Iron Goo
Iron Goo robot traces a reference card beside a laptop in a dark office lit by a desk lamp

Analytics for Small Business: Terms Explained

Atamyrat Hangeldiyev
Atamyrat Hangeldiyev
Systems Architect
February 19, 2026
On this page
Analytics & Data

Analytics for small business is the practice of turning recorded events from your website analytics, CRM and payment processor into decisions. Its first complication is vocabulary: the systems can all report a “conversion” or a “customer” while counting different things. For example, GA4 bounce rate is the percentage of sessions that were not engaged, not the percentage of people who viewed only one page.

Analytics terms quick-reference card

TermWhat it meansWhere you see itThe trap
EventA recorded action with a name and parametersGA4 Events reportA recorded form action does not establish that a lead reached the CRM.
SessionA period of interaction with a site or appGA4 Traffic acquisition reportA direct return can inherit an earlier non-direct source.
UserAn identity GA4 counts over a date rangeGA4 user metrics and reporting identity settingsA phone and laptop may count as separate users.
Engaged session and bounce rateA session meeting an engagement rule; bounce rate counts those that do notGA4 engagement metricsA bounce is no longer defined simply as a single-page visit.
Key eventAn event marked as important to business successGA4 Admin > Data display > EventsMarking it does not change historic data.
ChannelA grouping of traffic sourcesGA4 Traffic acquisition reportSession default channel group uses last click, even when the property uses another attribution model.
AttributionA rule for assigning credit to interactionsGA4 attribution settings and advertising reportsAdding platform claims together can count the same lead more than once.
FunnelA sequence of steps taken toward an actionGA4 Funnel explorationOpen funnels allow entry after the first step.
Conversion of recordThe business action you decide to count as the official resultCRM lead and customer recordsA GA4 key event and an Ads conversion need not match the CRM.
CACChannel spend divided by new customersMarketing spend records and CRMRevenue per customer is the wrong comparison if you want to assess margin.
MRRMonthly-normalized value of qualifying subscriptionsStripe Billing analyticsAnnual subscriptions need normalization; usage-based products are excluded.
Churn and retentionCustomers lost, or a starting group still paying, over a periodStripe Billing analyticsSubscriber retention and revenue retention answer different questions.
LTV and ARPUEstimated lifetime value and average revenue per userStripe Billing analyticsStripe's subscriber LTV is a revenue estimate, before gross margin.
CohortUsers grouped by a shared characteristicGA4 Cohort exploration or Stripe Billing analyticsA cohort's inclusion date and return condition change the answer.
First-party dataData collected directly from your customers and visitorsCRM, order records, forms and your own analyticsOwning a form does not guarantee its fields reach the CRM.
Consent modeTag behavior that responds to consent choicesGoogle tag and GA4 consent settingsModeling has traffic thresholds; declined visitors may be missing.
Data retentionHow long eligible user-level and event data remains availableGA4 Admin > Data retentionStandard aggregated reports can remain while older explorations are unavailable.
Sampling and thresholdsAn estimate from a subset, or data withheld by GA4GA4 explorations and report data-quality indicatorsA missing row can mean withheld data, not zero activity.
Data warehouseAn analysis database holding copies of multiple systems' dataBigQuery and connected exportsExporting data does not automatically reconcile customers or revenue.
Dashboard and reportA recurring monitoring view and a one-question analysis, respectivelyCRM, GA4 or a reporting toolA dashboard without an owner or decision rule becomes decoration.
Leading and lagging metricsAn earlier potential predictor and an outcome measured afterwardCRM pipeline and financial reportsA leading number needs evidence that it predicts the outcome.
Sample sizeThe number of observations behind a rate or countGA4 landing-page reports and CRM lead countsThree leads from 100 visits do not establish a precise conversion rate.
Owner checks an analytics reference card beside a report on a laptop

Check the definition and the reporting system before treating a number as a business result. The same label can hide different counting rules.

