44 Digital Marketing Interview Questions, Sorted by Channel

Relevance before authority. A high-authority site in a completely unrelated niche linking to you does less for rankings than a moderate-authority site that's actually topically adjacent, and a pattern of unrelated links can look manipulative if it repeats. Check whether the linking page has real traffic itself, rather than a high domain score with nobody actually visiting it.

Weak answers name a domain authority threshold and stop there. Stronger answers mention checking the linking page's own organic traffic and whether the link sits in editorial content versus a paid directory listing.

A negative keyword list stops your ad from showing for searches containing a given term, the opposite of a normal keyword match. Build the first pass from the search terms report after a week or two of live spend, not from guessing in advance, since real query data surfaces irrelevant matches faster than intuition does.

A specific example worth having ready: a campaign for a paid software product getting clicks from "free [product] alternative" searches. Those clicks cost money and convert at close to zero. Negative-matching them is a five-minute fix that most junior accounts skip for months.

Start from the format with the highest production cost, usually a long-form piece or an original data study, and break it down rather than building five pieces from scratch. A 2,000-word guide becomes a LinkedIn carousel, three standalone posts pulling out specific stats, one email newsletter section, and a short video script covering the single most surprising finding.

The trap: repurposing that's just copy-pasting the same paragraph into five formats. Interviewers want to hear you'd adjust the framing per platform, what works as a caption doesn't work as an email subject line, rather than simply resizing the same asset.

UTM source, medium, and campaign parameters tag a link so analytics can attribute the resulting session correctly. It breaks down constantly in practice: someone screenshots a link into a group chat and the tag disappears, an app strips query parameters on share, or two team members tag the same campaign two different ways and the data fragments into duplicate campaign names.

A UTM naming convention document, shared and actually enforced, fixes most of this. Candidates who've been burned by fragmented UTM data usually have a specific story about a reporting month that looked wrong until someone found the naming mismatch.

SMS for anything time-sensitive and short, a cart abandonment nudge within the hour, an appointment reminder, not for anything that needs explanation or a link-heavy layout. Email for anything that benefits from context: a full onboarding sequence, a newsletter, a win-back message with more than one sentence of reasoning behind it.

The decision usually comes down to urgency and message length more than channel preference. A candidate who defaults to "SMS has higher open rates so use it more" without weighing message fit tends to get pushed on it.

A session is one visit. A user is one person, potentially across many sessions. A pageview is one page load, potentially several within a single session. Reporting "users" when you mean "sessions" understates real traffic. Reporting "pageviews" when you mean "sessions" overstates engagement, since one visitor loading four pages looks like more activity than it represents.

These mix-ups sound minor until a stakeholder catches one in a board deck. Getting caught rounding numbers in the flattering direction, even by accident, erodes trust in every number you report afterward, well beyond the single one that was wrong.

A shared naming convention, enforced through a simple template or a UTM builder everyone actually uses, not a set of guidelines nobody opens. Without it, five marketers create five slightly different tagging patterns for the same campaign type within a month, and reporting fragments into duplicate rows that look like separate campaigns.

A quarterly audit catches drift before it compounds. Cleaning up eighteen months of inconsistent UTMs after the fact costs far more time than a five-minute template would have saved from day one.

SEO is the practice of earning visibility in the unpaid, organic results of a search engine by improving content relevance, site structure, and authority signals like backlinks. It's slow. A new page can take three to six months to see meaningful ranking movement, and results compound over time rather than disappearing the moment you stop investing.

SEM technically covers both, but in practice most people use it to mean paid search, buying placement through Google Ads or Microsoft Ads so you show up above or alongside organic results for the keywords you bid on. You pay per click, and traffic stops the instant you turn the budget off.

The practical answer for when to use each: SEM when you need traffic now, launching a new product, or competing for a keyword where you have no organic footprint yet. SEO when you're building a durable asset, content that keeps converting a year from now without ongoing spend. Most mature marketing teams run both at once, using paid search data, which keywords convert, what ad copy resonates, to inform the SEO content roadmap.

