All posts
Paid Media

The Paid Media System

An interactive map of how all the pieces of a paid media campaign connect, from first-party data and audience targeting to bid strategies, attribution, experimentation, and the metrics that actually matter.

By Matt Brown·March 2026

Paid media has a lot of moving parts. Budget, pacing, bid strategy, audience targeting, keywords, creatives, landing pages, measurement, attribution, and experimentation all interact with each other and with external factors you cannot control, like competition, demand shifts, and seasonality.

Most resources explain these pieces in isolation. This guide maps them as a connected system so you can see how a change in one area ripples through everything else. Click any node in the diagram below to explore how it works and why it matters.

The offer is the foundation

The offer is the biggest lever on performance. It needs to appeal to your audience and match across the ad and landing page. No amount of optimization rescues a weak offer.

First-party data

Data you upload about your business to help Google algorithms identify your most valuable customers. Providing first-party data to google can greatly improve performance.

Transaction data
Order value, repeat purchases, LTV signals
On-site behavior (GA4)
Page visits, events, cart abandons, form responses
Offline events
In-store visits, calls, in-person purchases
External signals
Weather, inventory, promotions

Transaction data

Pass order value and purchase history back to Google via your conversion tag. Over time, Google uses this to shift from optimising for conversion volume toward customer value — bidding more aggressively for users who are likely to spend more, not just convert once.

On-site behavior (GA4)

Page visits, events, time on site, cart abandons, and form responses feed remarketing lists and inform customer value signals.

Offline events

Uploading data on in-store visits, phone call outcomes, in-person purchases, etc gives Google a more complete picture of the customer journey.

External signals

Weather, inventory levels, foot traffic, stock price.

How to use this data to enhance performance?

Customer Match
Target or exclude specific customers
Remarketing audiences
Re-engage users based on past behaviour
Lookalike audiences
Find new users who resemble your best customers
Dynamic spend rules
Trigger budget shifts from external signals

Customer Match

Upload a list of emails, phone numbers, or physical addresses and Google matches them to signed-in users. Use this to target known customers directly, exclude them from campaigns, or use them as a seed for lookalike audiences.

  • ExclusionExclude recent purchasers from acquisition campaigns so you're not paying to re-acquire someone who just bought
  • Upsell / cross-sellTarget customers who bought X but not Y with a tailored message rather than a generic acquisition ad
  • Lapsed re-engagementIdentify customers who haven't purchased in 60, 90, or 180 days and target them with win-back campaigns

Remarketing audiences

Use the Google tag or GA4 to build audiences based on how users have already interacted with you.

  • Product viewersVisitors who viewed a product but didn't purchase
  • Cart abandonersUsers who added to cart but abandoned checkout
  • Lapsed usersApp users who haven't logged in for 30+ days

Lookalike audiences

You upload your highest value customers and Google identifies common characteristics to find new users who match that profile.

Mortgage leads

Someone fills out your mortgage form and provides information like loan size, credit score range, and property type. Some of those leads eventually close on a mortgage, while others don't. Without additional data on actual mortgage closures, Google can only optimize for people that fill out the form. However, you can upload which leads became customers back to Google so it can optimize towards the borrowers who are most likely to close.

Dynamic spend rules

Use external signals like weather, inventory levels, day of week, and promotions to trigger automatic spend adjustments. Rather than manually shifting budgets, you define rules that tell Google when conditions change.

  • Boost spendIncrease budget when conditions predict higher conversion rates, like warm weather for seasonal products
  • Reduce spendPull back when inventory of a product runs low to avoid paying for clicks you can't convert

The thresholds in these rules don't have to be guesses. Regression analysis on your historical data can surface which signals actually predict performance and by how much, so your rules are calibrated to real lift rather than intuition.

Ice cream & weather

An ice cream company knows sales spike when it's warm. They feed Google a weather signal. When temperature exceeds 20°C in a given location, increase spend. Google uses this to shift budget automatically toward warm regions on hot days, without anyone manually adjusting campaigns.

sharpens
What you control
$ Budget
How aggressively you buy traffic
Pacing
When budget is spent
Bid strategy
How you compete in each auction
Attribution
Which ad gets credit for a conversion
Audience
Demographics, location, interests, etc.
Keywords
What they're actively searching for
Creatives (ads)
Copy, visuals, format, CTA
Landing page
Speed, content, form integrity
Measurement
What counts as a conversion

$ Budget

How aggressively you buy traffic.

  • Increasing budget: Leads to more impressions, clicks, conversions and revenue at decreasing efficiency

Pacing

When is budget spent across the month?

