Review Quiz & Fixing Pitfalls

When the dashboard feels like a trick question

It’s the last day of the month. Your ad platforms insist performance is strong: ROAS is up, CPL is down, conversion counts look healthy. Then you open your website analytics and see fewer sessions than clicks, fewer purchases than the platform claims, and a channel report that shows “Direct” suddenly spiking. Finance asks a simple question—“Did marketing work?”—and you realize the numbers don’t line up well enough to answer confidently.

This lesson is a practical “review quiz in prose”: a fast, structured way to spot the most common pitfalls beginners hit in marketing analytics, and the fixes that make your reporting defensible. The goal isn’t to memorize definitions—it’s to build a habit: when metrics disagree, you know what to verify first, what to ignore until definitions are aligned, and what conclusions are safe to draw.

You’ll use four ideas as your safety rails: clear units and definitions, funnel diagnosis, attribution as a chosen credit rule, and incrementality as the reality check.

The definitions that prevent 80% of mistakes

Three small clarifications eliminate most “analytics arguments” before they start. First: metrics vs. dimensions. A metric is a number you aggregate (clicks, sessions, revenue). A dimension is how you slice it (channel, campaign, device, landing page). The pitfall is mixing levels: comparing a campaign-level metric from an ad platform to a session-level metric from web analytics without noticing they count different things. A classic symptom is “clicks > sessions,” which is often normal because not every click becomes a measurable session (blocked scripts, bounces before tag loads, redirects, privacy limits).

Second: KPI vs. supporting metric. A KPI is the outcome you commit to optimize because it reflects business success (qualified leads, paid orders, subscription starts, CAC). A supporting metric explains why the KPI moved (CTR, CVR, AOV, lead-to-close rate). The misconception is treating every visible number as a KPI, which encourages local optimization: celebrating higher CTR while conversion rate falls, or lower CPC while lead quality collapses. A clean mental model is: KPIs answer “are we winning?” and supporting metrics answer “why did it change?”

Third: unit of analysis—your most important sentence in any report. Are you counting users, sessions, clicks, orders, or leads? Each unit duplicates differently. One person can create multiple sessions, one session can include multiple events, and one conversion can be claimed by different touchpoints depending on the model. If you don’t declare the unit, you accidentally compare incompatible numbers and end up “fixing” the wrong thing.

The four pitfall zones—and how to fix each one

1) Funnel thinking: diagnose before you optimize

A funnel is a model of cause-and-effect, not a literal map of every customer journey. Used well, it keeps you from guessing. When revenue drops, the funnel forces a disciplined question: did the change happen at the top (traffic/reach), middle (conversion behavior), or bottom (value and retention)? Beginners often skip this and jump straight to channel blame (“paid social is broken”) because that’s what the dashboard is organized around.

A strong practice is to map each stage to an observable event with a crisp definition. For e-commerce, that might be: ad click → landing-page session → product view → add-to-cart → checkout → purchase. For lead gen: ad click → session → form submit → MQL → SQL → opportunity → closed-won. The pitfall is using micro-conversions (email signup, add-to-cart) as if they were the business outcome. Micro-conversions are useful because they move faster and give more data, but they can mislead when campaign changes increase low-intent behavior that doesn’t translate downstream.

The “fixing pitfalls” move is to treat funnels as a diagnostic map, then validate the instrumentation at the exact stage where the break appears. If clicks are steady but sessions fall, you investigate redirects, page load, tag firing, or cookie consent. If sessions are steady but checkout completion drops, you look at payment errors, shipping surprise costs, or broken promo codes. Funnels don’t settle attribution debates—but they do tell you where to look first, so you don’t waste time optimizing ads when the real problem is the site.

2) Attribution: not “truth,” but a consistent credit rule

Attribution answers “who gets credit?” and it’s always a choice. Last-click is operationally simple and often useful for short-term optimization, but it tends to over-credit bottom-of-funnel touches like branded search, retargeting, or email. First-click emphasizes discovery, but can under-credit the touches that close the deal. Multi-touch options (linear, position-based, time-decay) spread credit differently—and each tells a different story.

A core misconception is that one attribution model is “the correct one.” In practice, the goal is consistency and decision-fit. If you’re deciding which ad set to pause tomorrow, last-click may be acceptable. If you’re defending budget for a top-of-funnel channel, you need a model (and supporting evidence) that recognizes early influence, such as assisted conversions, cohort behavior, or structured tests. Another pitfall is comparing platform numbers to analytics-tool numbers without aligning the rules: platforms may claim view-through conversions, longer windows, or different identity matching than your website analytics can observe.

The fix is to make attribution assumptions explicit in the report itself: model, window, counting method, and source of truth for conversions. If your ad platform says 500 purchases and your analytics tool says 380, you don’t start by accusing either tool of being wrong—you ask: are they deduping the same way, observing the same users, and using the same time window? Once the rules are visible, the discrepancy becomes explainable, and your conclusions become defendable.

3) Incrementality: the reality check behind every “win”

Incrementality asks the question your dashboard can’t answer by default: did marketing cause additional outcomes beyond the baseline? Most reporting is observational. If conversions rise after spend increases, it might be causal—or it might be seasonality, pricing, inventory, PR, or competitor changes. Beginners often treat platform ROAS as causal proof, but ROAS usually reflects performance under the platform’s measurement system, not necessarily true incremental profit.

Where this bites hardest is branded search, aggressive retargeting, and campaigns aimed at existing customers. These can look extremely efficient because they intersect with people already close to buying. That doesn’t make them useless—it means you must be careful with the claim. A common pitfall is scaling spend based on high reported ROAS, only to discover overall revenue barely changes because the ads mostly re-attribute conversions that would have happened anyway.