Organic social is content you post to a brand's own channels that reaches followers and whoever the platform's algorithm decides to show it to, without paying for distribution. Paid social is running the same or purpose-built content as an ad, targeting specific audiences by demographic, interest, or behavior, and paying per impression or click to guarantee reach.

The reason organic alone rarely works anymore comes down to algorithm changes. Independent tracking from firms like Socialinsider and Rival IQ has shown organic reach for business pages sitting in the low single digits, often under 5% of followers seeing an unpaid post on Facebook. Instagram and LinkedIn are more generous but still prioritize content that gets early engagement, which creates a chicken-and-egg problem for smaller accounts.

In practice, most brands treat organic as a brand-voice and community channel, and use paid as the actual distribution mechanism. A common pattern is to post organically, watch which posts get engagement in the first hour, then put ad spend behind the ones that are already working rather than boosting everything equally.

A meta description is the HTML tag,

<meta name="description" content="...">

It's not a direct ranking factor. Google has confirmed this repeatedly, and it isn't part of the algorithm's scoring the way title tags, content relevance, or backlinks are. What it does affect is click-through rate. A well-written description that speaks to search intent and gives a reason to click can meaningfully lift CTR, and CTR does appear to act as a secondary relevance signal over time.

So the honest interview answer is that it won't move you from position eight to position three, but a page with a strong description competing against generic auto-generated snippets can win more clicks at the same rank, which indirectly helps rankings and definitely helps traffic.

CTR, click-through rate, measures how many people who saw an ad or search result clicked it, calculated as clicks divided by impressions. Conversion rate measures how many of the people who landed on your page after clicking actually completed the action you wanted, a purchase, a signup, a form fill, calculated as conversions divided by clicks or sessions.

They describe completely different parts of the funnel, and a strong number on one paired with a weak number on the other usually points to a specific problem. High CTR with low conversion rate typically means the ad or headline is over-promising or attracting the wrong audience, someone clicks because the copy is compelling but the landing page doesn't deliver what was promised. Low CTR with a high conversion rate on the traffic that does click often means the offer is fine but the creative or targeting isn't reaching enough of the right people.

I've seen teams optimize an ad purely for CTR and end up training the algorithm to serve clickbait-style creative to bargain hunters who bounce immediately. Tracking both, and treating cost per conversion as the real north star, keeps you from winning the vanity metric and losing the business outcome.

Retargeting, or remarketing, is showing ads to people who already visited your site or app rather than to a cold audience. Someone who added a product to cart or read a pricing page is a much warmer prospect than someone who's never heard of you, so you follow up with ads reminding them to come back.

Technically it works through a tracking pixel or tag, a small snippet of JavaScript from Google Ads, Meta, or a platform like Criteo, placed on your site. When a visitor loads a page, that pixel drops a cookie, or increasingly uses server-side tagging and first-party data matching, that adds them to an audience list. The ad platform then matches that list against its logged-in users on other sites or apps and serves your ads to them.

Worth mentioning in an interview: third-party cookie retargeting has gotten less reliable. Safari and Firefox block third-party cookies by default, and Chrome has been phasing them out too. That's pushed serious retargeting programs toward first-party data, building audiences from CRM or logged-in user data and uploading those as customer match lists, and toward Meta's and Google's own logged-in ecosystems, which don't depend on the same cookie technology.

Reach is the number of unique people who saw a piece of content. Impressions is the total number of times it was displayed, including repeat views by the same person. If one person sees your ad three times, that's a reach of one and three impressions.

Neither is inherently the metric that matters, they answer different questions. Reach tells you how big your audience footprint was, useful for awareness goals where you're trying to get in front of as many distinct people as possible without over-saturating any one of them. Impressions tell you about frequency, and frequency drives ad fatigue. If impressions climb much faster than reach, the same people are seeing your ad over and over, and past a certain point, a common rule of thumb is somewhere around three to five exposures a week for a given campaign, performance drops and cost per result climbs because you're paying to re-show an ad to someone who's already tuned it out.

