Analytics and metrics, plainly defined

A KPI, or key performance indicator, is a metric a business chooses as an objective because it shows progress toward a goal that matters, such as qualified leads, sales or margin. Not every metric deserves the role: engagement signals are useful, but none is a universal KPI, and the right set depends on product, market and sales process. KPIs work best in pairs that should move together, so traffic growing while conversions fall shows where to look. A badly framed target such as "double the traffic" tends to be met with loosely related queries.

More: Why Luxury E-commerce Breaks Every Rule of Conversion Optimization

Conversion rate is the share of visitors who complete a goal, such as an order, a form request or a call, out of all visitors to a page or site. It turns ad spend into the cost of a lead: at the same $2 click, a page converting at 4% brings leads at $50, one at 2% at $100. In our projects design changes alone have raised it by 8% to 17%. In luxury the key conversion may be a boutique visit days after the online session.

More: Design You Don't Notice: Why the Best Design Disappears

A qualified lead, or MQL (marketing qualified lead), is an inquiry that fits the business by need, budget, size or timeline, as opposed to any contact who fills in a form. Counting all leads hides the difference: organic traffic up 80% with qualified leads up 2% means the site attracts the wrong people. A more honest report tracks organic leads, calls, booked consultations and revenue per landing page. An AI website assistant can do the first qualification by project type, size, budget, timeline and stage.

More: SEO, AEO, and GEO in 2026: What Actually Matters to Google, and What Your Marketer's Report Should Show

Lead-to-sale conversion, or close rate, is the share of leads that become paying customers within a given period. Without it, cost per lead says nothing about the cost of a customer, because a cheap lead that never buys is the most expensive one. Sales handling often matters as much as the ad: US dealership data cited in our analysis of rising CPC shows a 16.2% close rate within 30 days, with 43.2% of leads mishandled.

More: Rising CPC Isn't the Problem. Here's What Is

Customer acquisition cost, or CAC, is the total marketing and sales spend needed to win one paying customer. It sits on the third level of advertising economics: the first is the cost of attention (CPC), the second the cost of a lead (CPL and lead-to-sale conversion), the third the result (CAC, gross margin, repeat purchases and LTV). Most arguments happen about clicks, while most money is made or lost on this third level. Organic search is one way to lower CAC over time.

More: Rising CPC Isn't the Problem. Here's What Is

Customer lifetime value, or LTV, is the total revenue or margin a business earns from one customer over the whole relationship, repeat purchases included. Set against CAC and gross margin, it shows whether a channel pays: a cheap click that never becomes a sale is expensive, while a costly customer who keeps buying can be the best result in the account. LTV is why retention and repeat orders belong in the same report as ad spend.

More: Rising CPC Isn't the Problem. Here's What Is

Vanity metrics are easy-to-measure numbers, such as likes, followers, reach, page views or articles published, that look like progress but do not show whether marketing makes money. Business metrics sit further down the path: profile visits, link clicks, messages, form fills, cost per lead, leads attributed in the CRM and revenue. The two can diverge sharply. One client's content brought 34 direct sales inquiries without moving the like count, and one owner judged a site by falling page views while forms and calls held steady.

More: How to Build a Social Media Strategy That Actually Generates Leads

Bounce rate is the share of visits that leave a page without interacting with it. A rate above 60% is one of the signals that a site may need a redesign, and a high rate on mobile alone often points to slow loading. The figure can also look normal while small moments of confusion stay hidden, so it is read next to session recordings. Bounce rate on product pages is one of the store signals checked in a usability audit, which with us starts from $1,500.

More: Do You Need a Website Redesign? Start With a Usability Audit

An engaged session is a visit with meaningful interaction; in Google Analytics 4 it is a session that lasts longer than 10 seconds, includes a key event or views at least two pages. It replaced bounce as the main quality signal in GA4. Together with returning users, product views, scroll depth and video views, it helps read long buying journeys where the sale comes weeks later. A luxury store can even make a third return to the same product a tracked signal of intent.

