Pay-Per-Click Advertising Explained: Auctions, Targeting, Costs, and Attribution

Pay-per-click (PPC) advertising is a digital advertising model in which an advertiser is charged when someone clicks an ad. In paid search, advertisers choose keywords and other targeting settings, set bids and budgets, and compete in an auction for each eligible query. Paying for the click does not buy a lead or a customer. A useful PPC model therefore connects four layers: targeting selects the auctions, the auction controls delivery and price, the landing experience turns traffic into action, and attribution assigns credit for the result.

PPC is the payment model; paid search is the channel#

PPC tells you how an ad interaction is priced, not where the ad appears. Paid search tells you the inventory: sponsored placements associated with search results. Many paid-search campaigns use cost-per-click billing, but PPC can also appear outside search, and advertising platforms can sell inventory under other pricing models.

The adjacent terms are easy to collapse:

Google describes Google Ads as its PPC advertising solution and explicitly separates it from SEO: buying ads does not improve organic rankings. Microsoft likewise uses PPC inside its search-engine marketing explanation. For a clean operating vocabulary, call the channel paid search, the billing model PPC, and the recorded price CPC.

The five formulas that keep PPC reporting honest#

PPC itself has no score that says a campaign is good. The core cost calculation is:

Average CPC = total click cost ÷ total clicks

Four supporting formulas connect delivery to a business outcome:

CTR = clicks ÷ impressions × 100

Conversion rate = attributed conversions ÷ eligible ad interactions × 100

CPA = conversion cost ÷ attributed conversions

Reported ROAS = attributed conversion value ÷ ad cost

Google’s average-CPC definition makes an important distinction: average CPC is based on the actual charges recorded for clicks and can differ from a maximum CPC bid. A bid is an auction input or constraint; it is not the invoice for every click.

Here is illustrative arithmetic, not real account or monetary data. A campaign records 10,000 impressions, 400 clicks, 20 defined conversions, 2,000 cost units, and 6,000 units of attributed conversion value.

• CTR is 400 ÷ 10,000 = 4%.
• Average CPC is 2,000 ÷ 400 = 5 cost units.
• Conversion rate is 20 ÷ 400 = 5%.
• CPA is 2,000 ÷ 20 = 100 cost units.
• Reported ROAS is 6,000 ÷ 2,000 = 3.0.

That row is arithmetically complete and commercially incomplete. The 20 conversions might be purchases, qualified opportunities, raw form fills, or several event types counted together. The attributed value might be revenue, a fixed lead value, or an imported downstream value. Change the conversion definition, eligible interaction set, window, or attribution model and the reported CPA and ROAS can change without a corresponding change in the underlying business.

Every eligible search enters a new auction#

A search campaign does not buy a permanent position for a keyword. On Google, an auction runs for each search when an ad may be eligible to appear. The sequence is roughly:

1. The platform identifies ads whose keywords or other matching logic relate to the query.
2. It removes ads that are ineligible because of targeting, policy, budget, or other campaign conditions.
3. It calculates Ad Rank for the remaining candidates.
4. Ads clearing the relevant thresholds may appear, and their rank determines relative placement.

Ad Rank is not simply bid × Quality Score. Google’s Ad Rank documentation names several factors: the bid; the expected quality of the ad and landing page; Ad Rank thresholds; auction competitiveness; the person’s search context, including query, location, device, and time; and the expected impact of assets and formats.

This explains two results that look contradictory until the auction is understood. A higher bidder can lose a better position to a more relevant ad. A lone eligible advertiser can still face a reserve threshold rather than receive a nearly free click. The auction is recalculated in context, so position and CPC can move even when the advertiser changes nothing.

Quality Score deserves its own boundary. Google calls the visible 1–10 Quality Score a diagnostic tool, not a direct auction input. Its components can identify weak expected click-through rate, ad relevance, or landing-page experience, but optimizing the displayed score is not the same as improving qualified acquisition.

Targeting decides which auctions you are willing to enter#

In paid search, targeting begins with the query but does not end there. A search term is what the person typed or otherwise submitted. A keyword is an advertiser-controlled input used by the platform’s matching system. Treating them as the same object is one of the fastest ways to misunderstand spend.

Google documents three keyword match types: broad, phrase, and exact. Their reach overlaps, with broad covering the widest set and using signals beyond literal wording. “Exact” is therefore a control category, not a promise that every served query will be character-for-character identical to the keyword.

The operating loop is simple: choose an intent hypothesis, observe the queries that actually received traffic, and revise. Google’s search terms report is designed for that feedback. Relevant queries can become more deliberate targets; irrelevant queries can become negative keywords or lead to a change in match type, structure, offer, or landing page.

