retargeting program.
Meta's interest targeting is broad by default and gets broader over time as the platform optimises for delivery. Stacking multiple interests in a single ad set (AND logic) restricts the audience more than the platform intends and often performs worse than single-interest targeting at a larger scale. We test interest segments in separate ad sets so the data is readable and the algorithm has enough room to find buyers within each.
Lookalike audiences ask the platform to find people statistically similar to a seed group you supply. The quality of the lookalike is entirely determined by the quality of the seed: a lookalike built from your top 100 customers by LTV will outperform one built from all site visitors, because the seed is more signal-dense. We always build multiple lookalike percentage ranges (1 percent for precision, 3 to 5 percent for scale) and test them before allocating significant budget.
Creative is the biggest variable in social advertising performance. Two identical targeting setups with different creative can produce CAC differences of 2x to 5x. Yet creative testing is consistently underfunded and poorly designed. Most accounts run two ads per ad group and call it a test. A real creative test starts with a hypothesis about why the challenger should outperform the control, runs long enough to reach statistical significance, and produces a learning rather than just a winner.
Not all creative elements are equal in their impact on performance. The elements that move results the most, in roughly decreasing order of impact, are:
Test one variable at a time where possible. On Meta, use the A/B test feature to isolate audiences from creative variables: if the audiences differ, you are not learning about creative. On Google, responsive search ads surface headline and description combinations algorithmically, which means individual element performance is readable but overall ad performance reflects the best combination, not any single variant.
Set a minimum test duration based on expected conversion volume, not calendar time. If an ad group gets thirty conversions a month, a two-week test produces roughly fifteen conversions per variant: not enough data to distinguish signal from noise at a 95 percent confidence level. Either run longer or accept a lower confidence threshold with the understanding that your conclusion is directional, not definitive.
The landing page is where ad spend either pays off or disappears. A conversion rate difference of 2 percent versus 6 percent on a landing page triples the cost per lead without changing a single element of the ad campaign. Yet landing pages receive far less attention than the ads that lead to them, because they feel like a product or design problem rather than a media problem. They are both.
The most reliable way to lift conversion rate with no other changes is to ensure the headline on the landing page mirrors the claim in the ad. A user who clicked because the ad said "cut timesheet admin to ten minutes" and arrives at a page headed "The leading workforce management platform" has to rebuild their mental model from scratch. The conversion rate drops because cognitive work increases and relevance decreases.
For campaigns with multiple messages being tested, consider dedicated landing pages per message, or at minimum dynamic headline insertion that pulls the ad's value proposition into the page header. The extra implementation cost pays back in conversion rate within weeks.
A landing page should have one purpose and one conversion action. Navigation to the rest of the site, links to blog posts, and secondary CTAs for other products all provide escape routes for people who are close to converting but not yet certain. The longer and more distracted the conversion path, the more people leave before completing it. For high-intent ad traffic, a stripped-back page outperforms a full site page nearly every time.
A landing page that takes more than three seconds to load on mobile loses roughly half its visitors before a single word is read. This is not a performance recommendation: it is a conversion rate calculation. If a page converts at 5 percent when it loads in one second and at 2.5 percent when it loads in four seconds, the slow page costs 50 percent more per lead. Google also uses landing page experience as a Quality Score signal, so a slow page raises your CPCs in addition to lowering your conversion rate.
Bidding strategy determines how the platform spends your budget: toward what objective, at what pace, and with what constraints. The default recommendations from every platform favour the platform's revenue, not yours. Understanding the tradeoffs lets you choose the strategy that fits your stage.
Manual CPC bidding gives you explicit control over how much you pay per click. It is predictable and transparent, but it does not react to auction dynamics in real time. It suits accounts with low conversion volume where automated strategies cannot learn fast enough to outperform a human setting bids based on historical data.
Automated bidding strategies (Target CPA, Target ROAS, Maximise Conversions) use machine learning to adjust bids in real time based on signals at the moment of each auction: device, location, time, audience membership, and historical conversion patterns for that context. They outperform manual bidding reliably when conversion volume is sufficient, typically thirty or more conversions per month per campaign as a minimum threshold.