In practice I watch frequency, impressions divided by reach, as the real signal. Low frequency with flat reach means there's more audience to find. Rising frequency with flattening reach is the cue to refresh creative or expand targeting before the campaign burns out.

CAC, customer acquisition cost, is total sales and marketing spend over a period divided by the number of new customers acquired in that same period. If you spent $50,000 across ads, tools, and marketing salaries in a quarter and closed 100 new customers, your CAC is $500.

The most common mistake is scoping the spend inconsistently. Some teams calculate CAC using only paid media spend, which flatters the number but ignores the salaries of the marketers and salespeople actually running the motion, the tools, and any content or agency costs. A fully loaded CAC that includes headcount and tooling will always look worse than a media-only number, and comparing your media-only CAC to a competitor's fully loaded CAC in a board deck is a good way to make a bad decision look good.

The second mistake is measuring CAC in isolation from lifetime value. A $500 CAC is fine for a product with a $5,000 average lifetime value and terrible for one with $600 lifetime value. Any solid answer to a CAC question should mention that it's only meaningful next to LTV, payback period, how many months of revenue it takes to recover the acquisition cost, and how it trends as channels saturate.

An MQL, marketing qualified lead, is a prospect who's shown enough interest and fit, downloading a gated report, attending a webinar, hitting a lead score threshold, to be worth marketing's continued attention, but hasn't been vetted for budget, authority, or immediate intent to buy. An SQL, sales qualified lead, is a lead that's been reviewed, usually by an SDR or sales rep, and confirmed to have a real need, budget, and timeline, meaning it's worth a sales rep's time to actively work the deal.

The handoff point is where a lot of marketing and sales tension actually lives. In a healthy setup, marketing and sales agree in advance on the criteria, often in a signed SLA, so an MQL that meets a lead score threshold or fills out a demo request form gets routed to an SDR within a set window, commonly under 24 hours since response time correlates heavily with conversion, and the SDR confirms it's an SQL after a qualifying call, typically using a framework like BANT or MEDDIC.

Where this breaks down in real companies is when marketing is measured on MQL volume and sales is measured on closed revenue. Marketing has an incentive to loosen the MQL definition to hit a number, sales gets flooded with leads that don't convert, and starts ignoring the queue entirely. The fix isn't a better dashboard, it's realigning both teams' incentives around a shared metric further down the funnel, like SQL-to-close rate or pipeline generated, not raw lead count.

Traffic is the beginner's answer. A more useful one tracks assisted conversions from organic content, pipeline influenced in the CRM, and a comparison of organic versus paid traffic for equivalent intent levels, since a page pulling in high-intent visitors at zero marginal cost is worth more than its raw session count suggests.

Weak answers stop at sessions and time on page. Strong answers connect a specific content initiative to a specific downstream number, even an imperfect one, rather than treating "traffic went up" as the finish line.

Depends entirely on the content type, and that's the honest answer even though it sounds like a dodge. Thin, purely informational content, "what is email marketing," "how to write a subject line," is exactly the kind of page Google's helpful content systems increasingly deindex regardless of whether a human or a model wrote it. Original research, a documented test, a genuine opinion backed by specifics, still performs, because nothing else on the page can be swapped in from a competitor's site.

One opinion that could be wrong: most marketing teams are using AI content on exactly the wrong pages, high-volume informational posts, and skipping it on the pages where speed would actually help, internal briefs, first-draft outlines, meeting notes. That's backwards from where the deindexing risk actually sits.

Search intent mismatch first. A page can rank well because Google decided it answers the query, without that query actually matching what the page is selling. If the keyword is informational and the page is a hard product pitch, high rank and low conversion is expected, not a bug.

If intent does match, it's a page-level problem: load time, a call to action buried below three scrolls, a mobile layout that breaks on the exact device your traffic actually uses. Pull the page speed and mobile usability reports before touching the copy. A slow page loses conversions long before anyone reads the headline.

