Google Ads ROI & KPIs: What to Track and Why Most Reports Lie

Most guides to Google Ads metrics list 25 to 30 of them, define each one, and leave you to figure out which actually matter for your account. That's comprehensive, not useful — a business trying to check performance once a week doesn't need a full glossary, it needs the short list that actually drives a decision.

This covers that short list, how to read those metrics together rather than one at a time (a single number in isolation is often misleading), and something no metrics guide addresses directly: which of these numbers get quietly distorted by invalid traffic, and which stay reliable regardless.

None of the metrics or techniques below require anything beyond what's already visible inside a standard Google Ads account and Google Analytics — the barrier for most businesses isn't access to the data, it's knowing which pieces of it actually matter and how to read them together.

The starter set: six metrics worth checking weekly

Conversions and conversion rate. The actual count of genuine business outcomes (purchases, qualified leads, bookings), and that count divided by clicks. This is the metric everything else exists to support — a healthy-looking account with weak or absent conversion data is not actually healthy, it's unmeasured.

Cost per conversion (CPA). Total cost divided by total conversions — what you're actually paying to acquire one real outcome. This is the number worth judging against your own break-even math (covered in our guide to Google Ads costs), not against a generic industry benchmark.

ROAS or conversion value, where applicable. Total conversion value divided by total cost, expressed as a ratio — a 4:1 ROAS means $4 in tracked revenue for every $1 spent. Ecommerce accounts should have this running by default; lead-generation accounts need to assign a reasonable estimated value to conversions first (covered below) before this number means anything.

Invalid click rate. Visible under Campaigns, then Modify columns — what Google's own systems have already caught and filtered. A rate trending upward over several weeks is worth investigating before trusting any of the metrics above at face value.

Impression share, and specifically impression share lost to budget versus lost to rank. The first tells you whether you're leaving profitable, already-proven volume on the table because of a budget cap; the second tells you whether Ad Rank (bid and Quality Score combined) is the limiting factor instead. These call for genuinely different fixes, which is why the breakdown matters more than the single combined impression share number.

Quality Score, on your core, highest-spend keywords specifically. Not every keyword needs daily attention, but a handful of keywords carrying most of your spend quietly sitting at a 3 or 4 is worth knowing about, since it's directly inflating your cost-per-click on exactly the terms that matter most.

It's worth resisting the temptation to add more metrics to this list just because they're available — every additional column competing for attention in a weekly check dilutes the time spent on the ones that actually drive a decision. If a specific metric hasn't changed a real decision in the last few reviews, it likely belongs in the monthly deeper-dive set covered further down, not the weekly quick check.

These six work together as a set deliberately — conversions and CPA tell you what's happening at the outcome level, invalid click rate tells you whether that outcome data is trustworthy, and impression share plus Quality Score tell you whether you're constrained by budget, competition, or your own account's relevance. Missing any one of the six leaves a real blind spot in the picture the other five are trying to give you.

Reading metrics as diagnostic pairs, not in isolation

A single metric rarely tells a complete story on its own — the same number can mean genuinely different things depending on what's happening alongside it. A handful of pairings are worth internalizing specifically because they come up constantly.

High CTR paired with low conversion rate almost always points at a landing page or offer mismatch, not an ad copy problem — your ad is compelling enough to earn the click, but something after the click is losing the visitor. Chasing better ad copy here usually doesn't fix the actual issue.

Low CTR paired with high conversion rate suggests the ad itself may be under-selling a genuinely strong, relevant offer — worth testing more direct or benefit-focused copy, since the traffic that does click is clearly finding real value once it arrives.

Impression share lost to budget on a campaign with a strong, healthy conversion rate is one of the more straightforward, high-confidence fixes available: the demand and the conversion mechanism are both already proven, and the constraint is purely the budget cap — reallocating from a weaker campaign is usually a safe, fast win.

Impression share lost to rank on the same kind of campaign calls for a different, slower fix: better Quality Score through tighter ad relevance and landing page alignment, or a considered bid increase specifically on the contested keywords, rather than a blanket budget increase that won't address the actual constraint.

A third pairing worth knowing: a rising Quality Score alongside a flat or rising CPC is a genuine anomaly worth investigating rather than accepting at face value, since these two normally move in opposite directions — a real Quality Score improvement should be lowering your effective cost per click for the same position, and if it isn't, something else (a competitor escalating, as covered in our Auction Insights guide, or a broader market-wide CPC increase) is offsetting the improvement you're not otherwise seeing reflected.