The trap is deploying Target CPA before the algorithm has enough conversion data to set bids correctly. In the learning phase (the first two to four weeks after a major change, or on campaigns with less than 30 conversions per month), automated strategies can bid erratically and waste budget exploring the wrong corners of the auction. We keep accounts on manual or enhanced CPC until conversion volume justifies the switch.
Daily budget caps prevent overspend but do not guarantee even delivery. On Google, daily budgets can be exceeded by up to twice the daily amount on high-traffic days, compensated by under-delivery on other days (the monthly cap is enforced). On Meta, delivery is optimised toward your objective within the budget, which can mean front-loading spend on days when the algorithm predicts good audience availability.
For accounts with strict monthly budget limits, we use campaign-level monthly budget controls where available, monitor daily spend against the monthly target proactively, and set alerts for overspend thresholds rather than relying on platform-side enforcement.
Attribution is the practice of assigning credit for a conversion to the touchpoints that preceded it. It sounds administrative. It is actually the mechanism by which you decide whether a campaign is working, and therefore whether to spend more or less on it. Bad attribution means bad decisions.
Before any campaign goes live, every conversion event that matters to the business must fire correctly on the right action. For an e-commerce store, that is a purchase with the correct revenue value attached. For a SaaS product, it might be a free trial start, a demo booked, or a paid subscription event. For a lead generation business, it is a form submission or a phone call. Getting this wrong is more common than most advertisers realise: doubled conversion tags, thank-you page visits counted instead of actual form submissions, and mobile conversion events that never fire are regular findings in our audits.
Browser-side tracking fires conversion tags from the user's browser when they complete an action. It is straightforward to implement but is increasingly unreliable as ad blockers, browser privacy restrictions, and iOS tracking prevention (ATT) block or delay the signals before they reach the platform. Signal loss of 20 to 40 percent is common on platforms that rely heavily on browser-side events from Safari users.
Server-side tracking fires conversion signals from your server to the platform's API, bypassing browser-level blocking. Meta's Conversions API (CAPI) and Google's Enhanced Conversions are the two most material implementations. Server-side tracking recovers a significant share of the signal loss and improves optimisation because the algorithm gets more conversion data to learn from. It requires engineering time to implement properly, but the CAC improvement from better signal justifies it on any account spending above a few thousand a month.
An attribution window defines how far back before a conversion a touchpoint can receive credit. A 7-day click, 1-day view window (common on Meta) means that a click within the past week or an ad view within the past 24 hours is credited if a conversion happens. A 30-day click window is more appropriate for a B2B product with a 3-week consideration cycle.
Attribution models decide how to distribute credit when multiple touchpoints occur within the window. Last-click gives all credit to the final touchpoint before conversion. Data-driven (available on accounts with sufficient conversion volume) distributes credit based on which touchpoints statistically contributed to conversions. The choice of model changes which campaigns look profitable, which is why we align it to the actual sales cycle rather than leaving the platform default in place.
The post-cookie reality. Third-party cookies, which historically enabled cross-site tracking and audience building, are being deprecated or restricted across major browsers. Safari has blocked them for years. Chrome's deprecation is ongoing. First-party data, server-side signals, and consent-based tracking are the durable infrastructure. We build attribution setups that work in a first-party world from the start, rather than retrofitting them when the signal disappears.
Reporting that shows click-through rates and impressions alongside spend tells you nothing about whether the campaign is financially viable. The metrics that matter are the ones that connect ad spend to business outcomes.
