A reference you can use while doing the work
We went through the pain of this so you don’t have to. Here is the plain-English version, with a practical example for every term.
A team can send more messages, ship more features and still leave customers stuck. Useful measurement connects the work to the change people came for. These terms help you define that change, follow cohorts fairly, and distinguish observed improvement from evidence that your intervention caused it.
About the numbers: these are hypothetical teaching examples, not business results or recommended targets. Dollar amounts use USD for illustration.
1. Define progress before choosing a dashboard
Outcome
A meaningful change in a customer’s situation or behavior. Define it in observable terms without pretending the proxy is the whole experience.
Example: A new customer completes their first usable project plan; receiving a welcome email is only an activity along the way.
Output
Something the team produces or does. Outputs can support outcomes but do not guarantee them.
Example: Publishing three tutorials is an output. Whether eligible customers can complete setup afterward is a separate question.
Leading and lagging indicators
A leading indicator may signal future progress; a lagging indicator records a later result. Predictive relationships need validation.
Example: First-week use may precede renewal. Verify whether it predicts renewal in your own cohorts rather than assuming it does.
North Star metric
A central measure intended to reflect recurring customer value, supported by economic and quality checks.
Example: Weekly customer plans used to run real work is more informative than templates generated, but does not replace retention or margin.
Activation
A defined early behavior indicating that a customer reached a useful milestone.
Example: A trial account activates when it creates and shares a usable project plan within seven days.
Time to value
Time from a defined starting event to the first useful result. State the population included in the calculation.
Example: Among users who activate, the median time is two days. Also report how many never activate, so a fast minority does not hide failure.
Conversion rate
The share of an eligible group completing a defined action within a stated window.
Example: Forty of 100 eligible signups complete the first plan in seven days: 40% activation conversion.
Guardrail
A measure that detects an unacceptable downside while the primary outcome improves.
Example: A faster onboarding flow is not a win if support escalations or incorrect customer data rise beyond the agreed limit.
2. Follow customers over time
Cohort
Customers grouped by a common starting event or trait and followed over comparable time windows.
Example: Compare September and October signups at day 30, rather than comparing one mature group with one that joined yesterday.
Retention
The share of a starting cohort still active or returning under a specified rule and interval.
Example: Sixty of 100 starting accounts perform the key task during week four: 60% week-four activity retention.
Customer churn
Starting customers lost during the period divided by starting customers. State how pauses and reactivations are treated.
Example: Ten of 100 starting subscribers leave: 10% customer churn. New subscribers do not enter that denominator.
GRR — gross revenue retention
Starting recurring revenue less churn and contraction, divided by starting recurring revenue. Excludes expansion and new customers.
Example: ($10,000 − $1,000 churn − $500 contraction) ÷ $10,000 = 85% GRR.
NRR — net revenue retention
Starting recurring revenue plus expansion less contraction and churn, divided by starting recurring revenue. Excludes new customers.
Example: ($10,000 + $2,000 expansion − $1,000 churn − $500 contraction) ÷ $10,000 = 105% NRR.
Expansion and contraction
Changes in recurring revenue from existing customers: upgrades increase it; downgrades reduce it.
Example: An existing customer moves from $100 to $150 monthly: $50 expansion. A move to $80 would be $20 contraction.
Reactivation
A previously inactive or churned customer returning under an explicit rule. Distinguish return use from a new acquisition.
Example: An account inactive for 60 days resumes the core task. Count it as reactivated under that definition, not as a first-time customer.
Lifecycle stage
A useful customer state that informs the next action or support. State transitions should reflect behavior.
Example: A signed-up but inactive customer needs help reaching first value; an active customer may need guidance on a recurring workflow.
3. Connect experience to sustainable economics
MRR and ARR
MRR normalizes eligible monthly recurring revenue; ARR annualizes it. They are not cash collections or guaranteed future revenue.
Example: $8,000 MRR corresponds to $96,000 annualized ARR, excluding one-time setup fees.
CAC — customer acquisition cost
Included sales and marketing acquisition costs divided by new customers. Match the period and cost scope to the customer journey.
Example: $6,000 acquisition cost ÷ 100 new customers = $60 CAC.
Contribution margin
Revenue less the costs that vary with serving the customer, before or after acquisition as explicitly labeled.
Example: $100 revenue minus $60 variable service and transaction costs leaves $40 contribution before acquisition and overhead.
LTV — customer lifetime value
Expected revenue or contribution over a customer relationship. Report the horizon and assumptions; early cohort value is not a proven lifetime.
Example: One hundred customers produce $12,000 revenue and $4,800 contribution over 90 days: observed 90-day value is $120 revenue or $48 contribution each.
LTV:CAC
Customer value divided by acquisition cost. Use contribution value to assess how acquisition may be recovered.
Example: $240 modeled contribution LTV ÷ $60 CAC = 4×. A 90-day observed contribution of $48 would be only 0.8× over that shorter window.
CAC payback
The point when cumulative customer contribution covers acquisition cost. A projected payback is different from an observed one.
Example: At $32 contribution per order and $60 CAC, one order leaves $28 unrecovered; two contribute $64 and cover CAC.
CSAT — customer satisfaction
A survey measure with a defined question, scale, threshold and respondent denominator. It describes respondents, not automatically all customers.