Other settings narrow or reshape eligibility:

Geography is a particularly quiet source of leakage. Google says its default location option can include people in or regularly in a place and people showing interest in that place. It also describes location inference as a best effort rather than 100% accurate. A local operator should therefore define whether location interest is useful, inspect geographic performance, and not assume a city name in campaign settings is a hard physical boundary.

PPC cost is an economic outcome, not a rate card#

There is no fixed price for “doing PPC.” Click cost varies by the auctions entered and the value competitors place on them. Microsoft summarizes paid-search CPC as depending on bid, competition, and relevance; Google adds query context, thresholds, quality, and asset impact. A daily or monthly budget constrains participation. It does not set a uniform CPC, guarantee click volume, or make the resulting traffic valuable.

Public benchmarks are useful only when their denominator and population resemble yours. LocaliQ’s June 2026 search-advertising report summarizes thousands of its customer campaigns across Google Ads and Microsoft Ads. It reports cross-industry figures of $5.42 CPC, 6.64% CTR, $66.69 cost per lead, and 8.18% conversion rate. The industry rows vary materially.

Those figures are a reference set, not a target. They mix businesses, offers, geographies, query portfolios, conversion definitions, margins, and account maturity. A campaign with a CPC above the aggregate can be excellent if it acquires valuable customers efficiently. A campaign with cheap clicks can be wasteful if the queries, offer, or downstream lead quality are poor.

The most useful planning identity works backward from economics:

Maximum affordable CPC ≈ target CPA × expected click-to-conversion rate

For example, if an economically acceptable CPA is 120 cost units and the expected conversion rate is 4%, the implied CPC ceiling is 120 × 0.04 = 4.8 cost units. This is illustrative planning arithmetic, not a bid recommendation. It holds only if CPA and conversion rate use consistent events, windows, and traffic. If only one in four recorded leads becomes qualified, the model must use that downstream qualification rate or the CPC ceiling will be too generous.

Measure from the click to the qualified outcome#

A clean PPC report separates diagnostic metrics from decision metrics.

CTR and CPC help diagnose message and auction performance. They cannot decide success on their own. A high CTR may mean an ad is compelling, or that its promise is broader than the landing page can honor. A low CPC may reflect efficient buying, or low-value traffic. The decision metric belongs as far downstream as measurement quality permits.

Google defines a conversion as an advertiser-selected valuable action such as a sale, lead, sign-up, call, or download. That flexibility is useful and dangerous. If an account marks page views, form starts, demo requests, qualified opportunities, and purchases as equally important outcomes, reported conversion volume becomes a mixture that automated bidding may optimize toward.

Choose one primary outcome for the campaign’s decision. Keep earlier events as diagnostics. For B2B lead generation, pass a stable click or campaign identifier through the landing experience, preserve it in the CRM, and return the qualified or converted-lead outcome when the platform and consent design permit. The point is not technical completeness for its own sake. It is to prevent a cheap form submission from being mistaken for efficient customer acquisition.

Attribution assigns credit; it does not create the result#

Attribution answers: which eligible interaction receives how much reported credit for a conversion? A person may click a non-brand ad, return through a brand ad, and later convert. A last-click model assigns all credit to the last eligible click. Google’s data-driven model distributes credit based on patterns in the account’s eligible conversion paths.

Google Ads currently documents last-click and data-driven attribution; first-click, linear, time-decay, and position-based models are no longer supported there. The choice affects conversion reporting and any automated bidding that consumes those conversions. Changing the model can therefore move credit among campaigns or keywords without creating additional customers.

Three boundaries belong in every PPC readout:

1. Event boundary: What exact action counts as a conversion?
2. Time boundary: Which interaction and conversion windows are eligible?
3. Channel boundary: Which platforms and offline steps can the measurement system actually observe and credit?

Within those boundaries, attribution is useful for optimization. Outside them, it is silent. A Google Ads path report cannot, by itself, value an unobserved sales conversation, an organic exposure, or a competitor comparison on another device.

Google’s Conversion Lift documentation makes that distinction operational: it compares an exposed group with a control group and reports incremental conversions, incremental CPA, or incremental ROAS when the design is eligible. Lift studies have their own power and implementation limits, but they ask a different question from attribution.

Use a four-gate PPC review#

Before approving a campaign or increasing its budget, write one line for each gate:

If one line cannot be written, more platform configuration will not repair the missing decision. A keyword list without demand logic buys ambiguity. A relevant ad without a matching landing experience buys disappointment. A conversion without downstream value buys a flattering dashboard. An attribution model without a declared scope buys false certainty.

Use PPC when identifiable demand exists, the offer can answer it, the value of a qualified outcome supports auction prices, and measurement can follow the click far enough to make a decision. It is especially useful when speed and query-level control matter. Do not use it merely because traffic can begin quickly.