Iron Goo
A woman and the Iron Goo robot review a printed plan and a tablet of event rows in a lamplit room.

How to Create a Measurement Plan on One Page

Atamyrat Hangeldiyev
Atamyrat Hangeldiyev
Systems Architect
February 23, 2026
On this page
Analytics & Data

To create a measurement plan, write the business objective and the decisions it drives, give each decision one KPI with a formula, source of record and action threshold, list the GA4 events and parameters that feed it, test them in DebugView, set a UTM convention, and keep the plan on one page with a review date and change log. A standard GA4 property allows 40-character event names, 25 parameters per event, 30 key events and 50 event-scoped custom dimensions.

A measurement plan in analytics starts with decisions: who will make them, using which numbers, and what will cause them to act. A tracking plan specifies the events and parameters that collect supporting data. The tracking plan procedure expands the event list into an implementation specification.

The example throughout is fictional. Northgate IT is a hypothetical 35-person managed-services company selling monthly support contracts. Its leads come from a quote form, phone calls and a free network-assessment booking; assume a $4,000 monthly ad budget and an average new contract worth $1,500 per month.

Procedure 1: Write the objective and decisions

Check. Write one business objective, then list choices that named roles will make on a date or review cadence. Each choice should have an action available to its owner. Northgate's objective is to grow signed support contracts:

D1  Monthly: marketing lead decides how to split $4,000 in ad spend.
D2  Weekly: managing director decides whether to open a sales engineer hire.
D3  Monthly: web lead decides whether to rewrite the assessment page.

The owner is a role in this example. On your own plan, replace each role with the person responsible for making the decision. The business-model metrics guide helps you choose decisions and measures that fit the way your company earns money.

Expected. Each decision has an owner, a review date or cadence, and at least two possible actions. Someone reading D3 can tell who may commission a rewrite and when that choice comes up.

Bad result. “Improve marketing” is an objective without a decision. “Review the website” has no defined choice, and a decision nobody owns can stay on a dashboard indefinitely without changing anything.

A one-page chain connecting a business objective to decisions, KPIs and events

Follow the chain from the objective to the decision, KPI and supporting event. Each event should have a reason to exist on the sheet.

Procedure 2: Give each decision a KPI and threshold

Check. For every decision, write one KPI, its formula, its source of record and the threshold that triggers an action. Use the CRM as the count of record for leads and customers; GA4 describes site behavior. A GA4 key event count and the CRM count need not match.

DecisionKPI and formulaSource of recordThreshold and action
D1Cost per qualified lead by channel = channel spend / qualified leads, rolling 12 weeksSpend records and CRM qualified leadsAbove $400 for two consecutive monthly reviews: marketing lead halves that channel's budget and moves the money to the cheapest channel
D2Qualified leads per weekCRMAbove 12 for eight consecutive weeks: managing director opens the hire
D3Assessment bookings per 100 landing sessions = assessment_booked key events / sessions landing on /assessment × 100GA4Under 2 per 100 with at least 400 landing sessions in the period: web lead rewrites the page

Define the comparison period before checking D3, and use the same period for bookings and landing sessions. The event count must correspond to those landing sessions for the ratio to answer the page decision. At 400 landing sessions, fewer than eight bookings meets the stated rate threshold.

For other decisions, keep the definitions consistent: CPL = spend / leads; cost per qualified lead = spend / qualified leads; close rate = new customers / leads, grouped by the month the lead arrived. The CRM supplies the lead and customer counts. An ad platform's reported conversion can help investigate a channel, but it is not the count of record for those business outcomes.

Northgate's D2 baseline is nine qualified leads per week. Our rough count-noise rule uses the square root: √9 = 3, so a weekly count from six to 12 is ordinary variation around that baseline. The eight-week threshold requires a sustained count above that band before opening the hire.

Expected. Each KPI has a named owner, formula, source, threshold and action. The KPI card standard adds the baseline and noise band, counter-metric and other fields for a full card; the small-business analytics vocabulary distinguishes events, key events, sessions and the conversion of record.

