A campaign may get many clicks and a sales team may generate dozens of opportunities, yet the business can still end up with less money or more orders than it can fulfill.
Measuring marketing and sales is not about collecting numbers on a dashboard. It means choosing what to observe to answer a specific commercial question, linking the results to actual orders and costs, and acknowledging what the evidence cannot establish. Should we maintain a channel? Does order volume drop before payment? Are we acquiring customers who return? Is a promotion contributing enough after commissions and refunds?
Google Ads defines metrics such as conversion value per cost and conversion rate within its measurement system [R01]. Google Analytics distinguishes key events representing important behavior from conversions used to report on and optimize advertising [R02]. These definitions help implement measurement, but platform reports are not a profit-and-loss statement and cannot by themselves demonstrate causation.
1. Start with a decision, not a list of metrics
Begin with a question that can produce action. “How many views did we get?” rarely determines a business investment. A more useful question is “Should we keep paying for Channel A when it produces confirmed orders but also many refunds?” That tells us which events, costs, and quality conditions to record.
Build a simple record: possible decision, relevant metric, source, period, owner, warning threshold, information that might change the interpretation, and provisional action. Do not redefine targets after seeing the results. With a small number of orders, describe outcomes as preliminary evidence rather than reliable statistical forecasts.
Not every metric belongs solely to marketing. If a promotion attracts interest but checkout fails, the barrier may be commercial, technical, or payment-related. If accepted orders arrive late, operations is affecting the commercial result. Joint review reduces blame directed at the wrong function and prevents teams from improving their own metrics at the expense of the business.
2. Define events, units, sources, and sales statuses
A basic map is impression → click → visit → inquiry → qualified lead → opportunity → accepted sale → payment collected → fulfillment → cancellation/refund → repeat purchase. Businesses need not use every stage. An online store may not track named opportunities; a B2B provider may have meetings and proposals without thousands of impressions.
Each event requires an appropriate identifier, timestamp, channel, campaign, counted unit, status, and source. A visit is not necessarily a unique visitor; an inquiry is not a customer; an accepted order may not be paid or delivered. A customer may place several orders. Refunds after month-end must be accounted for without silently erasing the original transaction.
| Event or record | Recommended unit | Illustrative source |
|---|---|---|
| Impression or click | Ad event as defined by platform | Platform and period |
| Website visit and action | Defined session, user, or event | Owned analytics |
| Inquiry or lead | Inquiry ID; avoid duplicates | Form or recorded messaging |
| B2B opportunity | Opportunity ID and stage | CRM |
| Order and payment | Order ID; amount and status | Sales, payment records, reconciliation |
| Refund and repeat purchase | Order/customer ID and date | Operations and support |
Reconcile advertising-reported orders against sales systems and bank or payment-gateway records when appropriate. A tag can fire twice; a purchase can happen by phone; a customer can block tracking. Record the mismatch instead of silently changing values until dashboards agree.
3. Separate exposure, interaction, acquisition, and economic results
Exposure: impressions, estimated reach, and frequency. Interaction: clicks, valid clicks, visits, inquiries, and CTR. Acquisition: qualified leads, new customers, defined CPA, CPL, and CAC. Commercial outcomes: net orders, invoiced or collected value under a stated definition, cancellations, repeat business, and economic contribution. These are different levels: reach is not paid demand, and one purchase does not always mean one newly acquired customer.
To distinguish a temporary spike from recurring operating needs, compare sales and inquiries with the demand forecast.
CTR is commonly clicks divided by impressions when both are from the same campaign and period. For 200 clicks and 4,000 impressions, descriptive CTR is 200 / 4,000 = 5%. But high CTR can coexist with out-of-area inquiries, misunderstood pricing, and out-of-stock products. There is no universally “good” CTR without comparable context, objectives, and data.
When comparing B2C with B2B, define valuable conversion according to the business: a net order in a store, a qualified opportunity for a service provider, or a repeat purchase in a recurring business. Do not claim a B2B campaign “sells less” simply because it closes fewer contracts when deal size, lead time, and operating capacity differ. Monitor customer experience and fulfillment as well.
4. Calculate rates using compatible denominators and cohorts
A rate is the number of qualifying events / the number of eligible units at the starting point. The numerator and denominator must refer to the same segment and a window long enough for the process. If 40 qualified leads enter a cohort and 20 complete a purchase, lead-to-net-sale conversion is 20 / 40 = 50%, provided those 20 belong to that cohort and the treatment of open cases and refunds is defined.
Do not divide November sales by October clicks without checking cohorts and attribution windows. Nor should “attributed sales” be divided by “unique users” as if all systems share the same identity. Google notes that attribution models and lookback windows affect how credit is assigned to earlier interactions; credit may even be fractional [R03].
A cohort record should retain entry date, initial stage, observed close, still-open opportunities, exclusions, and change reasons. A B2B proposal that takes weeks to close cannot fairly be compared at an immature stage to a B2C checkout completed in minutes. Report “immature cohort” or “incomplete follow-up” rather than inventing a final rate.