Basic answer: shift budget toward the highest ROAS campaign until it's maxed out. Reasonable starting point, wrong ending point, since most channels hit diminishing returns from audience saturation well before they're "maxed out" on paper.

Stronger answers bring up marginal ROAS, not average ROAS, meaning the return on the next dollar spent rather than the return on the dollars already spent. A campaign with excellent average ROAS can have terrible marginal ROAS if its best audience segment is already saturated.

Quality Score still exists and still affects cost per click, but it matters less as a direct optimization lever than it used to for accounts running Smart Bidding, since the algorithm is already optimizing toward conversions using signals well beyond the Quality Score inputs. Candidates who spend the whole answer on ad relevance and expected click-through rate, without mentioning that shift, read as behind on how the platform actually runs in 2026.

The honest nuance: Quality Score is still a useful diagnostic. A low score usually points to a real landing page or relevance problem worth fixing, it's just not the optimization target it was five years ago.

Say so directly. A Google Skillshop certification proves you know how bidding strategies, match types, and conversion tracking are supposed to work. It doesn't prove you've made a real budget call under pressure, and interviewers know the difference. Certifications test what you know. Interviews test what you'd do. Those are different skills, and conflating them in an answer is an easy way to get caught on a follow-up.

If your real experience is a personal account, a class project, or supporting a senior campaign manager rather than owning the account, name that specifically and talk about the decision you were closest to. A small, honest example beats an inflated one the moment a follow-up asks for detail.

Structure by product category and margin, not one catalog-wide campaign. A 40-SKU catalog with wildly different margins and conversion rates performs badly under one shared budget and bid strategy, since the algorithm optimizes toward whatever converts most easily, usually the cheapest items, and starves the higher-margin ones of spend.

A reasonable split: group by category, set separate budgets and target ROAS per group, and use one shared negative keyword list across the account instead of duplicating exclusions in six places. Watch for cannibalization between similar product ad groups bidding against each other on the same query.

Audience research before platform selection, in that order. A common mistake is starting with "I'd start on Instagram because it's visual," which picks the channel before knowing anything about where the actual audience spends time or what they're there to do.

The stronger sequence: define who the audience is, find where they already have conversations relevant to the brand's category, then pick platforms and formats. Sometimes that answer is boring, LinkedIn and a niche Slack community, not exciting, but it's the honest one for a B2B brand with a small team.

Follower count rarely belongs in a serious answer here. Stronger metrics: dark social attribution (traffic that shows up as direct because it came from a screenshot or a DM share, not a trackable click), UTM-tagged pipeline or revenue contribution, and share of voice against named competitors during a specific campaign window.

The honest gap worth naming: dark social is genuinely hard to measure precisely, and candidates who claim they've solved it completely usually haven't. Acknowledging the limitation while describing a workaround, post-click surveys, branded search lift, reads better than pretending the attribution problem doesn't exist.

One platform done consistently beats three platforms done sporadically, and a small team almost never has the bandwidth to do three well. Pick the platform where the target buyer already spends professional attention, usually LinkedIn for most B2B categories, and treat any second platform as distribution for content already built for the first, not a separate content operation.

Candidates who name three or four platforms as an ideal state without acknowledging the resourcing trade-off tend to read as inexperienced with small-team constraints specifically.

Respond fast, specifically, and without a copy-pasted brand voice line that ignores what was actually said. A generic "we hear you, thanks for your feedback" under a specific, valid complaint reads worse than silence to anyone else reading the thread.

If the complaint is accurate, say so plainly and describe the fix or the timeline. If it's inaccurate, correct it factually without getting defensive. Escalate anything involving a safety or legal issue immediately rather than trying to handle it solo in the comments.

One opinion that could be wrong: short-form video is overrated as a primary B2B demand-gen channel for mid-market companies specifically, not because it doesn't get views, but because the production cost per piece is high relative to the attribution most teams can actually prove back to pipeline.

It works well for brand awareness and for companies with a strong, credible on-camera presence already. It works badly as the first channel a small team invests in before nailing distribution basics. Someone could reasonably argue the opposite, and a candidate who takes either side with specifics is in better shape than one who avoids picking a side at all.