Bad result. A D2 threshold of 10 qualified leads in one week sits inside the ordinary noise band and could trigger a hire on a routine swing. A KPI sourced solely from an ad platform's own conversion claim cannot establish the CRM lead or customer count.

Procedure 3: Specify events, parameters and names

Check. List the event that records each observable site action, its firing condition and the parameters needed to interpret it. Fire the quote event on a confirmed success state, after validation and successful submission. A button click alone does not establish that a lead was generated.

Event nameTypeFires whenParametersKey event?
generate_leadRecommendedQuote form reaches its success statelead_source, value, currencyYes
phone_clickCustomA visitor clicks a tel: linklink_locationNo; calls are counted in the CRM
assessment_bookedCustomBooking confirmation appearsbooking_typeYes
qualify_leadRecommendedCRM marks a lead qualified, if sent to GA4value, currencyYes, if implemented
close_convert_leadRecommendedA qualified lead becomes a customer, if sent to GA4value, currencyYes, if implemented
purchaseRecommendedA purchase completes, if applicabletransaction_id, value, currency, items; optional tax, shipping, coupon, customer_typeOnly if relevant
sign_upRecommendedA real account is created, if applicablemethodNo

The last two rows are reference choices, not assumed Northgate actions. For purchase, use a unique transaction_id and an items array whose items each have item_id or item_name. Its value is the sum of item price times quantity, excluding shipping and tax; customer_type, if supplied, is new or returning.

For generate_lead, lead_source is optional. If you send value, send a three-letter ISO 4217 currency too; Google recommends a value when the event is a key event. The following adapts Google's recommended event reference to Northgate's assumed monthly contract value. It is an assigned event value, not proof that a lead has become revenue.

// Run only after the quote form confirms a successful submission.
gtag("event", "generate_lead", {
  currency: "USD",
  value: 1500,
  lead_source: "quote_form"
});

Apply Google's event naming rules: start with a letter and use only letters, numbers and underscores. Names are case sensitive. Lowercase snake case is our convention, matching Google's recommended names; Google does not require lowercase.

GOOD  assessment_booked  17 characters; starts with a letter
GOOD  phone_click        letters and underscore

BAD   Assessment Booked  space is invalid; capitals break our convention
BAD   form_submit        reserved web event name
BAD   Generate_Lead      separate event from generate_lead
BAD   2nd_call           starts with a number
BAD   aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
                         41 characters; exceeds the 40-character limit
Do not send a custom form_submit event from your tag

form_submit, form_start, click and scroll are reserved web event names collected by GA4's enhanced measurement. GA4 can use these names when you create or modify events inside GA4. form_submit can already appear in reports, but do not mark it as the lead key event: use generate_lead for a confirmed lead and reserve key-event status for actions that represent the result you intend to count.

Google's collection limits and configuration limits, read on 29 October 2026, set these boundaries for a standard property:

LimitStandard property
Event name length40 characters
Parameters per event25
Parameter name length40 characters
Parameter value length100 characters, with documented page-field exceptions
Event-scoped custom dimensions50
Key events30

GA4 doesn't log an event whose name is over 40 characters, so it cannot count as a key event either. Web data streams have no limit on distinctly named events, which is not a reason to create a new name for every button.

Register lead_source, link_location and booking_type as event-scoped custom dimensions before launch. They use three of the 50 slots. Do not register a timestamp or lead ID as a custom dimension: high-cardinality values can push data into the (other) row. Google says a new custom dimension can take 24 to 48 hours to become available for reporting, and freeing a slot at the limit takes 48 hours after deletion.

The success-state trigger and hidden UTM fields depend on the site implementation. Iron Goo's website foundation build includes instrumenting analytics.

Expected. Every implemented event has a defined trigger, valid name and a parameter list that answers a decision or helps interpret its KPI. A successful quote submission produces one generate_lead; a validation failure produces none.

Bad result. A tag on both the submit click and confirmation produces duplicate leads in GA4. A parameter sent but never registered as a custom dimension may not be available in the reporting detail the plan expects.

Procedure 4: Mark key events and test the full path

Check. In GA4, open Admin > Data display > Events and select the star for each collected event that should be a key event. For an event not yet collected, use Create event and turn on Mark as key event. On Northgate's standard property, mark generate_lead and assessment_booked; mark qualify_lead and close_convert_lead if the optional CRM events are implemented.