| Metric | Formula | What it tells you |
|---|---|---|
| CAC (cost per acquisition) | Total spend divided by customers acquired | Whether the channel is viable at all |
| ROAS (return on ad spend) | Revenue from ads divided by ad spend | Revenue efficiency of the spend; most useful for e-commerce |
| Payback period | CAC divided by monthly gross margin per customer | How long until a customer pays back their acquisition cost |
| LTV to CAC ratio | Customer lifetime value divided by CAC | Long-run profitability of the acquisition; target above 3 for most SaaS |
| MER (marketing efficiency ratio) | Total revenue divided by total marketing spend | Blended efficiency across all channels, not attributable to any one |
| CPL (cost per lead) | Spend divided by leads generated | Efficiency of the top of the funnel; meaningful only paired with lead quality data |
Every platform over-reports its own ROAS because it attributes conversions that would have happened anyway. A customer who saw a Meta ad and then purchased three weeks later via a direct visit would be attributed to Meta in the last-click model, but also to the organic search visit that happened in between, and possibly to a Google Search ad clicked on the day of purchase. Triple-counting is normal across platforms with separate attribution systems.
Marketing efficiency ratio (MER) corrects for this by measuring total revenue against total spend, without platform-level attribution. It is a blunt instrument because it cannot tell you which channel drove which sale, but it gives you a ground truth to check channel-reported ROAS against. If your platforms collectively report a combined ROAS of 6x but your MER is 2.5x, something in the attribution is duplicating credit significantly.
We report weekly during active management. Each report contains:
We do not report impressions as a primary metric unless brand awareness is the defined objective. We do not report CTR as a success metric unless it is used as a proxy for creative quality in isolation from conversion data. The numbers in the report should map directly to the metrics the business uses to evaluate growth.
Scaling a paid media account is not just a matter of increasing the budget. Budget increases that outpace the algorithm's ability to find buyers at the current CAC produce a predictable pattern: CPCs rise, conversion rates stay flat, and CAC climbs. The account that ran at a 180 CAC at 10,000 a month is running at 280 at 30,000 because the audiences that convert efficiently at scale are not the same as the ones that convert at a small budget.
We scale in stages, validating each before moving to the next:
On Meta especially, creative fatigue is a real and measurable phenomenon. The same ad shown to the same audience enough times produces declining engagement and rising CPCs as the platform serves it to progressively less-engaged members of the audience. We track the frequency metric (average number of times each person in the audience has seen the ad) and refresh creative before fatigue becomes visible in the performance data rather than after.
The practical target is new creative every two to six weeks for prospecting audiences, depending on audience size and spend level. Small audiences at high spend fatigue faster. We keep a creative backlog so new variants are ready to deploy without a gap in delivery.
A direct list of the things we find most often in audits, with the spend impact each tends to carry:
Paid media management is not a one-time setup. The accounts that maintain efficient CAC over 12 months are the ones with a consistent optimisation cadence: weekly data review, monthly creative refreshes, quarterly channel and audience audits, and a feedback loop that connects campaign performance to sales pipeline quality.
Our engagement follows a four-phase structure:
One point of contact. You get one person who owns the account, the reporting, and the decisions. Campaign changes are documented in the weekly report with the rationale. You have full visibility into the account at all times and can export the data independently. When we stop working together, the account, the audiences, the creative library, and the historical data are yours.
How long before paid media shows results? Most accounts see measurable performance data within the first two to four weeks. Stable CAC typically takes six to ten weeks as automated bidding strategies exit the learning phase. Meaningful improvements from creative and audience testing accrue from month three onward.
Do you work with existing campaigns or start fresh? We start from the audit. If existing campaigns have a structure worth preserving, we work within it and fix the specific issues. If the structure is so fragmented that a rebuild is faster and cleaner, we rebuild. We will tell you which situation you are in within the first two days.
How do you decide what to test? Every test has a hypothesis written before it runs, based on what the current data shows about where performance is leaking. We do not test randomly. The hypothesis defines what we expect to learn, which means the test result is useful whether the challenger wins or loses.
What happens to the account data when we stop working together? Everything stays in your ad accounts. The campaigns, the audiences, the conversion history, the creative library, and the reporting structure are all yours. We do not create dependencies that make it expensive to switch.
That is the whole method. When you want it applied to your account, the next step is a free audit: real findings on your real data, in about two days, with no obligation.
A free audit on your real account data, the wasted spend quantified, and a managed engagement run to the CAC your business supports. Findings in two days.
Get a free account audit