Choose the source of truth before comparing numbers

A source of truth is the system you designate to settle a particular business question. Use the system closest to the money for the result you are counting, while recognizing that each system has a different job.

QuestionSource of truthWhat to check
How much money came in?Payment processor or accountingCheck payments, refunds and the reporting period against the financial record.
How many leads and customers did we get?CRMDefine which records qualify as leads and which became customers.
What did visitors do on the site?GA4Check events, sessions and the applicable reporting identity.

The payment system is closest to money, but it cannot tell you how many qualified leads never bought. The CRM is the count of record for leads and customers; GA4 describes observed site behavior. When two counts disagree, first ask whether they measure the same action and period, then use the system closest to the business outcome being decided.

Website analytics, CRM and payment records feed a decision, with payment records closest to money

Each system owns a different kind of number. Decisions draw on all three, while payment or accounting records carry the most weight for money.

Write the chosen events, key events and source of truth for each outcome into a measurement plan. Google and Stripe documentation was read on 29 October 2026.

Terms you meet in website analytics

Event

An event is one recorded action with a name and parameters. GA4 uses event names such as page_view, session_start, first_visit and generate_lead; you can inspect their counts in the Events report.

The trap is treating an event name as proof of the business outcome. A generate_lead event can record a form action, but you still need to check that the corresponding record arrived in the CRM and met your lead definition.

Session

A session is a period during which a user interacts with your website or app. GA4 starts one when a user opens an app in the foreground or views a page or screen, and the default timeout is 30 minutes of inactivity; you can adjust that timeout.

The session_start event carries attribution information, including a Google Ads click identifier when present, UTM parameters and the referrer. Google gives an example in which someone arrives from Google organic on Day 1 and returns directly on Day 68: the later session is also attributed to google / organic under the 90-day default key event lookback window. A session's reported source is therefore an attribution result, not necessarily the literal source of that return visit. See Google's session definition and example.

30 minutes
Default GA4 session inactivity timeout
30
Maximum key events on a standard GA4 property
10 steps
Maximum steps in a GA4 Funnel exploration
10 million
Standard-property event-level query quota

Total users, Active users and New users

Google defines Total users as the number of unique users who triggered any event in the specified date range. Active users are the unique users who engaged with your site or app in that range. New users are unique users who logged the first_open or first_visit event. These are different metrics, so label them precisely when moving a number into a spreadsheet. See Google's user metric definitions.

Reporting identity changes how GA4 identifies users. Blended uses User-ID, then device ID, then modeling; Observed uses User-ID, then device ID; Device based uses only device ID. Our reading is that, without a sent User-ID, one buyer using a phone and a laptop can appear as two users. Device based reporting uses device ID even when you send a User-ID.

A user count is useful for trends, but it is not a customer count. Check the reporting identity before comparing a GA4 user figure with people in the CRM, and keep the same identity setting in mind when comparing periods.

Engaged session, engagement rate and bounce rate

An engaged session meets at least one of three conditions: it lasts longer than 10 seconds, has a key event, or has two or more page or screen views. Engagement rate is the percentage of sessions that were engaged; bounce rate is the percentage that were not engaged. The two rates are opposites.

This differs from the old single-page interpretation of bounce rate. A hypothetical visitor who reads one page for nine seconds, triggers no key event and views no other page has a bounced session under the rule, even if they found the answer they wanted. Use the rate as a prompt to inspect behavior, not a verdict on whether every visitor was satisfied. Google's engagement definitions give the exact conditions.

Key event

Events that measure actions important to the success of your business are now called key events. In a standard GA4 property, you can mark up to 30. Go to Admin > Data display > Events and use the star icon for an existing event; marking it does not change historic data.

A GA4 key event supports analysis in Analytics. A Google Ads conversion is used for Ads measurement and bidding, while the CRM holds your chosen lead or customer count of record. Even when all three concern a form submission, their totals can differ because their recording and attribution rules differ.

