Economics and profitability
KPI
Key performance indicator
The metric you use to decide whether an experiment or channel continues, pauses or scales. In SmokeMetrics it is not “how many forms came in”.
Purpose and measurement
- What it is for
- Align the team on one question: does this spend create customers at an allowable cost?
- How it is measured
- You pick a primary indicator (CAC, MQL→SQL rate or maximum allowable CPL) and read it with the report evidence, not in isolation.
CPL
Cost per lead
Average cost of each captured contact. A low CPL with poor-quality leads destroys margin.
Purpose and measurement
- What it is for
- Comparing channels and creatives, never as the only success criterion.
- How it is measured
- Ad spend for the period ÷ number of leads. Computed by channel, campaign and creative from daily metrics and attributed leads.
CAC
Customer acquisition cost
What it costs to get a paying customer, not a form. It is the central economics KPI.
Purpose and measurement
- What it is for
- Deciding whether a niche or channel scales: if CAC exceeds expected customer value, you stop or pivot.
- How it is measured
- Ad spend ÷ number of customers (leads in won stage or confirmed payment). With no customers, CAC cannot be closed.
CPA
Cost per acquisition / action
Cost of a defined action (signup, purchase, SQL). Ads platforms sometimes use it loosely for CAC; here we separate lead (CPL) from customer (CAC).
Purpose and measurement
- What it is for
- Talking to ad platforms without treating “a lead is worth X” as the same as a customer.
- How it is measured
- Spend ÷ number of conversions of the type you defined (do not mix types).
ROAS
Return on ad spend
Estimated revenue euros per euro of media. Greater than 1 does not guarantee profit if customer value is mis-estimated.
Purpose and measurement
- What it is for
- Seeing whether Ads spend is covered by the value of experiment customers.
- How it is measured
- Estimated revenue (customers × average customer value from config) ÷ spend. If spend is 0, there is no ROAS.
Max allowable CPL
Expected value of a lead
Ceiling: above this CPL the channel destroys margin even if volume looks “cheap”.
Purpose and measurement
- What it is for
- Pausing channels whose CPL is high relative to how often a lead becomes a customer.
- How it is measured
- (Customers ÷ leads) × average customer value. Alert thresholds live in config, not on this page.
MQL rate
Marketing-qualified lead rate
What share of leads reach MQL quality or above. A low CPL with a low MQL rate is a trap.
Purpose and measurement
- What it is for
- Prioritising channels that bring qualifiable people, not just clicks.
- How it is measured
- MQLs (MQL, SQL or won stages) ÷ total leads in the same slice (channel, campaign, period).
Funnel and qualification
Lead
Experiment contact
Someone who showed intent on a smoke-test landing (visit, quiz, payment intent). Not a buyer of SmokeMetrics.
Purpose and measurement
- What it is for
- Measuring real demand for a specific offer, with source attribution.
- How it is measured
- Persisted in the leads collection with channel, campaign, creative, landing and capture time. Missing attribution is incomplete data.
Prospect
Sales opportunity
A CRM contact: either a potential buyer of the platform (site form), or an already qualified experiment lead (MQL/SQL) promoted to sales.
Purpose and measurement
- What it is for
- Not mixing “experiment leads” with “someone who wants to hire Studio”.
- How it is measured
- Pipeline with stages, score and today’s plan. Experiment economics do not treat site prospects as niche customers.
MQL
Marketing-qualified lead
A lead that clears the marketing qualification threshold (BANT and config weights). Ready for sales time, not yet to close.
Purpose and measurement
- What it is for
- Filtering noise before spending commercial time.
- How it is measured
- Answer score ≥ MQL threshold in config. The agent or rules set the stage.
SQL
Sales-qualified lead
A lead with a strong commercial fit (budget, authority, need and timeline verified). Next step is closing, not more mass filtering.
Purpose and measurement
- What it is for
- Prioritising today’s plan instead of treating every MQL the same.
- How it is measured
- Score ≥ SQL threshold in config, or a human promotion. MQL→SQL rate is a quality KPI.
BANT
Budget, Authority, Need, Timeline
Qualification frame: budget, who decides, what problem, and by when.
Purpose and measurement
- What it is for
- Not promoting a qualified prospect when we do not know if they can buy.
- How it is measured
- Questions and weights in config. Each answer adds points; the total decides MQL/SQL.
SLA
Service-level agreement (first contact)
Maximum time to first automatic contact with the lead. If it is missed, qualification drops sharply.
Purpose and measurement
- What it is for
- Honouring the minutes rule: do not leave the lead to “business hours” as the only mechanism.
- How it is measured
- Seconds between capture and firstResponseAt, compared with the config SLA. Activity is logged and retries exist.
Funnel
Conversion funnel
Landing steps (visit → quiz → payment intent / lead) and then commercial stages (contacted, MQL, SQL, won).
