Performance marketing is paid media judged by what reaches your bank account: every dollar has to tie back to a sale, a qualified lead or a booked appointment. To measure it for real, you track CAC, ROAS, LTV, cost per opportunity and attributed revenue, and you check ad platform dashboards against your CRM and your books.
This guide uses aggregate numbers from the accounts under NWLAB management, 2023 to 2025, to show how each metric is calculated, why the ad dashboard almost always reports more than you actually banked, and what it costs to run this with an agency.
What is performance marketing, and how is it different from reach-based media?#
With reach-based media, a campaign is judged by how many people saw the message and how often. With performance marketing, the same budget is judged by a business event: a completed purchase, a ticket sold, an appointment booked, a proposal sent. If you buy reach, you optimize CPM and frequency. If you buy performance, you optimize cost per result and move budget to the ad, audience and channel that bring in customers for less.
The two don't compete. Les Binet and Peter Field's 2013 report for the IPA found that the optimal budget split was 60:40 between brand building and activation (IPA, 2023). That's a starting point for brands with a strategic budget, not a law. For most small and midsize businesses, activation comes first because it pays this month's bills, and brand gets more room once you know what a customer costs.
US digital ad revenue hit $294.6 billion in 2025, up 13.9%, with $114.2 billion in search and $117.7 billion in social (IAB, 2026). Almost all of that inventory is sold at auction with conversion-based optimization, which is why bad measurement is expensive: the algorithm learns from whatever event you send it.
What do NWLAB's numbers show about measuring real revenue?#
These are the totals across the accounts under NWLAB management, 2023 to 2025, all clients and channels combined:
- Attributed and tracked revenue for clients: US$ 2.4M+ (BRL 12.6M)
- Ad spend under management: US$ 500k+ (BRL 2.6M)
- Sales and tickets attributed to campaigns: 60,000+
- WhatsApp conversations generated by ads: 50,000+
What these numbers have in common is where they come from. Revenue only counts when there's a recorded event with a value (a checkout purchase, a ticket issued, a deal closed in the CRM) and a path back to the ad that brought the person in. A WhatsApp conversation only counts when it started from a click-to-message ad. Anything the platform estimates without that record is left out, which is why the total is smaller than the sum of the dashboards.
These totals mix very different industries and sales cycles: they show scale and method, not a ROAS target for your business. Your target comes from your margin.
Which metrics decide where your budget goes?#
Five metrics cover almost every decision, and they all come from ad spend, revenue, margin and customer count.
CAC: what each new customer costs#
CAC = (ad spend + management + sales cost for the period) ÷ new customers in the period. If you spent $6,110 on media, paid $890 for management and closed 28 new customers, your CAC is $250. Calculating it from ad spend alone makes the number look good and wrong.
ROAS and break-even ROAS#
ROAS = attributed revenue ÷ ad spend. On its own, it doesn't tell you whether a campaign makes money. That's what break-even ROAS is for: 1 ÷ contribution margin. At a 40% margin, your minimum ROAS is 2.5. Below that, the margin on the sale doesn't cover the ad cost. A ROAS of 4 is profitable at a 40% margin and loses money at a 20% margin, where break-even is 5.
LTV and the LTV:CAC ratio#
LTV = average order value × purchases per customer in the period × contribution margin. It tells you how much you can pay for a customer who comes back. The most cited reference comes from David Skok, for subscription businesses: LTV above 3 times CAC and CAC recovered in under 12 months (For Entrepreneurs, 2011). Outside SaaS, treat it as a warning sign, not a rule.
Cost per opportunity#
Cost per opportunity = ad spend ÷ qualified opportunities, where an opportunity is a lead that turned into a quote, a booked consultation or a proposal. It fixes the bias of cost per lead, which rewards cheap form fills. For context, the average cost per lead on search ads is $66.69, with wide gaps by industry (LocaliQ, 2026):
- All industries: $66.69 per lead, 8.18% conversion rate, $5.42 per click
- Dentists and dental services: about $72.97 per lead
- Home and home improvement: about $90.92 per lead
- Attorneys and legal services: about $131.63 per lead
Two hypothetical channels with the same $4,000 budget:
- Channel A: 100 leads at $40, 10 opportunities at $400, 4 sales, $1,000 in media per sale
- Channel B: 40 leads at $100, 16 opportunities at $250, 8 sales, $500 in media per sale
By cost per lead, channel A looks two and a half times cheaper. By sales, it costs twice as much.