Choose a meaningful event such as generate_lead, then test the full path from the site to the CRM. A site must capture events and fields such as hidden UTM values correctly for either system to receive useful information; Iron Goo's website foundation service is relevant when that capture needs work.

Channel

A channel groups traffic sources into categories. In GA4, the Default channel group dimension uses the property's attribution model. Session default channel group and First user default channel group use paid and organic last click; our reading is that Traffic acquisition, when using Session default channel group, is a session-level last-click view.

Google defines AI Assistant as traffic arriving from sources such as ChatGPT, Gemini, Deepseek, Copilot or Grok. AI Overviews and AI Mode are counted in Organic Search. These labels describe the source classification GA4 can make; they do not establish every interaction a buyer had before arriving.

For tagged campaigns, supply utm_source, utm_medium and utm_campaign. Google says its channel definitions are not case sensitive, while UTM values are. Keep a consistent naming convention so differently capitalized campaign values do not fragment your analysis. See Google's default channel group definitions.

Attribution

Attribution is the rule used to assign credit for an outcome to marketing interactions. GA4 offers Data-driven attribution, Paid and organic last click, and Google paid channels last click. Data-driven attribution is the default; it distributes credit based on data for each key event and is specific to the property. Google's attribution documentation does not state a minimum data requirement for the model.

First click, linear, time decay and position-based models were removed as of November 2023. Direct visits receive no credit unless the entire path is direct. The default lookback is 30 days for first_open and first_visit key events and 90 days for other key events, so a reported source can depend on interactions well before the action.

Do not add platform conversion claims

Google Ads and other ad platforms can each claim credit for the same lead. Use the CRM count as the lead total, then examine platform reports as separate views of contribution.

A platform's credit is useful for decisions inside that platform, but a sum of overlapping claims is not a company-wide lead count. The marketing measurement guide works through channel measurement with a CRM scorecard.

Funnel

A funnel is a sequence of steps you define to examine progression toward an action. In GA4 Funnel exploration, you can define up to 10 steps, such as viewing a service page, starting a form and submitting it.

An open funnel lets users enter at any step. A closed funnel requires entry at the first step. If someone arrives directly at the form, an open funnel can include that person at the form step; a closed funnel requires that person to have entered at step one. State the entry rule before comparing funnel counts or drop-off rates.

Terms you meet in the CRM and payment processor

Conversion of record

A conversion of record is the business action you choose as the official result for a decision. For a lead-generation company, that might be a CRM record that meets an agreed lead definition; for a customer count, it might be a CRM record marked as won.

The useful detail is the rule, not the label. Write down whether duplicates, existing customers and incomplete inquiries count, then apply that definition consistently. GA4 can record a key event and an ad platform can claim a conversion without either establishing that a qualified CRM lead exists.

Customer acquisition cost and cost per lead

Customer acquisition cost, or CAC, is channel spend divided by new customers from that channel. Cost per lead, or CPL, is channel spend divided by leads. For a hypothetical channel with $2,000 in spend, 40 recorded leads and five new customers, CPL is $50 and CAC is $400.

Use the CRM's agreed lead and customer definitions for the denominators, and state the spend period and attribution rule. CAC payback in months is CAC divided by monthly gross margin per customer. If that hypothetical $400 CAC produces $100 of monthly gross margin per customer, payback is four months under those assumptions.

Monthly recurring revenue

Stripe defines MRR as the monthly-normalized value of subscriptions in active and past_due status. It excludes taxes, free plans and usage-based products. An annual subscription therefore contributes its annual value divided by 12 to this metric.

Stripe's example has 100 subscribers at $100 per month and 50 at $600 per year. The calculation is (100 × $100) + (50 × $600 / 12) = $12,500 in MRR. It is a subscription metric, so do not use it as a substitute for a complete accounting revenue figure.

Churn and retention

Subscriber churn rate measures the share of subscribers lost over a period. Stripe calculates its rate as subscribers churned in the last 30 days divided by active subscribers 30 days ago plus new subscribers in the last 30 days. Its example is 100 / (1,000 + 100) = 9.1%.