Marking an event does not change historic data. Standard reports can take up to 24 hours to show it, while Realtime updates a few minutes after a key event occurs. A standard property has 30 key-event slots; a 360 property has 50. See Google's key-event instructions for the setup.

Enable debug mode with Tag Assistant or preview mode, or set debug_mode: true on the Google tag configuration or individual event. In Tag Manager, set it on the Google tag or event tag. Setting debug_mode to false does not disable debug mode; remove the parameter.

Open Admin > Data display > DebugView, select the test device, and submit the form once. The seconds stream shows the last 60 seconds; the minutes stream and Top Events cover 30 minutes. Test a failed validation and a booking confirmation as separate actions.

GOOD: hypothetical DebugView mock
12:00:01  generate_lead     lead_source=quote_form value=1500 currency=USD (flagged as key event)
12:00:20  assessment_booked booking_type=network_assessment (flagged as key event)

BAD: hypothetical DebugView mock
12:00:01  generate_lead     first event for one successful submit
12:00:01  generate_lead     duplicate event for the same submit
12:00:02  Generate_Lead     separate name caused by capitalization
12:00:03  generate_lead     value=1500, currency missing
12:00:10  generate_lead     fired despite failed validation
Clean DebugView stream beside a stream with duplicate lead events for one form submission

Compare the single lead event in the clean stream with the two lead events produced by one submission in the faulty stream.

DebugView has a blind spot: client-side privacy controls, or consent mode without consent to Analytics cookies, can prevent events from appearing. Check the test browser's consent and privacy settings before treating an empty stream as proof of a broken tag. Google's DebugView documentation also recommends filtering developer traffic while debugging.

Expected. DebugView shows one correctly named event with the expected parameters for each successful test, and none for failed validation. Then check the CRM: each successful test should have one corresponding record. Realtime provides a further check in its event and key-event cards, though Acquisition reports are the place to examine attribution.

Bad result. Two generate_lead entries for one submit indicate duplicate firing. A clean DebugView event with no CRM record means the tag fired but the lead did not reach the count of record.

Procedure 5: Set and test the UTM convention

Check. Keep one shared link sheet with an owner and destination for each campaign URL. Use lowercase values and hyphens within values. Set utm_medium from a fixed list: email, paid_social, social, referral, and cpc only outside Google Ads. Keep the underscore in paid_social; it matches Google's paid-medium pattern.

Require utm_source, utm_medium and utm_campaign. Strongly recommend utm_id and utm_source_platform; use utm_term and utm_content when needed. Google Ads uses auto-tagging with no manual UTMs under this convention.

https://example.com/assessment?utm_source=newsletter&utm_medium=email&utm_campaign=monthly-tips

The address is a placeholder for a test. UTM values are case sensitive, so Facebook and facebook create separate source rows; channel definitions are not case sensitive. The marketing measurement guide has the full UTM before-and-after table and CRM scorecard.

Open a tagged test link, accept the relevant analytics consent and check the expected source and medium in Realtime. Submit the form and inspect the hidden UTM fields on the corresponding CRM record. Use Acquisition reports for attribution analysis because Realtime is a best-effort view.

Expected. The shared sheet, test visit and CRM record agree on the campaign values. The example uses newsletter as utm_source and email as utm_medium, giving the channel rules a recognizable medium.

Bad result. A link using utm_source=mailchimp and utm_medium=newsletter lands in Unassigned, on our reading of Google's default channel group rules. If Realtime sees the visit but the CRM fields are empty, inspect the site's field capture and form submission path.

Procedure 6: Assemble the one-page YAML plan

Check. Put the objective, decisions, event choices, UTM rules, review cadence and change log in one file. Print it and check that it fits on one page at a readable size. Trace each implemented event back to a decision; use a fuller tracking plan for implementation details that do not fit.