Name the specific piece and the specific miss, not a vague "it didn't resonate." Maybe the promotion channel didn't match the audience the content was written for, or the headline tested well in isolation but didn't match what the body actually delivered, creating a bounce.

What separates a strong answer here is naming what changed in the process afterward, not just the postmortem. A content brief template that now includes an explicit audience-and-channel match check before writing starts is a lot more credible than "we just try harder now."

Start from behavior and firmographic data together, not one or the other alone. Job title tells you who someone is; what they've actually clicked, downloaded, or ignored tells you what they're ready to hear next. A persona segment that ignores engagement history sends the same nurture sequence to someone who opened every email and someone who's opened none of the last ten.

A concrete example worth having ready: three personas at different stages of a buying committee, an economic buyer, a technical evaluator, and an end user, each need a different first email even if all three downloaded the same whitepaper.

Anchor the sequence to a single first action, not a generic welcome message. If the product's core value shows up when a user completes one specific setup step, the first two or three emails should focus entirely on getting them to that step, not a general tour of every feature.

A common mistake: a five-email sequence that front-loads feature announcements before the user has done anything meaningful in the product yet. Sequence the emails around what the user has and hasn't done, triggered by behavior rather than by day-since-signup alone.

Be honest that a 4,000-person list makes standard A/B significance hard to hit on a single send, especially for a subtle subject-line variation. Either test a bigger, more obvious difference, not two similar phrasings but two genuinely different angles, or pool results across several sends rather than trusting one send's outcome.

Candidates who confidently claim statistical significance on a tiny one-off test without checking sample size math are usually bluffing. Better to say the list is too small for a clean single-send result and describe the workaround.

Personalization built on data you have explicit consent to use is fine. Personalization that relies on inferred or purchased data the recipient never agreed to share crosses into the territory GDPR and similar regulations exist to prevent, and it tends to feel invasive even where it's technically legal.

Practical answer: maintain a clear record of consent basis per contact, honor unsubscribe and data deletion requests within the required window, not eventually, and default to less personalization rather than more when the data's origin is unclear.

Basic formula: average revenue per account divided by churn rate. Fine as a starting point, wrong as a final answer, since it treats every customer as equally likely to churn and equally valuable, which is rarely true.

Stronger candidates segment by acquisition channel and cohort before trusting a blended LTV number. A channel bringing in customers with a 40% higher churn rate has a materially different LTV even at an identical average revenue per account. Blending everything together hides which channels are actually worth the acquisition spend.

Group users by acquisition month and channel, then track a specific downstream behavior, activation, retention at 30 and 90 days, revenue per cohort, rather than looking at aggregate channel volume alone. A channel can hold steady or grow in raw signups while the quality of who it's bringing in quietly degrades.

The signal to watch for: a cohort acquired six months ago retaining noticeably worse than one acquired a year ago from the same channel, even with similar acquisition cost. That's the pattern that shows up in cohort data well before it shows up in a top-line churn number.

Check the definition of "converting" before assuming sales is wrong. A campaign optimized toward form fills can convert well on that metric while producing leads that never had real budget or authority, since the ad targeting or the offer attracted the wrong intent even though the form-fill number looks healthy.

Pull a sample of the actual leads and talk to sales about specific ones, not the aggregate number. Sometimes the fix is the targeting. Sometimes it's the offer. Sometimes, less comfortably, sales is missing something in how they're qualifying, and a good answer doesn't assume the fault sits entirely on one side before checking.

Lead with the metric the CFO already trusts, then bridge to marketing's contribution rather than opening with impressions or brand lift, which reads as marketing defending its existence rather than reporting results. Cost per acquired customer against lifetime value, and the payback period on spend, translate directly into language a CFO already uses for every other budget line.

Marketing-influenced pipeline is a real number worth reporting, but presenting it as equivalent to closed-won revenue is where marketing loses credibility with finance fastest. Naming the difference explicitly, here's what we influenced, here's what we can defensibly claim we drove, tends to build more trust over time than inflating the softer number.