The general habit worth taking from all three pairings: before making any change based on a single metric, ask what its natural counterpart is doing at the same time, and treat a genuine explanation as one that accounts for both numbers moving together, not just the one that happened to catch your attention first.

How invalid traffic distorts specific KPIs

This is the part most metrics guides skip entirely, and it matters because not every KPI is equally vulnerable — knowing which ones to trust less when you suspect a traffic-quality issue changes how you read the whole dashboard.

CTR is only mildly affected in most cases, since invalid clicks are still clicks against real impressions — the ratio moves, but usually not dramatically unless the invalid volume is severe. CPC and total cost are directly and proportionally inflated by every invalid click, dollar for dollar, making these some of the most reliably distorted numbers on the dashboard when a real traffic-quality problem exists.

Conversion rate is where the distortion becomes seriously misleading. Invalid clicks add to the denominator (total clicks) without ever contributing to the numerator (conversions), which mechanically drags conversion rate down — a real invalid-traffic problem can look, on the surface, exactly like a weak offer or a bad landing page, even though the underlying real traffic is converting perfectly normally. This is precisely the confusion covered in more depth in our conversion tracking guide: you can't reliably diagnose a conversion-rate problem without first ruling out traffic quality as the actual cause.

CPA is doubly hit, since it inherits both the inflated cost (numerator) and the suppressed conversion count (denominator) — a moderate invalid-click rate can make a genuinely well-performing campaign's CPA look considerably worse than its real, underlying efficiency, which is exactly the mechanism worked through numerically in our guide to Google Ads costs.

Impression share, by contrast, stays comparatively reliable as a signal even in the presence of invalid traffic — it's calculated from auction eligibility and participation, not from click quality, so a traffic-quality problem doesn't directly distort this particular number the way it distorts cost-based and conversion-based metrics. Quality Score sits similarly comparatively insulated in the short term, though a sustained, severe invalid-click problem can eventually drag it down indirectly through its effect on engagement signals over time.

Worth stating plainly, since it's easy to lose track of amid the mechanics above: none of this means every disappointing metric is secretly a traffic-quality problem in disguise. Genuine market conditions, real creative fatigue, and honest landing page issues remain far more common explanations for most performance shifts than invalid traffic. The point of this section is knowing which numbers to double-check specifically once you've reasonably ruled out the more common, boring explanations first — not treating traffic quality as the default suspect for every dip.

A worked example: the same dashboard, two different diagnoses

Say a campaign's conversion rate drops from 6% to 4% over two weeks, cost stays flat, and CPA rises accordingly. Two entirely different root causes can produce this exact same dashboard, and reading only conversion rate and CPA won't tell you which one you're looking at.

Diagnosis one: a landing page change introduced friction (a longer form, a confusing new layout), genuinely reducing how well real visitors convert. Check this by looking at on-site behavior metrics in Analytics — session duration, bounce rate, and scroll depth on the affected page — for a corresponding drop in engagement alongside the conversion rate drop.

Diagnosis two: invalid click volume increased over the same period, adding non-converting clicks to the denominator without anything actually changing about how real visitors behave once they land. Check this by looking at invalid click rate and the gap between Google Ads clicks and Analytics sessions for the same window — if that gap widened at the same time conversion rate dropped, traffic quality is the more likely explanation, not the landing page.

The two diagnoses call for completely different fixes — reverting or fixing the landing page versus investigating and blocking a traffic source — and applying the wrong one wastes real time while the actual cause continues unaddressed. This is exactly why the pairing and cross-referencing approach throughout this piece matters more than staring at any single metric in isolation.

This pattern — the same surface-level numbers, two genuinely different root causes — is common enough across the diagnostic pairs in this piece that it's worth treating as a general habit rather than a one-off technique: whenever a metric moves, ask what else changed at the same time before settling on the first explanation that comes to mind.

A second-tier list, worth checking monthly

Beyond the weekly starter set, a handful of metrics reward a less frequent, deeper look rather than daily attention. Search terms report activity — new queries appearing, spend accumulating on non-converting terms — is genuinely a monthly-review item for most accounts, tied directly to the negative keyword maintenance covered in our dedicated guide.

Device-segmented performance is worth a monthly check specifically because mobile and desktop behavior can diverge meaningfully without either showing up as a problem in the blended, account-level numbers — a mobile conversion rate quietly half of desktop's, for instance, can sit hidden inside a healthy-looking overall average indefinitely unless someone segments and looks.