Example: Forty of 50 respondents select satisfied or very satisfied: 80% top-two-box CSAT, with the response rate reported separately.
NPS — Net Promoter Score
The share of 9–10 recommendation ratings minus the share of 0–6 ratings on a 0–10 scale. Ratings of 7–8 are passives.
Example: Among 100 respondents, 60 promoters and 20 detractors yield NPS +40. It is a score, not 40% retention.
4. Interpret the evidence before acting
CES — customer effort score
A survey measure of how easy or difficult a task felt. State the question, scale and which direction is favorable.
Example: On a 1–7 scale where 7 means very easy, respondents average 5.8. Do not compare it with a reversed scale without conversion.
Attribution
Rules assigning credit to touchpoints; it does not establish what would have happened without them.
Example: A reminder receives credit for a renewal, but the customer may have renewed without receiving it.
Incrementality
The additional outcome caused by an intervention compared with a credible counterfactual.
Example: A randomized holdout may estimate how many extra customers activate because of a new onboarding message.
A/B test
An experiment comparing alternatives using a defined allocation, outcome and analysis plan. Randomization helps isolate the effect.
Example: Assign eligible new accounts to one of two onboarding flows, keep assignments stable, and compare seven-day activation with guardrails.
Absolute change and relative lift
Absolute rate change is measured in percentage points; relative lift divides that change by the baseline.
Example: Activation moves from 40% to 50%: +10 percentage points and +25% relative lift. The observation still needs an uncertainty assessment.
Measurement contract
A shared definition of the event, eligible population, denominator, time window, source, owner and decision it informs.
Example: Activation means one valid plan shared within seven days of signup, excluding staff and test accounts, reported by the product analyst each Monday.
See the concepts work together
Worked example · common scale
Activity is not the same as first value
Worked example · calculation steps
GRR and NRR tell different retention stories
Worked example · common scale
When contribution actually covers acquisition
Use an app marketplace to fill one real gap
An app marketplace is a directory of extensions for a particular platform. It can help you add a missing capability to your current site or workflow. Start with the customer’s stuck step and the metric you expect to change, then search for the capability.
| Need | What to investigate |
|---|---|
| Customers receive reminders after completing the task | Look for event-based segmentation, suppression rules and reliable event sync. Test an inactive customer, a completed customer, and a delayed or duplicated event. |
| Where to look | Use your current CRM or lifecycle platform’s official integrations. The HubSpot Marketplace and Zapier’s app directory are starting points for checking supported connections, not proof that a workflow is correct. |
- Name the job. Write one sentence: “When this event happens, this person needs this useful next step.” Check whether your current plan already supports it.
- Compare two or three candidates. Check the required plan, current fees and usage limits, recent maintenance, support, permissions, supported data fields and cancellation terms.
- Calculate the full cost. Include setup and maintenance. A $60 monthly app needs three genuinely additional sales at $20 contribution per sale to cover its fee. With $120 setup, the first month needs nine such sales. Sales merely credited to the app are not necessarily additional.
- Test a complete journey. Use a safe test account or draft environment. Check success, failure, cancellation and duplicate events; verify mobile behavior and what happens after uninstalling.
- Keep or remove it on evidence. Record a baseline, an owner and a review date. Measure contribution, reliability or usable time saved. Hours freed are capacity until you actually turn them into value.
Directory links are ordinary official links. Availability, compatibility and pricing can change; verify the listing before installing. An app recommendation should explain fit and trade-offs rather than treating every popular extension as essential.
Give an AI agent a useful, reviewable job
Ask an agent to reconcile definitions, check arithmetic, research app capabilities, classify anonymized feedback or draft a weekly review. Provide the source records and your cost definitions. Keep purchases, live installations, customer messages and budget changes subject to the permissions you explicitly grant.
Review these de-identified customer cohorts. Check eligibility, event definitions, observation windows, activation, retention and recurring-revenue calculations. Separate survey responses from behavior and attribution from causal evidence. Draft a measurement contract and one experiment with a primary outcome and guardrail. Mark unknowns; do not fabricate lift.
Check the result against the original records. A polished explanation is not proof that the data is complete or that the proposed change will work.
Put five terms to work this week
- Choose one decision. What are you trying to improve or learn?
- Select five measures at most. Define the event, denominator, time window, costs and source for each.
- Calculate one real example. Reconcile it with an order, customer, invoice or observed task.
- Make one bounded change. Record the hypothesis, owner, cost limit and review date.
- Review the evidence. Continue, revise or stop. Record uncertainty and customer feedback alongside the numbers.
Turn these definitions into the weekly customer-progress scorecard and the Outcome Experiment Canvas. If a specific journey needs implementation help, see the Growth Fix scope and bring the evidence behind the stuck step.
Definitions and further reading
Numerical examples, graphics and exercises in this guide are original hypothetical teaching scenarios. They are not client results, industry benchmarks or forecasts. Different examples are independent unless explicitly connected. For platform-specific reporting definitions, consult Google Ads on CTR and Shopify on customer lifetime value. Report your own cost scope and observed cohort horizon alongside any simplified LTV model. Official marketplace directories were checked November 2, 2026.