Not a reconsideration request. Those apply only to manual actions, not algorithmic ranking changes, and asking for one after a broad core update is the fastest way to signal you don't understand the difference. Pull Google Search Console first, filter to the update's date range, and see which specific pages and query types actually lost traffic instead of assuming the whole site dropped evenly.

Then compare the pages that lost the most against Google's quality rater guidelines: thin informational content, aggregator pages with no original point of view, anything that reads like it was written to rank rather than to answer the question. Core updates rarely punish a domain evenly. They tend to concentrate on specific page types, and finding which type is the actual diagnostic work.

Split the funnel into three separate problems before touching spend: ad performance, landing page performance, and offer performance. A high CPA with a strong click-through rate and a weak landing page conversion rate needs a different fix than a high CPA with poor click-through and a fine landing page.

A specific trap worth naming: a low cost-per-lead from the ads themselves, but 80% of leads coming from one underperforming ad group that's technically hitting its CPL target on volume alone. The blended number looks fine. The individual ad group is quietly wasting budget. Segment before you conclude anything.

Apple's App Tracking Transparency framework cut the reliability of device-level, last-click attribution for a meaningful share of iOS traffic, since users who decline tracking can't be matched across sessions the way they used to be. Relying on Google Ads or Meta's own reported conversions alone means trusting numbers a platform has an incentive to report favorably.

Stronger candidates describe triangulating with blended metrics instead: correlating paid spend against branded search volume and direct traffic over time, using post-purchase surveys, or building a marketing mix model that doesn't depend on individual-level tracking at all. None of these are perfect substitutes, and a good candidate says that plainly instead of pretending attribution is a solved problem.

Five likely causes, checked against a timeline rather than guessed at: an algorithm change, a shift in content type, a change in posting frequency, audience fatigue from repetitive formats, or a competitor absorbing attention with a specific campaign. The diagnostic work is lining up the drop's exact start date against each of those.

Weak answers pick one cause and stop, usually "the algorithm changed," without checking whether posting frequency or format also shifted in the same window. Strong answers rule causes out methodically instead of pattern-matching to whatever explanation is currently trending.

Few companies solve this cleanly, and saying so upfront is more credible than pretending there's a tidy formula. A practical approach: first-touch and last-touch attribution from UTM data, supplemented by a pipeline-influenced report from the CRM that credits any deal touched by content at any stage, including touches long before the final form fill.

The follow-up worth preparing for: what do you do when the pipeline-influenced number looks great but a specific piece of content clearly didn't drive it, someone just happened to visit the blog once during a six-month sales cycle. Good candidates flag that noise instead of taking the flattering number at face value.

Sender reputation and authentication records, SPF, DKIM, DMARC, before assuming the content or subject lines caused it. A sudden overnight drop with no campaign change usually points to an infrastructure issue: a shared IP got flagged, a domain authentication record lapsed, or a spam trap got hit in a recent send.

Major mailbox providers have gotten stricter about authentication requirements for anyone sending in volume, and accounts running on loose or missing DMARC records are exactly the ones that get caught when enforcement tightens. Weak answers jump straight to rewriting subject lines. Deliverability problems and content problems produce a similar symptom, fewer opens, from completely different causes, and treating them the same wastes a week chasing the wrong fix.

Combine explicit data (job title, company size, stated intent from a form) with implicit behavior (page visits, email engagement, content downloads), weighted toward whichever signals have actually correlated with closed-won deals historically, not toward whichever signals are easiest to collect.

You know it's working when sales stops complaining that "MQLs are garbage." Until that complaint changes shape, the model needs revisiting. A lead scoring system that marketing loves and sales ignores isn't actually working, regardless of the internal reporting.

Name a specific breakage, not a vague "it was a hard migration." Common ones: historical engagement data that doesn't map cleanly to the new platform's scoring model, workflows that fire differently because the new tool's trigger logic works on a different event schema, or a segment definition that pulls a different, wrong list of contacts because a field name changed.