View-through and assisted conversions matter primarily for accounts running Display or YouTube alongside Search, where a pure last-click view can meaningfully undervalue upper-funnel activity that's genuinely contributing to eventual conversions elsewhere in the account. This is a real, useful signal for those specific account types and largely irrelevant noise for a Search-only account, which is part of why it doesn't belong in a universal weekly starter set.

Time-of-day and day-of-week segmentation is worth a periodic look for any account with a meaningful ad-scheduling opportunity — genuine demand rarely distributes perfectly evenly across all hours, and a pattern showing a cluster of low-converting spend concentrated in specific windows is a concrete, actionable finding rather than a vague sense that "some hours perform worse."

None of these second-tier items need weekly attention specifically because they tend to move slowly relative to the starter-set metrics — checking them too frequently mostly just adds noise and time without producing new decisions, which is the same principle behind keeping the weekly list short in the first place.

Getting ROAS and conversion value right

For ecommerce, conversion value is usually straightforward — the actual transaction revenue, passed automatically if your tracking is set up correctly. The main thing worth double-checking is that the value being passed is genuinely the transaction total and not, say, a flat estimated figure left over from an early, rough setup that nobody's revisited since.

For lead generation, this takes deliberate work most accounts skip. Without an assigned value, a conversion action can't feed ROAS or value-based bidding strategies at all — it's worth assigning at least a rough estimated average value, based on your typical close rate and average deal size, even if it's not perfectly precise. A demo request and a newsletter signup are not equally valuable, and treating them as equal (or leaving both unvalued) gives Smart Bidding nothing meaningful to optimize toward beyond raw conversion count.

Where lead value genuinely varies by source or by specific form, assigning different values to different conversion actions lets the bidding algorithm correctly prioritize the higher-value action over the lower-value one — a distinction that matters considerably more than most lead-gen accounts currently account for in their setup.

Revisit these estimated values periodically rather than setting them once and forgetting them — close rates and average deal sizes shift as a business matures, a product line changes, or a market becomes more or less competitive, and a value assigned a year ago on outdated assumptions can be quietly steering Smart Bidding toward a target that no longer reflects reality.

Vanity-metric red flags in a report someone else hands you

If you're reviewing a report from an agency, freelancer, or internal team member rather than pulling data yourself, a few patterns are worth watching for specifically. A report emphasizing clicks, impressions, or CTR prominently while conversion data and CPA sit in small print or get left out entirely is a report built to look active rather than to show real business impact — activity is not the same as results.

A report showing a suspiciously low CPA with no visible primary-conversion designation shown is worth checking directly against the account, since (as covered in our conversion tracking guide) a soft action like a page view or a phone-number tap sitting in the Primary conversion slot inflates apparent efficiency without reflecting anything close to a real business outcome.

A report with no mention of invalid click rate or traffic quality at all isn't necessarily evidence of anything wrong — plenty of legitimate, well-run accounts simply don't lead with that metric in a summary report. But it's a reasonable, fair question to ask directly: what's the current invalid click rate, and how is it trending? A confident, specific answer is a good sign; a vague one is worth following up on.

A related pattern worth naming: month-over-month percentage improvements presented without the underlying absolute numbers. "Conversion rate improved 40%" sounds meaningful whether it moved from 1% to 1.4% (still weak in absolute terms) or from 5% to 7% (a genuinely strong result) — always ask for, or check, the actual absolute figures behind any percentage-change claim before treating it as evidence of real progress.

None of this is an argument for distrusting every report by default — most people presenting Google Ads performance are doing so honestly and competently. It's an argument for knowing specifically what to ask for when a report feels incomplete or unusually rosy, rather than either accepting it uncritically or dismissing it without a concrete, specific basis.

Comparing KPIs across a Google Ads manager account

If you're reviewing these metrics across several client accounts under an MCC, resist comparing raw KPI values directly account to account — a $50 CPA is excellent in one industry and alarming in another, for the same reasons covered in our cost guide, and comparing accounts by absolute numbers rather than by each account's own trend and its own break-even math produces misleading conclusions about which accounts are actually performing well.

What is genuinely worth comparing across accounts: the trend direction and rate of change on each account's own starter-set metrics, and whether the same diagnostic patterns (rising invalid click rate, impression share increasingly lost to rank rather than budget) are showing up simultaneously across multiple accounts — a pattern repeating across several unrelated client accounts at once is more often a platform-wide or market-wide shift than something specific to any one account, and worth investigating as such rather than diagnosing each account in isolation.