Retention measures the share of a starting group still active or paying after a stated number of periods. Stripe illustrates 10% monthly churn with retention of 90%, 81%, 73%, 66% and 59% in months one through five. The same percentage lost each month compounds against a shrinking group.

Subscriber retention and revenue retention are different measures. Revenue retention can exceed 100% when upgrades expand revenue from the retained group. Also check the time zone when reconciling Stripe: pre-built reports use UTC dates, while Billing Analytics charts use the account time zone, so monthly totals may differ.

ARPU and lifetime value

Stripe's ARPU is total MRR divided by active subscribers. Its example is $50,000 / 100 = $500. Stripe estimates subscriber LTV by dividing ARPU by subscriber churn rate, using point-in-time values.

In Stripe's example, $500 divided by 9% produces approximately $5,555. Our reading is that this is a revenue estimate before gross margin, not profit. For a hypothetical comparison with CAC, assume a 40% gross margin and apply it to the estimate:

Stripe revenue LTV and a hypothetical margin LTV

revenue_LTV = $500 ARPU / 0.09 monthly subscriber churn = about $5,555 margin_LTV = $5,555 revenue LTV × 0.40 assumed gross margin = about $2,222

Compare CAC with margin LTV using compatible customer definitions and periods. The calculation is an estimate, and its usefulness depends on whether the current ARPU, churn and assumed margin are reasonable representations of the customers you are evaluating. Stripe's Billing analytics definitions specify the inputs.

Cohort

A cohort is a group of users sharing a characteristic identified by an Analytics dimension. Cohort analysis follows that group across time, so you can ask how many people acquired in a particular period return later.

In GA4 Cohort exploration, granularity can be Daily, Weekly or Monthly; a week runs Sunday through Saturday. An inclusion criterion such as First touch groups users by acquisition date, and a return criterion defines what counts as coming back. A cohort of first-time visitors who trigger any event later answers a different question from a cohort of paying subscribers retained in Stripe.

Always record the inclusion rule, return rule and period beside the result. Without them, two retention percentages can look comparable while describing different groups.

Terms for data collection and interpretation

First-party data

Our working definition of first-party data is information you collect directly from your customers and visitors. CRM records, orders, form submissions and your own analytics are examples.

Direct collection does not guarantee completeness or accuracy. A form may collect a campaign value that never reaches the CRM, and an order record may not identify the marketing interaction that brought the buyer. Check the fields and handoffs needed for the decision you want to make.

Consent mode changes how Google tags behave according to consent choices. With analytics_storage denied, tags do not read or write first-party analytics cookies. With ad_storage denied, no new advertising cookies are written.

In basic consent mode, tags are blocked until consent; declined visitors produce no data through those blocked tags. In advanced consent mode, tags load before the banner and can send cookieless pings when consent is denied, enabling modeling when its requirements are met.

Behavioral modeling requires at least 1,000 events per day with analytics_storage denied for at least seven days and at least 1,000 daily users sending events with analytics_storage granted for at least seven of the previous 28 days. Without modeling, Analytics misses data for users who decline. Our inference is that many small sites are unlikely to meet both thresholds. Treat modeled recovery as uncertain and compare observed website activity with CRM outcomes. See Google's behavioral modeling requirements.

Data retention

Data retention controls how long eligible user-level and event data remains available for certain GA4 analysis. Standard properties offer two or 14 months; the longer event-data options of 26, 38 and 50 months are for Analytics 360. User-level data retention is limited to two or 14 months.

Check Data retention in Admin and select 14 months if you need the longer standard-property window. The setting does not affect standard aggregated reports; it affects explorations and funnel reports. A standard report can therefore show an older period even when the underlying period is unavailable for an exploration.

Sampling and data thresholds

Sampling means a query uses a subset of data to estimate a result. For standard GA4 properties, the event-level query quota is 10 million events; sampling may occur when the events used exceed that quota. Check the data-quality information in an exploration before treating its figures as exact counts.