  • Even: A consistent daily target ($ budget per month / 30.4 days)Good for always-on campaigns focused on awareness or steady lead generation.
  • Front-loaded: Spend more at the start of the monthEffective for new product launches, short-term promotions, and gathering performance data quickly
  • Back-loaded: Spend more at the end of the monthGood when conversion rates spike later e.g. around payday
  • Day-parting: Budget adjusted by day or hourUseful when certain times yield meaningfully higher conversion rates
  • Active pacing: Algorithm sets budget dynamically.Effective but cedes control to the AI

Bid strategy

How do you compete in each auction?

  • Manual CPC: You set the maximum bid per clickFull control, but requires active management
  • Target CPA (tCPA): Platform optimizes bids to hit a target cost per acquisition
  • Target ROAS (tROAS): Platform optimizes bids to hit a target return on ad spend
  • Max conversions: Platform spends the full budget to maximize conversion volume regardless of cost

Attribution

Determines which ad gets credit for a conversion. The model you choose changes how spend is allocated and how performance appears

  • First-click: 100% of credit goes to the first ad the user clicked. Favors awareness and prospecting campaigns that initiate the journey.
  • Last-click: 100% of credit goes to the final ad clicked before converting. The default in most platforms. Over-credits retargeting and brand keywords.
  • Linear: credit split evenly across every touchpoint in the path. More balanced but treats every interaction as equally important.
  • Data-driven: machine learning distributes credit based on which touchpoints actually influenced conversion. Requires sufficient conversion volume to train the model.

Cross-channel attribution connects touchpoints across Google, Meta, email, and other platforms into a single conversion path. Without it, every platform claims the same conversion and total attributed revenue looks inflated.

Audience

Who sees your ads. Google automatically targets users most likely to convert, but you can manually refine by demographics, location, and device. And also improve the signal by uploading first-party data.

  • Demographicsage, gender, income, job title
  • Locationcountry, region, city, radius
  • Devicemobile vs desktop

Keywords

What they're actively searching for

  • Broad match: platform shows ad for related searches. Maximum reach, lower relevance
  • Phrase match: ad shows when query contains the keyword phrase. Balanced reach and relevance
  • Exact match: ad shows only for that specific query. Highest relevance, lowest reach
  • Negative keywords: queries you're actively blocking. Prevents wasted spend on irrelevant searches

Creatives (ads)

Creative quality is the primary lever for CTR. A better ad means more clicks from the same impressions

  • Higher CTR: more clicks from same budget
  • Relevance score: better quality score leads lower CPCs
  • Ad fatigue: CTR declines over time so rotate creative regularly

Landing page

The page a user sees after clicking. Page speed regressions, content changes, and broken forms directly suppress conversion rate: without any change to ad spend or delivery.

  • Slow load time: +1s can reduce CVR by 7–20%
  • Form breakage: zero conversions tracked despite normal click volume
  • Message mismatch: ad and page out of sync, high bounce, low CVR

Measurement

Measurement is how you tell the platform what success looks like. The conversion action you choose becomes the signal the algorithm optimizes toward. Choose wrong and you're training it on the wrong behavior.

  • Primary conversion actionThe single action that defines success for this campaign. For e-commerce this is a purchase. For lead gen it might be a form submission, a phone call, or a booked meeting. The bidding algorithm optimizes entirely toward this.
  • Secondary conversion actionsActions you want to observe but not optimize toward, like page views or video plays. Secondary actions inform reporting without influencing bids.
Use one primary conversion action

If you mark both "Qualified Lead" and "Closed Deal" as primary conversions, the algorithm sees them as equivalent signals and doesn't know which to chase. A lead generates far more signal volume than a closed deal, so the algorithm gravitates toward optimizing for leads, not revenue. Pick the conversion that is closest to money and set everything else as secondary or observation-only.

  • Double-countingThe most common measurement mistake. If your GA4 import and your Google tag both fire on the same thank-you page, every purchase gets counted twice. Inflated conversion numbers make CPA look half what it actually is and cause the algorithm to overbid. Audit your active conversion actions regularly and remove duplicates.
validated by
Experimentation

The only way to know if a change actually caused a result, not just correlated with it. Without a controlled test, you're reading noise as signal.

A/B testing
Isolate one variable, compare two versions
Incrementality
Would the conversion have happened anyway?
Holdout groups
Measure true lift by withholding ads from a control group

A/B testing

Change one thing, a headline, a landing page, a bid strategy, and split traffic evenly between the two. The difference in outcome is causally attributable to that change. Most platforms have native experiment tools; use them instead of making changes mid-flight and comparing before/after.

Incrementality

Incrementality answers whether your ads are actually driving conversions or just taking credit for purchases that would have happened organically. High ROAS on brand campaigns often collapses under an incrementality test because those users were going to convert anyway.

Holdout groups

Suppress ads from a randomly selected control group while running normally for everyone else. The gap in conversion rate between exposed and holdout is the true lift your campaigns are generating.

also influenced by
External factors
Competition
Other advertisers in your auction
Demand shifts
Changes in search volume & intent
Seasonality
Predictable performance swings

Competition

More advertisers in the auction drives up CPCs and suppresses ROAS without any change on your end.