Keep a lightweight, shared log of each account's baseline starter-set numbers, updated monthly, specifically so a new pattern is genuinely comparable against that account's own recent history rather than being judged against whatever number happens to feel intuitively normal in the moment — memory is a poor substitute for an actual recorded baseline once you're tracking more than a couple of accounts.

Setting up a view that actually works

Rather than scrolling through dozens of default columns every time, build a custom column set specifically around the starter-set metrics above: Modify columns, then select conversions, conversion rate, cost per conversion, conversion value or ROAS, invalid clicks, invalid click rate, impression share, impression share lost (budget), impression share lost (rank), and Quality Score components. Save this as a named custom set so it's one click to load rather than reassembled from scratch each time.

It's worth building a second, separate saved view for deeper diagnostic work — segmented by device, time of day, and geography — kept apart from the weekly quick-check set specifically so the fast, routine review stays fast and doesn't get bogged down trying to serve two different purposes with one overloaded column configuration.

Revisit the saved column set itself every few months too — as an account matures, or as you add new campaign types, the specific metrics that belong in a fast weekly check can shift, and a column set built a year ago may no longer reflect what actually needs the closest attention today.

What's a good CTR for Google Ads?

There's no single universal figure worth chasing — CTR varies enormously by industry, campaign type, and even by keyword intent within the same account. A meaningfully higher CTR than your own account's historical average on similar keywords is a stronger, more relevant signal than comparing against any generic cross-industry benchmark, for the same reason covered in our cost guide: blended averages describe almost nobody's specific situation precisely.

What's a good ROAS?

It depends entirely on your gross margin, not on a generic target. Break-even ROAS is 1 divided by your gross margin as a decimal — a 50% margin needs at least 2.0x just to break even before accounting for other real costs beyond the product itself. Whatever figure you see cited as a general "good ROAS" benchmark is only meaningful once you've compared it against your own break-even math, covered in more detail in our guide to Google Ads costs.

How often should I actually review these numbers?

The starter set above is worth a genuine weekly check on any actively managed account — five to ten minutes once the custom column view is set up. A deeper monthly review, incorporating the diagnostic pairs and a broader look at segmented data, catches slower-moving trends the weekly quick-check isn't designed to surface.

Should I trust Google's own automated recommendations and optimization score?

Treat Google's optimization score and in-account recommendations as suggestions to evaluate against your own data, not instructions to follow automatically. Several commonly suggested actions (broadening match types, applying every suggested keyword, raising budgets or bids) genuinely can improve the optimization score number while simultaneously increasing spend without a proportional improvement in the KPIs that actually matter to your business — read any recommendation against the starter-set metrics above before applying it, not against the optimization score itself.

How many conversions do I need before these metrics are statistically meaningful?

There's no universal hard threshold, but a common, reasonable rule of thumb: treat conversion rate and CPA as directional rather than firm below roughly 30 conversions in the period you're measuring, since small sample sizes make these ratios genuinely noisy and prone to swinging on a handful of outcomes either way. This is also close to the volume Google's own Smart Bidding guidance cites as a reasonable minimum before trusting an automated strategy's performance judgment for the same underlying statistical reason.

Should I look at these metrics per campaign, per ad group, or account-wide?

Account-wide numbers are useful for a quick pulse check but hide real variation — a strong campaign and a genuinely struggling one can average out to a perfectly ordinary-looking account-level number. The weekly starter set is worth checking at the campaign level at minimum, and at the ad group level for your highest-spend campaigns specifically, since that's where a real problem is most likely to be hiding inside an acceptable-looking blended average.

The short version

Six metrics, checked weekly, cover most of what actually drives a decision: conversions and conversion rate, cost per conversion, ROAS or conversion value, invalid click rate, impression share (split by budget versus rank), and Quality Score on your top-spend keywords. Read them in pairs rather than isolation, remember that cost and conversion-based metrics are the ones most distorted by invalid traffic while impression share stays comparatively reliable, and build a saved column view so this becomes a five-minute habit rather than something reconstructed from scratch every time you check.

Does Performance Max need a different set of metrics?

The starter set above still applies as the core outcome metrics, but PMax gives you less granular visibility to diagnose problems within those numbers — no keyword-level breakdown the way Search offers. For PMax specifically, asset group performance ratings and the search terms insights report (PMax's closest equivalent to a traditional search terms report) become the practical diagnostic layer sitting underneath the same conversions, CPA, and ROAS numbers you're already tracking, since you can't negative-keyword your way to precision the way you can in Search.