5. Calculate CPL, CPA, CAC, and ROAS without calling them profit
Before using an acronym, state the numerator and denominator:
When a channel appears profitable on ROAS alone, check its real economics through costs, margins, and break-even analysis, including operating costs.
- CPL (cost per lead): defined acquisition spend / valid leads generated. If it includes only advertising, label it advertising CPL, not total acquisition cost.
- CPA (cost per action/acquisition): defined spend / defined actions, which might mean net sales. Different platforms and businesses use CPA for different events; always specify the action.
- CAC (customer acquisition cost): attributable acquisition costs over a defined period or cohort / unique new customers acquired under that definition. Advertising-only CAC excludes sales labor, creative production, and tools; it is not fully loaded CAC.
- ROAS (return on ad spend): advertising-attributed revenue or conversion value / advertising spend. The ratio does not automatically deduct production, commissions, support, refunds, or fixed costs.
In Google Ads, conversion value per cost divides the value assigned to conversions by the cost of eligible ad interactions [R01]. If the platform is configured with estimated or default conversion values instead of verified revenue, its displayed “ROAS” may represent configured values, not cash received. Always inspect what the numerator actually contains.
For business economics, additionally calculate contribution after channel = net revenue after refunds − variable costs of the orders − channel commissions − relevant acquisition spend. State which costs were excluded and why. Do not call that result net profit while fixed staffing, rent, taxes, and other obligations remain unaccounted for.
6. Use different funnels for B2B and B2C
For B2C, you may track eligible visitors, inquiries, initiated carts, confirmed orders, successful payments, deliveries, cancellations, new buyers, and repeat purchases. If phone sales exist, apply consistent definitions and flag transactions with unknown source. Calculate average order value using a consistent treatment of sales and refunds across periods.
For B2B, prioritize genuinely qualified opportunities, expected value with stated assumptions, proposals, approvals, sales-cycle duration, closed deals, and recurring contracts. A large pipeline of quotes without next steps can be less useful than a small pipeline with documented decisions. Do not apply checkout conversion rates to complex corporate opportunities or count a website visit as a negotiation.
Keep number of agreements won, contracted amount, and cash actually collected separate. A closed contract may involve monthly deliveries and later payment. This matters when deciding whether a campaign can finance more operating capacity; expected contract value does not create immediate liquidity.
7. Compare channels with hidden costs and capacity constraints
Two channels may sell the same unit volume yet demand very different resources: marketplace fees, discounts, sales time, fraud or bad debt, packaging, support, returns, collection delays, logistics, and out-of-area orders. A channel with more reach need not produce more contribution. A profitable unit sale does not justify additional orders when the operating slot is already full.
When comparing channels, also track when collections and fees settle through cash, collection, and payment controls.
Compare three layers: commercial performance (qualified inquiries and net sales), economics (contribution after costs), and capacity/quality (on-time deliveries, incidents, refunds). Explain which costs can be directly assigned and which remain estimates or unallocated.
An indirect channel may save selling time while charging high commissions. Another may have inexpensive ads but require numerous time-consuming calls. Include relevant support or sales effort when it can be measured reasonably. If it cannot, make that limitation explicit rather than declaring a winner based on one indicator.
8. Recognize imperfect attribution, missing data, and privacy
A customer might hear about the business from a friend, see an ad, compare offers on the site, and ultimately buy through messaging. Crediting the entire sale to the last click is a reporting rule, not proof that earlier contacts did not matter. Attribution models, time windows, and available data can produce different outcomes for the same campaign [R03]. Correlation should not be treated as proof of incremental sales caused by advertising.
Missing data have legitimate causes: consent requirements, blocked cookies, device switches, offline purchases, incomplete forms, duplicate tags, and privacy conditions. Respect applicable law rather than trying to fill gaps through covert tracking. Collect only needed data, explain its use where required, and set appropriate access, retention, and deletion controls.
With small samples, use ranges and notes about uncertainty. If a channel generates 3 sales one week and 5 the next, that does not establish that a creative change caused the increase. Describe the difference and identify what would be needed to investigate it. Causal inference and continuous optimization belong to the next stage of the Atlas.
9. Build a short report with decisions and owners
A useful weekly report may show channel, spend, qualified leads, net orders, new customers, net revenue, variable costs, commissions, contribution after acquisition, capacity used, incidents, and decision. A monthly version adds repeat buying, later refunds, B2B sales cycles, collections, and mature cohorts. Always specify the actual cutoff date; do not mix final results with transactions still pending.
Use four operating decisions: continue when outcomes and capacity justify it; adjust when a specific friction is observed; pause for losses, insufficient capacity, quality issues, or compliance concerns; investigate when evidence is insufficient. Arbitrary thresholds are not universal laws. Confirm that operations can absorb extra demand before expanding the advertising budget.
Record who validates the figures and who decides. Marketing can explain acquisition cost; sales can assess opportunity quality and stage progression; operations can report delivery and returns; finance can check contribution and cash collected. Even in a one-person business, these different information sources need to be compared rather than taking the ad dashboard at face value.