This is a hypothetical measurement plan template for Northgate IT. D1 and D2 use CRM counts, while the site events help inspect the lead path; D3 uses the booking event in its KPI. The optional CRM events provide GA4 context for D1 and D2 if implemented.

objective: Grow signed monthly support contracts.
decisions:
  - id: D1
    decision: Split the $4000 monthly ad budget.
    owner: Marketing lead
    kpi: Cost per qualified lead by channel, rolling 12 weeks
    formula: Channel spend / qualified leads
    source_of_record: Spend records and CRM
    baseline: Establish from the rolling 12-week period
    action_threshold: Above $400 at two consecutive monthly reviews
    action: Halve that channel and move funds to the cheapest channel.
  - id: D2
    decision: Open a second sales engineer hire.
    owner: Managing director
    kpi: Qualified leads per week
    formula: Count of qualified leads per week
    source_of_record: CRM
    baseline: 9 per week; ordinary range approximately 6 to 12
    action_threshold: Above 12 for eight consecutive weeks
    action: Open the hire.
  - id: D3
    decision: Rewrite the assessment page.
    owner: Web lead
    kpi: Assessment bookings per 100 landing sessions
    formula: assessment_booked / sessions landing on /assessment * 100
    source_of_record: GA4
    baseline: Establish for the chosen review period
    action_threshold: Under 2 per 100 with at least 400 landing sessions
    action: Rewrite the assessment page.
events:
  generate_lead: Quote success; lead_source, value, currency; D1 and D2
  phone_click: Telephone link click; link_location; D1 and D2
  assessment_booked: Booking confirmation; booking_type; D3
  qualify_lead: Optional CRM qualification; value, currency; D1 and D2
  close_convert_lead: Optional CRM customer conversion; value, currency; D1
key_events: [generate_lead, assessment_booked, qualify_lead, close_convert_lead]
custom_dimensions: [lead_source, link_location, booking_type]
utm:
  required: [utm_source, utm_medium, utm_campaign]
  recommended: [utm_id, utm_source_platform]
  optional: [utm_term, utm_content]
  convention: Lowercase; hyphens within values; shared link sheet.
  mediums: [email, paid_social, social, referral, cpc]
  google_ads: Auto-tagging; no manual UTMs.
review:
  kpis: Weekly, 20 minutes; D1 and D3 decisions monthly
  plan: Quarterly
changelog:
  - date: "2026-10-29"
    change: Hypothetical plan established
    owner: Marketing lead
    expected_effect: Establishes definitions; no result assumed

The two optional CRM events belong in key_events only if Northgate implements them. Record an established baseline for D1 and D3 once the chosen periods contain data; do not invent one to fill the field.

Expected. The printed page names an owner and action for every decision, shows the count of record and identifies why each event is collected. The YAML remains readable by both the person making the decision and the person implementing the tags.

Bad result. An event with no decision link is a candidate for removal. A plan that omits the source or threshold leaves reviewers arguing over which number to use after they have seen it.

Procedure 7: Review the numbers and record changes

Check. Schedule a 20-minute KPI review every week and a plan review every quarter. At the quarterly review, identify events nobody uses, retire those with no remaining decision or diagnostic purpose, and recheck thresholds against current baselines and noise. Keep the monthly D1 and D3 decisions on their stated cadence.

Add a change-log entry whenever a tag, form, consent setting or definition changes. Record the date, change, owner and expected effect on the numbers before looking for the effect in a chart.

- date: "2026-11-05"
  change: Hypothetical quote form success trigger updated
  owner: Web lead
  expected_effect: Fewer duplicate generate_lead events per submission

Treat this as an example entry, not a claim that a change occurred. If the number later steps down, the log gives the reviewer a specific change to investigate alongside other possible causes.

Standard-property daily data is typically available in about 12 hours, but processing can take 24 to 48 hours and report data may change during that window. Do not judge yesterday's results at a morning review; use a settled period and consult Google's data freshness guidance when a number is still moving.

Expected. The calendar contains both review cadences, thresholds are checked against current data, and every relevant change has a dated owner and expected effect. A reviewer can connect a chart change to a documented implementation change and investigate it.

Bad result. An unexplained change in generate_lead after a form release leaves the team unable to tell from the plan whether demand changed or the counting method changed. Missing change-log entries make that investigation harder.

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