  • CPCs rise: auction becomes more competitive, you pay more per click
  • ROAS falls: same conversions, higher spend
  • Signals: impression share loss, rising CPCs with stable CVR

Demand shifts

Changes in search volume or buyer intent affect how much impression opportunity exists: independent of your campaign settings.

  • Volume drops: fewer searches, fewer impressions even at full budget
  • Intent shifts: same keywords, different buyer mindset, CVR changes
  • Signals: impression volume moves without bid or budget changes

Seasonality

Predictable CVR & AOV swings driven by time of year, holidays, and pay cycles.

  • Peak periods: Black Friday, Christmas, back to school, CVR spikes
  • Slow periods: Q1 dips, summer lulls, higher CPA, lower ROAS
  • Plan ahead: adjust budgets and bids before peaks, not during
flows into
Results
Platform metrics
Impressions
Times ad is shown
Clicks
Visits to landing page
Requires tracking
Conversions
Completed actions
AOV
Average order value

Impressions

The raw reach of your campaign: how many times your ad was shown. Driven primarily by budget, audience size, and bid. High impressions with low CTR signals a creative or audience relevance problem.

  • Impressions ÷ Reach = frequency (watch for fatigue above ~3–5x)
  • CPM = Spend ÷ (Impressions / 1000): your cost to reach 1,000 people

Clicks

Impressions that converted to a visit. CTR (Clicks ÷ Impressions) is the primary signal of creative and audience fit.

  • Low CTR: poor creative or wrong audience
  • High CPC: competitive auction or low quality score

Conversions

Clicks that turned into the desired action (purchase, lead, sign-up). CVR (Conversions ÷ Clicks) is the joint function of landing page quality and offer relevance.

  • Clicks steady, conversions drop generally is a landing page or measurement issue
  • CVR improvement is the highest-leverage way to lower CPA

Average order value (AOV)

The average revenue per conversion. Multiplied by Conversions gives total conversion value

  • Promos & discounts can lead to AOV drops which can turn profitable campaigns unprofitable
combined into
Performance metrics — calculated
CPA
Cost per acquisition
Spend ÷ Conversions
ROAS
Return on ad spend
Conv. value ÷ Spend
Profit
Net return
Conv. value − Spend

Cost per acquisition (CPA)

The average cost for one conversion. Your target CPA is determined by what you can afford to pay for a customer and still be profitable. CPA is the primary optimization target for most campaigns.

  • CPA rises: check CVR, audience quality, landing page, or bid strategy
  • Target CPA (tCPA) bidding instructs the platform to optimize toward your threshold
  • CPA must be below (AOV × margin) to be profitable

Return on ad spend (ROAS)

Revenue generated per dollar spent. A ROAS of 4x means every $1 in ad spend produced $4 in revenue

  • Break-even ROAS = 1 ÷ gross margin (e.g. 40% margin leads to break-even at 2.5x)
  • ROAS drops: check AOV, CVR, and attribution window
  • Blended ROAS masks channel-level efficiency so it's best to always segment

Profit

The ultimate metric, revenue generated minus ad spend. ROAS can look healthy while profit is negative if margins are thin.

  • Profit = (AOV × CVR × Clicks) − Spend
  • Small AOV improvements compound into large profit gains at scale
  • The goal is max profit, not max ROAS.
monitored by
Alerts

Every layer above can break silently. Alerts watch your metrics and notify you the moment something moves outside expected ranges.

Budget
Underspend, overspend, pacing off track
Performance
CPA spike, ROAS drop, CVR anomaly
Tracking
Pixel fires drop, conversion gaps

Built by Avisos

Why a systems view matters

When performance changes, the instinct is usually to look at the most visible metric and try to fix it directly. CPA spiked, so lower the target. ROAS dropped, so cut budget. But those metrics are outputs of the system, not inputs.

A CPA spike could trace back to a landing page that slowed down, a conversion tag that broke, a competitor entering the auction, or a creative that fatigued. Without understanding the connections between layers, you end up treating symptoms instead of causes.

The diagram above is designed to make those connections visible. First-party data sharpens the inputs you control. Those inputs are validated by experimentation. External factors influence the results alongside your inputs. Results combine into the performance metrics you report on. And alerts monitor every layer so problems surface before they compound.

How to use this guide

If you manage paid media campaigns, use this as a diagnostic map. When something changes in your results, trace it upstream through the system to find the actual cause. When you are planning a test, follow the connections downstream to predict which metrics it will affect.

If you are newer to paid media, start at the top and work down. The system flows from data collection through campaign inputs, into experimentation and external factors, and finally into the results and calculated performance metrics at the bottom.

Back to all posts