More: Why Luxury E-commerce Breaks Every Rule of Conversion Optimization

Average session duration, or time on site, is the mean length of visits over a period. A high number can be good or bad: deep reading and comparison look the same as a visitor lost in the navigation. That is why it works as a secondary metric next to conversion, not as a goal. After one jewellery site rebuild, average time on site went from 30 seconds to just over six minutes, and monthly visitors grew from about 500 to 15,000 in six months.

More: Why Luxury E-commerce Breaks Every Rule of Conversion Optimization

Organic CTR (click-through rate) is the share of a page's impressions in unpaid search results that turn into clicks, as reported in Google Search Console. It shows whether the title and snippet earn the click once the page is visible: rising impressions with flat clicks point to a snippet that does not. CTR also depends on site structure and on what the result promises. On one of our store projects, information architecture and UX work took organic CTR from 0.58% to over 2%, with recent weeks above 3%.

More: SHTAYER: Building a Brand for a New Generation

An activation metric is the one measure, agreed before launch, that defines when a new user counts as having reached the product's value, such as a first project created or a first order placed. Without it a team counts sign-ups and still cannot tell whether the product works. Users who sign up but do not activate or return are a sign the product needs design work, not more traffic. Research cited in our MVP articles found that about 80% of features in the average product are almost never used.

More: Architecting the MVP: What Marketing Owes a Product Before Launch

Cohort analysis compares groups of customers who started in the same period, for example everyone who first bought in March, and follows how each group behaves over time: repeat purchases, retention and revenue. It separates real improvement from growth that only hides churn, because new customers can mask old ones leaving. Investors read it closely: cohorts that hold up predictably over time can support the same confidence in a valuation as signed contracts.

More: Why Tech Companies Get Valued on Different Math Than Everyone Else

Traffic concentration is how much of a site's organic traffic depends on a small number of pages. If 70% of visits come from ten URLs, the business is fragile: one algorithm update or a migration mistake on those pages can cut a large share of its leads at once. It is one of the checks when buying a digital business, next to who owns the accounts and how the traffic converts. A spread across many pages and queries is a sign of a sturdier asset.

More: How to Audit a Website or Digital Product Before You Buy It

Containment rate is the share of conversations an AI assistant or chatbot closes without handing them to a human. On its own it is a poor metric, because a visitor who simply gives up also counts as contained. A useful report separates AI involvement, automated resolution, handoff to a person, abandonment, qualified leads, recommendations, bookings and transactions. A claim of "80% resolution" means little until someone defines what counts as resolved.

More: AI Website Assistants Are Not Chatbots. They Are a New Business Interface.

Seasonality is the regular rise and fall of demand across the year, driven by holidays, weather, budgets or the industry calendar. It is why traffic, cost per lead and sales are compared year over year, the same period against the same period, rather than month over month. A drop that looks alarming next to last month is often normal next to last year, so seasonality is checked first, before a change is blamed on ads, a redesign or an algorithm update.

More: Why Your Cost Per Lead Keeps Going Up in 2026 (10 Fixes)

Correlation versus causation is the difference between two numbers moving together and one actually causing the other. Marketing reports mix them up constantly: organic growth and a falling CPC can be observed together in an account without any proven link between them. Studies are no exception: an Ahrefs study found brand mentions correlated with AI visibility at 0.664, a strong correlation but not proof. Causation needs a test, such as an A/B split or a before-and-after comparison with everything else held steady.

More: Rising CPC Isn't the Problem. Here's What Is

A conversion funnel is the sequence of steps a visitor passes from first contact to a sale: for a store landing page, category, product, cart and checkout; for SaaS registration, activation and retention. The full chain runs from rankings and impressions through clicks, sessions, conversions and qualified leads to revenue and profit. A funnel report shows where people leave but not why; walking the customer's real path on a phone usually does. Every step should be tested before money goes into ads.

More: Traffic But No Leads? Your Customer Is Doing Math You Never See

Conversion rate optimization, or CRO, is systematic work on raising the share of visitors who take a target action, based on analytics, heatmaps, session recordings and tests rather than opinions. For one retailer a redesign built on an audit, heatmaps and analytics raised conversion by 17% in three months: rarely clicked header links went, price and buy button stayed on screen, and mobile, about 76% of orders, got priority. Standard CRO assumes easier buying means more buying, but considered purchases over $20,000 need information first, and luxury must also protect perception.