Data thresholds are a separate issue. GA4 can withhold data, including when a report contains demographic information or has low user or event counts in a narrow date range. Thresholds are system-defined and cannot be adjusted, so a missing row is not necessarily evidence of zero activity.

Data warehouse

Our working definition of a data warehouse is a database built for analysis that holds copies of data from several systems so they can be joined. You might combine GA4 events, CRM outcomes and Stripe records to examine the path from site visit to customer.

For a company with 10 to 200 people, our opinion is that a warehouse is usually not the first purchase. A CRM report or spreadsheet can answer many early questions with less setup, especially when the definitions and ownership of the numbers are still unsettled.

GA4 can export to BigQuery. Standard properties have a one-million-event-per-day limit for Daily batch export, while Streaming export has no event limit. Export to the BigQuery sandbox is free within sandbox limits. Outside the sandbox, BigQuery charges for storage and query processing. An export gives you data to work with, not an automatic match between a browser and a paying customer.

Dashboard and report

These are our working definitions: a dashboard is a standing view of a few numbers reviewed on a cadence by an owner. A report answers one question once and ends in a decision.

A dashboard needs an owner and a response rule. For example, a hypothetical weekly CRM dashboard could show qualified leads and identify who investigates a sustained decline; a report could investigate whether a particular campaign produced qualified leads. The guide to choosing metrics helps narrow the recurring view to numbers someone can act on.

Leading and lagging metrics

A lagging metric, in our working definition, records an outcome after the fact: revenue, churn and CAC are examples. A leading metric is an earlier number that should predict an outcome, such as qualified leads or quotes sent before sales close.

The word “should” carries the work. Check prior periods to see whether changes in the proposed leading metric preceded changes in the outcome. A rising count of quotes is a poor predictor if the quotes rarely become customers.

Email opens deserve particular caution as a leading metric. Apple says Mail Privacy Protection downloads remote content privately in the background; our reading is that this can register an open nobody actually read. Apple's protection does not extend to links, so report clicks and replies when assessing engagement. See Apple's Mail Privacy Protection description.

Sample size and conversion-rate uncertainty

Sample size is the number of observations behind a result. A conversion rate based on a small number of visits can move sharply when only one or two more people convert, so the displayed percentage should not be treated as a precise estimate.

Suppose a page gets 100 visits and three leads, for a measured rate of 3%. The calculation uses p = 0.03, n = 100, and 1.96 as the multiplier for a rough normal-approximation, or Wald, 95% interval:

Hypothetical conversion rate and rough 95% interval

rate = 3 leads / 100 visits = 0.03 = 3% standard_error = sqrt(0.03 × 0.97 / 100) = sqrt(0.000291) = about 0.0171 rough_95%_interval = 3% ± (1.96 × 1.71 percentage points) = 3% ± 3.34 points = about 0% to 6.3%, floored at 0%

This is a rough approximation with few conversions, not a precise boundary; Wilson and exact Clopper-Pearson methods put the upper end near 8.5%, making the uncertainty even clearer. A change from 2% to 4% on 100 visits in each period is insufficient evidence on its own that the underlying rate improved. Collect more observations and check that the lead definition and traffic mix stayed comparable.

Counts have noise too. As a rough rule, the typical variation is approximately the square root of the count: for 16 leads in a month, that is about four leads. A month with 12 or 20 leads can therefore be ordinary variation rather than proof that a campaign has changed.

Free ebook

Put your next steps in order

Build on what you have just read with a free ebook on semantic SEO and AI visibility. Get worked examples, decision rules, checklists and a do-this-week action in every chapter, plus a six-week plan for your site.

By signing up and confirming your address, you consent to receive the free ebook plus one email every two weeks with 3 or 4 new Iron Goo videos, guides and articles, and nothing else; unsubscribe any time with one click.

Related in Analytics & Data

Ready to move?

Tell us where your business is today. A written assessment lands in your inbox within two business days.

Talk to us