10. Separate daily commercial control from causal optimization
Day-to-day control asks: “What did we spend, attract, sell, collect, and deliver—and what can we reasonably do now?” Sustained optimization asks further questions: “What caused the change? What alternative explanations exist? What experiment is appropriate? How can we make a lasting improvement without creating new problems?” Those belong to the next pillar.
If a channel has high ROAS but lower contribution, reasons could include fees, refunds, delivery cost, or attribution. Each hypothesis requires different evidence. Do not maximize clicks while the business loses money on additional orders. Nor should immediate contribution be the only objective if poor service damages later purchases: state the time horizon and constraints of every decision.
Preserve definitions, sources, scenarios, and outcomes to provide a baseline for later reviews. That does not require enterprise analytics infrastructure; it requires traceability and a willingness to acknowledge uncertainty.
Worked example: two channels, identical order volume, different results
Entirely hypothetical example, in Peruvian soles for one closed reporting period. These are not real Google Ads results or market benchmarks. Each channel generated 20 delivered orders at a listed price of S/40. Variable cost to produce and deliver each order was S/24, including orders that were later fully refunded. Channel A had no refunds; Channel B had 2 fully refunded orders. For illustration only, net sales are assumed attributable to their respective channels.
| Period measure | Channel A | Channel B |
|---|---|---|
| Orders delivered | 20 | 20 |
| Full refunds | 0 | 2 |
| Revenue net of refunds | S/800 | S/720 |
| Variable cost of 20 orders | S/480 | S/480 |
| Advertising | S/100 | S/60 |
| Channel commission | S/20 | S/80 |
| Contribution after channel | S/200 | S/100 |
Calculations: A = S/800 − S/480 − S/100 − S/20 = S/200. B = (20 × S/40 − 2 × S/40) − S/480 − S/60 − S/80 = S/100. If ROAS uses assumed attributable net revenue, A = 800 / 100 = 8 and B = 720 / 60 = 12. B has higher ROAS but lower contribution after the channel. Neither figure represents the company's net profit.
For an illustrative acquisition definition, assume 10 unique new customers on A and 12 on B, counted among orders not refunded. Advertising-only CAC, which excludes other acquisition costs, would be A = 100 / 10 = S/10 and B = 60 / 12 = S/5. If CPA is defined as ad spend per nonrefunded order, A = 100 / 20 = S/5 and B = 60 / 18 ≈ S/3.33. Do not conflate customer-level CAC with order-level CPA, or present either as fully loaded acquisition cost.
A reasonable provisional decision is to monitor both channels, examine the causes of refunds and commissions on B, and avoid increasing spend before confirming contribution and capacity. This table cannot establish that advertising caused those orders; it illustrates a comparison based on assumed attribution.
Template: KPI dictionary and decision dashboard
| Field | What to write |
|---|---|
| Decision | Maintain, adjust, pause, or investigate channel/stage |
| KPI and formula | Unambiguous name, numerator, denominator, unit |
| Event / system | Ad event, CRM, order, payment, refund |
| Segment and cohort | B2C/B2B, location, entry date, cutoff date |
| Reference value | Baseline with source and comparable conditions |
| Spend and variable cost | Advertising, sales, commissions, operations, omissions |
| Outcome | Net orders, new customers, revenue, contribution |
| Quality/capacity | Delays, refunds, available slots, warnings |
| Attribution limits | Unknown touchpoints, lookback windows, missing data |
| Owner and review | Person, date, decision, unresolved evidence |
Minimum dictionary: CTR = clicks / impressions; lead-to-sale conversion = sales in cohort / eligible leads in the same cohort; CPA = defined spend / defined actions; CAC = defined customer acquisition cost / unique new customers; ROAS = defined attributable value / ad spend. If a denominator is zero, record not calculable rather than substituting zero or infinity.
Common mistakes and limitations
Mistakes include comparing gross revenue on one channel with net revenue on another; leaving refunds in unadjusted sales; treating repeat buyers as newly acquired; attributing offline purchases to ads without evidence; mixing entry and closure periods; equating ROAS with profitability; omitting fees and support costs; comparing B2B with B2C without accounting for their sales cycles; and tracking people in violation of privacy requirements. A mathematically precise rate can still be meaningless when its units or sources are inconsistent.
These formulas support commercial interpretation but do not guarantee performance. Accounting and taxation vary by jurisdiction. Examples exclude fixed costs, taxes, and many nuances of attribution. Advanced causal inference and the complete improvement program belong to Pillar 7.
What you should have by the end
A small set of useful commercial decisions; identified events and sources; reproducible KPI definitions; appropriate B2B and B2C funnels and cohorts; a dashboard combining acquisition, net sales, contribution, quality, and capacity; and a review process assigning owners and documenting limitations. The goal is to know what happened, what it costs, and which decision the evidence supports, without treating one advertising metric as the entire truth about the business.
References
- [R01] Google Ads Help, Understand your conversion tracking data. https://support.google.com/google-ads/answer/6270625?hl=en
- [R02] Google Analytics Help, Conversions vs. key events in Google Analytics. https://support.google.com/analytics/answer/13965727
- [R03] Google Analytics Help, Select attribution settings. https://support.google.com/analytics/answer/10597962