More: Why Luxury E-commerce Breaks Every Rule of Conversion Optimization

A/B testing is a method of showing two versions of a page, ad or email to randomly split traffic and comparing which performs better on a chosen goal. Only one variable should change at a time, such as the creative, the audience or the offer, and a test should run at least 7 days, ideally 14, to cover weekly patterns. Each variant gets only half the visitors, so with a modest budget it pays to test large differences first: the offer, the headline, the page length.

More: Landing Page Development

Session recordings are replays of real visits that show what one person actually did on a site: where they moved, clicked, hesitated and left. They reveal what aggregate reports cannot, such as customers relying on breadcrumbs that a new design had removed, or repeated clicks on a catalogue that kept collapsing. On one store, recordings guided product page and catalogue changes over five years. They inform design judgment rather than replace it.

More: Do You Need a Website Redesign? Start With a Usability Audit

Microsoft Clarity is a free behaviour analytics tool that records sessions, builds click and scroll heatmaps and flags rage clicks and dead clicks. It shows how people actually use a site at no license cost, which makes it a practical first step of any usability review. We use it before deciding whether a site needs an upgrade or a full redesign.

More: Do You Need a Website Redesign? Start With a Usability Audit

Rage clicks are rapid repeated clicks on an element that does not respond as the visitor expects; dead clicks are clicks on elements that are not interactive at all. Both are clear signs of frustration and point straight to broken buttons, misleading links or elements that only look clickable. Behaviour tools such as Microsoft Clarity flag them automatically. Their near absence is useful evidence too: on one studio site we checked, almost none appeared, so a redesign was not justified.

More: Do You Need a Website Redesign? Start With a Usability Audit

Scroll depth is how far down a page visitors get, usually measured as a percentage of the page length. It shows whether the content that sells, such as prices, proof or the request form, is ever seen or sits below the point where most people stop. In analytics it is tracked as an engagement event, and during a usability audit it is shown as a scroll map over the page itself.

More: Why Luxury E-commerce Breaks Every Rule of Conversion Optimization

Form monitoring is regular checking that a site's inquiry forms work and receive requests: daily submission counts, and form starts compared with finished sends. A sudden drop to zero is usually a technical fault, not a fall in demand, while a wide gap between starts and finishes points to a form that asks too much. The cost of skipping it is real: broken inquiry forms at one institute cost 84 inquiries over two months.

More: Landing Page Development

A before-and-after comparison measures the effect of a change such as a redesign against a recorded baseline instead of impressions. Before the work, record 3 to 6 months of inquiries by form, call and email, entry pages, sources, mobile versus desktop conversion and form starts versus finishes. After launch, compare the same 60 to 90 day period with the same tracking. Nothing on the site should change until tracking records inquiries, otherwise the result cannot be proven either way.

More: Website Redesign Seattle: Why Most Business Websites Fail to Generate Leads (And How to Fix It)

Event tracking is recording specific user actions, such as product views, scroll depth, video completion, return visits or form sends, as events in an analytics system. Events turn a count of page views into a picture of behaviour and feed funnels and conversions. For an MVP the minimum is events on each funnel step, UTM tags on every inbound link and one activation metric, about two or three days of work during the build. A custom event such as zero_results can turn failed filter combinations in a store into a list of what the assortment lacks.

More: Architecting the MVP: What Marketing Owes a Product Before Launch

Conversion tracking is recording completed goals, such as orders, inquiries and calls, together with the source that produced each of them. Without it, ad and SEO spend cannot be judged, and a site rebuilt without reconnecting analytics keeps receiving leads that nobody can trace. In e-commerce it has to be set up correctly at launch, not added later. The first fully tracked month often surprises: on one project it exposed a tracking discrepancy and bot traffic that had been invisible before.

More: What Actually Makes E-Commerce Development Expensive

UTM parameters are tags added to a link, such as utm_source, utm_medium and utm_campaign, that tell an analytics system where a visit came from. They make email, social, partner and offline links visible as separate sources instead of a vague line of direct traffic. They need discipline: the same naming on every inbound link. For partnerships, set unique tags, promo codes or affiliate links for each partner before launch and review the results at 30 and 90 days.