The stronger part of the answer is what you did about it: a parallel-run period where both systems sent nothing live while you validated segment counts matched, or a phased cutover by segment instead of an all-at-once switch. "We just switched over one weekend" tends to be the answer of someone who hasn't actually done this.

Sequence matters: hypothesis first, sample size calculation before launch (not after you've already collected some data and are wondering if it's enough), an even traffic split, and a pre-defined minimum detectable effect so you're not chasing statistical significance on a difference too small to matter for the business.

The most common failure candidates describe without realizing it's a failure: reaching statistical significance on a 0.3 percentage point conversion lift that took six weeks to detect. Technically significant, practically irrelevant. Defining the minimum effect worth caring about upfront avoids that trap.

Models disagree constantly, and pretending they don't is a tell that a candidate hasn't actually managed a real attribution stack. If the CRM credits email and the ad platform credits paid search for what looks like the same conversion, someone still has to make a budget call despite the conflict.

The better answer describes a specific resolution: a blended view weighting recent, high-confidence touches more than a distant first click, or a documented rule for which model wins in a disagreement, rather than re-litigating attribution philosophy every time two reports don't match.

First-party data becomes the foundation here, not a nice-to-have: logged-in user IDs, a hashed email match across devices where you have consent, and server-side tagging that doesn't rely entirely on a third-party cookie surviving a browser's privacy settings.

Honest gap worth naming: no first-party solution fully replaces what third-party cookies used to do for anonymous cross-device matching, and a candidate who claims a perfect substitute is overselling. A directionally useful, consent-based approach beats a claim of full parity with the old system.

An incrementality test holds out a comparable audience from a channel entirely, a geo holdout, a PSA-versus-ad-exposed split, and measures the actual lift against that baseline rather than trusting a platform's self-reported attributed conversions, which have an obvious incentive to look favorable.

Run one when a channel's reported ROAS looks suspiciously strong and the budget riding on it justifies the cost of a holdout, real conversions you're deliberately not serving ads to, in exchange for a cleaner read on whether the channel is driving incremental revenue or just claiming credit for demand that existed anyway.

Start with search intent, not volume. Group the keywords you find into informational, navigational, and transactional buckets, then be honest that a lot of the transactional terms won't be winnable in the first six months against sites with three years of backlink history. A believable timeline matters more than an ambitious one.

Add competitor gap analysis: which queries do the top five ranking pages answer that your planned content doesn't cover yet. Candidates who bring only a keyword list, no intent grouping, no gap analysis, read as someone who's used a tool once rather than someone who plans content around it every week.

Segment by why they likely lapsed, not one blanket win-back email to everyone who's gone quiet. Someone who hit a pricing objection needs a different message than someone who never finished onboarding and probably forgot the product exists.

A specific, honest detail worth including: win-back campaigns typically convert at a small fraction of the rate of a fresh lead, and setting that expectation with stakeholders upfront matters as much as running the campaign well.

Define the conversion events first, before touching the interface. Building a data stream and enhanced measurement setup without knowing what "success" means for this launch produces a dashboard full of numbers nobody can act on. Then configure the data stream, decide which events count as key events (GA4's replacement for Universal Analytics' "goals" terminology, which trips up anyone who learned analytics on the old platform), and set up Google Tag Manager so marketing can adjust tracking without filing an engineering ticket for every change.

Search Engine Land covered Google's July 1, 2023 sunset of Universal Analytics when it happened, and it's still a useful marker for candidates: anyone who's only ever configured GA4 from a blank slate, never migrated legacy UA goals and audiences over, should say so rather than claiming migration experience they don't actually have.

Start with the questions the dashboard needs to answer, not the data that happens to be available. A dashboard built backward from whatever fields exist in the warehouse tends to have forty metrics and answer none of the three questions an executive actually asks in a meeting.

Seven primary KPIs is a reasonable ceiling for the top-level view, with anything else available as a drill-down rather than cluttering the main screen. More than that and executives stop reading the dashboard at all and just ask you to summarize it verbally, which defeats the point of building one.