The fix is incremental thinking as a habit, even when you aren’t running formal experiments. You look for signs of cannibalization: does paid growth coincide with organic or direct declines? Do new-customer conversions rise, or only total conversions? Does performance hold when you expand to new audiences rather than repeatedly hitting the same high-intent pool? When possible, you use controlled comparisons (holdouts, geo splits, or other structured tests). The key is separating two statements: “This channel is efficient under this attribution model” versus “This channel creates lift.” Confusing them is one of the most expensive analytics pitfalls in online marketing.

4) Data quality and definitions: the hidden cause of “conflicting truths”

Data quality isn’t glamorous, but it decides whether any dashboard is trustworthy. Online marketing measurement spans ad platforms, website analytics, pixels, server events, UTMs, and often a CRM. Small mismatches create big arguments: a “purchase” event that fires twice on refresh, UTMs that fragment “Paid Social” into multiple labels, or a CRM that defines “qualified” differently than marketing does. The result is dashboards that disagree, and teams that waste energy fighting the numbers instead of improving performance.

A best practice here is measurement governance: a small, documented set of definitions and conventions that everyone follows. That includes a KPI dictionary (what exactly is CAC, qualified lead, revenue), an event naming scheme, a campaign naming convention, and a simple change log so you can annotate shifts when tracking or site changes occur. Another critical practice is aligning on the business outcome source of truth: orders, subscription starts, qualified opportunities—something stable and auditable—then treating upstream engagement metrics as diagnostic.

Common pitfalls include mixing gross vs. net revenue across reports (phantom swings), comparing date ranges without controlling for weekday/promo patterns, and assuming “more tracking” automatically means “more accuracy.” Often, more tracking increases noise unless definitions tighten. The fix is not to chase perfect measurement—it’s to build measurement that’s consistent enough to support decisions, and transparent enough that discrepancies can be explained rather than argued.

A quick “spot the pitfall” guide

[[flowchart-placeholder]]

Two online marketing scenarios, fixed step-by-step

Example 1: Paid social looks profitable, but cash outcomes disappoint

You run paid social to a product landing page. The ad platform reports a high ROAS and a surge in “purchases.” Website analytics shows fewer transactions than the platform claims. Finance reports margin pressure and a weaker month than expected. The pitfall is picking a single screenshot as “the truth” and making budget decisions from it.

Start with definitions and unit of analysis. What is a “purchase” in each system—order created, paid order, shipped order? Is revenue gross, net of discounts, or net of returns? Are you counting orders, unique purchasers, or purchase events that might duplicate? Once you align those definitions (or at least document the mismatch), you can interpret gaps instead of arguing about them. Often, the ad platform is counting conversions with rules your site tool can’t perfectly verify, especially when identity and tracking differ.

Next, use the funnel as a diagnostic map. Check whether clicks stayed flat while landing-page sessions dropped (tracking loss, slow loads, consent issues). If sessions are steady but add-to-cart rate declines, the offer/page is likely the issue. If add-to-cart is steady but purchase completion drops, checkout friction or payment failures may be responsible. This step prevents you from “optimizing ads” when the real leak is on-site, which is a common and costly misallocation.

Finally, apply attribution and incrementality cautiously. Treat platform ROAS as “performance under that platform’s attribution,” not as profit proof. If retargeting is heavy, ask the incremental question: are you mostly reaching people who were already going to buy? If so, reported ROAS can be high while incremental lift is modest. The practical fix is often a re-balance—reduce over-creditable retargeting, strengthen prospecting, and improve landing/checkout—so performance improves in business terms, not just in platform-reported terms.

Example 2: Lead volume surges, CPL drops, and sales says quality collapsed

A B2B campaign sends search and LinkedIn traffic to a gated content page. Marketing reports cheaper leads and rising form submits. Sales reports that most leads are low-fit and pipeline quality is down. The pitfall is optimizing the easiest-to-move number—form submits—and calling it success.

Begin by tightening definitions across marketing and sales. What counts as a lead (form submit) versus a qualified lead (meets firmographic/intent criteria) versus an opportunity (sales-accepted and in pipeline)? If your KPI is “leads,” the system will naturally optimize toward low-resistance conversions. If your KPI is “qualified opportunities,” you apply pressure where the business outcome lives. This is a definitions problem first, not an ad-platform problem.

Then map a funnel that matches reality: click → session → form submit → MQL → SQL → opportunity → closed-won. Calculate stage conversion rates, and you’ll often find the real break: form submits rose, but the MQL or SQL rate fell sharply. That points to specific fixes: tighten targeting, adjust the offer to attract higher intent, add validation fields, or improve follow-up and routing. This funnel approach turns “sales is mad” into a measurable diagnosis.

Attribution and incrementality still matter. If you only use last-click, you may over-credit retargeting or branded search for leads that were actually created by earlier discovery. And even if CPL looks great, incremental thinking asks: did cheaper lead volume create additional qualified pipeline, or did it simply shift effort onto sales by flooding them with unqualified conversations? The limitation is timing—quality often lags by weeks—so short-term dashboards should be interpreted with care and tied to downstream outcomes as they mature.

The core fixes to keep your analytics defensible

Three habits keep you out of the most common traps:

• Write down the counting rules before comparing dashboards: unit of analysis, conversion definition, model, and window.

• Diagnose with the funnel first, then optimize based on the stage where behavior changed.

• Treat ROAS and platform conversions as a view, and use incrementality thinking to avoid scaling false wins.

In the next lesson, you’ll take this further with Next Steps & Learning Path [30 minutes].