More: Architecting the MVP: What Marketing Owes a Product Before Launch

Google Tag Manager, or GTM, is Google's free tool for adding and managing tracking tags, such as analytics, ad pixels and conversion events, without changing the site code each time. It keeps all tracking in one container with versions and a preview mode, and it is one route to a server-side Conversions API setup for Meta ads. A common problem is ownership: the container was often created by a past SEO specialist or developer and left without an owner.

More: What You Own vs What You're Renting: The Digital Ownership Audit

Marketing attribution is the method of assigning leads and sales to the channels and campaigns that produced them, whether by last click, first click or a model that shares credit across the path. It decides where the next budget goes, so broken attribution is expensive: a changed lead source or a broken integration after a migration makes the reports meaningless. A practical rule from our work: a test budget that cannot produce one attributable sale in three to four weeks is telling you something.

More: Architecting the MVP: What Marketing Owes a Product Before Launch

Assisted conversions are conversions in which a channel took part earlier in the customer's path without being the last click before the purchase or request. They show the value of channels that start or warm up a decision, such as social media or content, which last-click reports undervalue. Read together with qualified leads, cost per lead and revenue, they help justify keeping a channel that rarely closes a sale on its own.

More: Social Media Strategy Is Not a Content Calendar

Call tracking is a way of recording which source produced each phone call, usually by showing different phone numbers to visitors from different channels and logging the calls. For businesses where many customers call instead of filling in a form, a large share of leads has no source in reports without it. It needs care during a redesign: a rebuild that forgets to reconnect it leaves calls arriving with no visible source.

More: Website Redesign Seattle: Why Most Business Websites Fail to Generate Leads (And How to Fix It)

Offline conversion tracking ties a sale that happens outside the site, such as a boutique visit, a consultation or a signed contract, back to the online inquiry or ad click that started it, usually through the CRM. Without it, channels that drive considered and luxury purchases look unprofitable, because the sale never appears online. It is the hardest part of a luxury analytics setup and is rarely fully automatic: it needs a disciplined CRM process, not just a script.

More: Why Luxury E-commerce Breaks Every Rule of Conversion Optimization

Bot traffic, or invalid traffic, is visits and ad interactions made by automated programs rather than people. It inflates reach, clicks and engagement, distorts conversion rates and wastes ad budget; industry estimates put it at roughly 40% of ad traffic on major platforms even after filtering. It often becomes visible only once tracking is done properly: on one of our projects, the first fully tracked month revealed bot traffic that earlier reports had counted as visitors.

More: Is Social Media Marketing Dead in 2026? What the Data Actually Says

A measurement plan is a short document that defines, before a site or feature is built, what will be measured, how, and what counts as success. It lists the goals, the events and conversions, the sources and the reports that will show them. AI tools can now generate tracking code, but someone still has to design the model behind it. For an AI feature, the baseline is agreed before development starts: handling time, error rate or the number of manual steps it should remove.

More: Claude Can Build the Website. It Can't Build the Understanding. Not Yet.

First-party data is data a business collects itself from its own site, apps, CRM, sales and customers, as opposed to estimates bought or scraped from third parties. It is the most reliable basis for decisions: a reported metric is a fact, a third-party traffic estimate is an estimate, and an AI guess at a competitor's conversion rate is an inference dressed as a number. Real projects with before-and-after data are also the strongest material for content and case studies.

More: AI Can Build Your Analytics System in Two Days. Here's What That Number Actually Describes.

Analytics account ownership is the question of who actually controls Google Analytics, Search Console, Tag Manager, the ad accounts and the conversion events. If a former employee or contractor created them on a personal email, years of history can be lost when that person leaves. The safe setup is the business as owner, with agencies and freelancers working through user access. When hiring an SEO specialist, Google advises granting read access to Search Console rather than write access.