Purpose and measurement
- What it is for
- Seeing where people drop: ad, landing or qualification.
- How it is measured
- stepReached on the lead plus lifecycle stage. Each step is counted; conversions are not invented.
Ads and traffic
CTR
Click-through rate
What share of impressions become a click. A high CTR with a weak landing does not save CAC.
Purpose and measurement
- What it is for
- Comparing creatives when the sample is large enough, before calling a winner/loser.
- How it is measured
- Clicks ÷ impressions. Judging waits for a minimum impression count from config; then the Z-Test applies.
CPC
Cost per click
Average price of a click on the channel. It does not say whether the click is worth it.
Purpose and measurement
- What it is for
- Controlling bids and daily budget; read together with CVR and CPL.
- How it is measured
- Spend ÷ clicks (or the CPC reported by the ads platform).
CVR
Conversion rate
What share of visits or clicks become a lead (or another defined action).
Purpose and measurement
- What it is for
- Spotting whether the problem is the ad (low CTR) or the landing (low CVR / message match).
- How it is measured
- Leads (or the chosen conversion) ÷ landing views or clicks, according to the report slice. Daily views use no cookies or PII.
Impressions and clicks
Ads reach and traffic
Impression: the ad was shown. Click: someone went to the landing. Without both there is no demand experiment.
Purpose and measurement
- What it is for
- Knowing whether there is sample to pick winners and to attribute leads to the ad.
- How it is measured
- Metrics sync pulls impressions, clicks and spend from Google Ads into daily metrics and each creative.
Media
Advertising spend
Money Google charge to show ads. It is not part of the SmokeMetrics fee.
Purpose and measurement
- What it is for
- Separating “what you pay the ads platform” from “what you pay Studio”.
- How it is measured
- costMicros from the Ads APIs, aggregated by day, channel, campaign and creative.
RSA
Responsive search ad (Google)
Google Search format with several headlines and descriptions. SmokeMetrics publishes one RSA per approved variant, with attribution.
Purpose and measurement
- What it is for
- Launching copy without building the campaign by hand in Google’s editor.
- How it is measured
- Each RSA carries a creative id; metrics return to the same document for the Z-Test.
CAPI
Conversions API
Sending real conversions (qualified lead, customer) to Google Ads, with hashed identifiers and consent when required.
Purpose and measurement
- What it is for
- Training the algorithm on quality, not just clicks, without dumping clear-text PII.
- How it is measured
- Each send is logged (who, when, legal basis). Only what config allows is uploaded, with consent if the legal basis requires it.
UTM / attribution
Source parameters
Marks on the URL (and equivalents such as gclid) so you know which channel, campaign and creative the lead came from.
Purpose and measurement
- What it is for
- No lead without a source: without attribution there is no investment decision.
- How it is measured
- The lead stores channel, campaignId, creativeId, landingId and, if present, keyword and UTM. It is a product requirement, not an extra.
Z-Test
Two-proportion significance test
A statistical calculation to decide whether a creative or variant really wins, not from four lucky clicks.
Purpose and measurement
- What it is for
- Marking winner/loser and shifting budget when there is sample and the config confidence level is met.
- How it is measured
- Z-score and p-value on rates (e.g. clicks or conversions / impressions or visits). The confidence threshold comes from config.
Message match
Ad → landing consistency
The landing must repeat the ad’s promise. If it does not, it is a conversion bug, not a copy nuance.
Purpose and measurement
- What it is for
- Raising CVR and meeting policies (do not surprise the visitor with a different offer).
- How it is measured
- Qualitative review in the creatives flow plus CVR of that ad/landing pair versus others.
Product and compliance
Smoke test
Demand smoke test
A short experiment: offer, landing and real traffic to see if there are qualifiable leads before you build the product or scale Ads.
Purpose and measurement
- What it is for
- Lowering the cost of being wrong. It does not fake an operational product or a fake checkout.
- How it is measured
- Traffic, attributed leads, qualification, payment intent (Stripe) and economics. The verdict is scale, pivot or stop.
CTWA
Click-to-WhatsApp
An ad that opens WhatsApp. The user starts the chat. In the EU that does not authorise bulk marketing; marketing opt-in is a separate check.
Purpose and measurement
- What it is for
- Fast contact when the channel is enabled in config. Today it may be off.
- How it is measured
- Leads with channel whatsapp_ctwa and a transactional first message. Not used as a broadcast.
GDPR / opt-in
Data protection and consent
In the EU, marketing email or WhatsApp needs explicit, granular, unchecked consent. Replying to a form may be transactional.
Purpose and measurement
- What it is for
- Running leadgen without copying other markets (e.g. WhatsApp-first with no opt-in).
- How it is measured
- Separate checkboxes, a consent log, and CAPI only with a legal basis and hashed IDs.