More: What You Own vs What You're Renting: The Digital Ownership Audit

AI referral traffic is visits that come to a site from links in AI assistants such as ChatGPT and Perplexity. Retailers now report it as its own line, separate from organic and paid search, because it behaves differently. Figures cited in our articles: AI-referred sessions grew 527% year over year through mid-2025 and are claimed to convert at 13 times the rate of organic traffic. In analytics it often lands in referral or direct traffic unless it is grouped as a separate channel.

More: SEO Gets You Found. AEO Gets You Recommended. Your Commerce Infrastructure Gets You Bought.

Google Analytics 4, or GA4, is Google's event-based web analytics platform: every page view, click, scroll or purchase is recorded as an event, and the events that matter are marked as key events, its name for conversions. Engagement events such as product views, scroll depth, video completion and return visits feed its reports. Its behaviour flows, together with heatmaps, once showed a retailer that most header links were barely clicked. Before any work on it, check who owns the property: it belongs to the business's assets.

More: Why Luxury E-commerce Breaks Every Rule of Conversion Optimization

Google Search Console is Google's free tool for site owners that shows how a site appears in Google Search: which pages are indexed or excluded and why, duplicates, empty language versions, impressions, clicks, CTR, average position, Core Web Vitals and reports on AI features. A branded queries filter separates searches for the brand from everything else. Someone has to watch it: in one case nobody did, and pages kept slipping out of the index despite heavy publishing.

More: SEO, AEO, and GEO in 2026: What Actually Matters to Google, and What Your Marketer's Report Should Show

Average position is Search Console's average ranking of a site's pages across all the queries they appeared for. It is a diagnostic, not a business KPI, and the most misread number in the report. When a site starts appearing for many new queries below page one, the average falls while clicks and impressions grow: ours slid from about 10 to 17 for exactly that reason. It is read split by page, query and date. Structure work alone once moved a store from 40.8 to about 12 before any SEO campaign.

More: 3D on a Live Site: Our three.js Experiment and What Google Saw

Impressions and clicks are the two base numbers of Search Console: how many times a site's pages were shown in Google results, and how many times someone clicked through. Impressions usually arrive before stable traffic, so periods are compared over weeks or months, not days. Impressions without clicks mean visibility without value: on one platform many articles drew impressions and almost no clicks. The ratio of the two is organic CTR.

More: SEO, AEO, and GEO in 2026: What Actually Matters to Google, and What Your Marketer's Report Should Show

The generative AI performance report is a Search Console report, added in 2026, that shows how often a site's pages appeared in Google's AI Overviews and AI Mode, by page, country, device and date. It does not show queries, clicks or visibility in assistants outside Google. It is best treated as one link in the chain from AI visibility to clicks, leads and business results, not as a goal of its own.

More: SEO, AEO, and GEO in 2026: What Actually Matters to Google, and What Your Marketer's Report Should Show

Lighthouse is Google's open-source tool that audits a page's speed, accessibility and basic SEO in one controlled run; PageSpeed Insights runs Lighthouse in a Google data center, emulating a mid-range phone on a throttled network without a graphics chip. That is why heavy 3D or animation can score far worse there than on real devices: a local run with software graphics took our own home page from 32 to 72-76 after we adapted it. A lab score is a symptom; nothing replaces timing the site on a real phone.

More: 3D on a Live Site: Our three.js Experiment and What Google Saw

Lab data and field data are the two kinds of page speed measurement: lab data is one controlled test run, such as Lighthouse, while field data is 28 days of real Chrome users collected in the Chrome UX Report (CrUX) and shown in Search Console. Each Core Web Vital in the field is judged at the 75th percentile of visits, so slow experiences of a large minority still count. Lab data helps find causes; field data is the diagnosis.

More: 3D on a Live Site: Our three.js Experiment and What Google Saw

A business dashboard, or BI (business intelligence) dashboard, is a live screen of a company's key numbers, such as sales, leads, orders or workload, pulled automatically from the CRM, ERP, website and other systems. It replaces weekly manual reports and the spreadsheets people assemble for them. In marketing, a good dashboard connects channel activity to business outcomes rather than to likes. Dashboards are part of our business digitalization work, which starts from $15,000, often as a custom layer connecting CRM, ERP and portals.

More: Business Digitalization