Attributed revenue#
This is the sum of sales that trace back to a campaign and show up in your books or your CRM. It is not the conversion value column in the ad platform. The gap between the two is the size of your measurement problem.
Why does the ad platform report more sales than you actually banked?#
Every platform counts conversions by its own rules and gives itself the credit. If someone saw an Instagram ad, clicked a Google ad three days later and bought, both can record the same sale. The default windows show how big the overlap can get:
- Meta Ads: 7 days after a click and 1 day after a view, no click needed (Jon Loomer, 2025)
- Google Ads: 30 days after a click on Search and Display, and 1 day for view-through conversions (Google Ads Help, 2026)
- GA4: a 90-day lookback for most key events, and no first click or linear models since November 2023 (Google Analytics Help, 2026)
Adding up what each dashboard says counts the same sale two or three times. And there's a deeper problem: attribution isn't causation. In a study of 15 advertising experiments at Facebook, with 500 million observations and 1.6 billion impressions, the observational methods the industry relies on overestimated ad effectiveness, and in half the studies the estimated lift in purchases was off by a factor of three (Gordon et al., Marketing Science, 2019; MSI summary).
In practice: treat the platform number as a signal for optimization, not as revenue; check the sum of the dashboards against your books every week; and, once budget allows, measure incrementality with a geo holdout test or a marketing mix model such as Meridian, which Google released as open source (Google for Developers, 2026).
How do you build tracking that matches your books?#
Order matters. Start with the event that's worth money, and only then think about dashboards:
- 1. Define the business event and its value. A sale, a ticket, a booked appointment or an accepted proposal, always with a dollar value. No value, no ROAS.
- 2. Set up GA4 key events. Only business outcomes become key events, always with a value.
- 3. Put UTMs on every link outside Google Ads. Google says you should always use utm_source, utm_medium and utm_campaign (Google Analytics Help, 2026), with one naming convention.
- 4. Run the pixel and the Conversions API with the same event_id. Meta discards the duplicate when event_id and event name match within 48 hours (Meta for Developers, 2026). Without that, the sale counts twice.
- 5. Store the GCLID in your CRM and send the closed deal back. Offline conversion imports tell Google Ads about the sale that happened off your site (Google Ads Help, 2026). Google keeps the GCLID for only 90 days (Google Ads Help, 2026).
- 6. Connect calls and messages to the sale. Log every phone lead with its source. On click-to-WhatsApp ads, each click carries a ctwa_clid; stored with the conversation and sent back through the Conversions API for business messaging, it tells Meta which conversation became a purchase (Meta for Developers, 2026).
- 7. Make the CRM your source of truth. Every week, compare CRM sales, attributed revenue and the dashboard total.
In the US, step 6 is usually about phone calls and forms: only 32% of US adults use WhatsApp (Pew Research Center, 2025). For local businesses and med spas and aesthetics, the sale often closes on the phone or at the front desk, and without that link the campaign gets judged by the click.
How should you split budget across channels?#
There's no correct fixed percentage. There's an order of decisions:
- Start where demand already exists. If people search for what you sell, paid search captures that intent. If the product is new or an impulse buy, Meta and TikTok create the demand search doesn't have yet.
- One well-measured channel beats three running blind. Below $1,500 a month in a channel, tests run slowly and the data takes much longer to become reliable.
- Decide on marginal cost, not the average. The weekly question is where the next dollar does the most work. A channel with a good average CAC may already be saturated, and the extra dollar there can cost twice as much.
- Move budget in steps, on a schedule. Abrupt changes reset the algorithm's learning. Shift a portion, wait for the conversion window to close, then compare.
How much does it cost to hire a performance marketing agency?#
Cost comes in three parts: ad spend, which goes straight to Google, Meta or TikTok; management, which pays for the work; and, in many models, a fee tied to ad spend. US agencies quote all over the map. HawkSEM puts flat fees at $1,500 to $10,000 a month and percentage fees at 10% to 20% of the budget (HawkSEM, 2026). NWLAB's paid media pricing is public:
- Setup: $440, one time, covering account audit, pixels, events and CRM or form integration
- Monthly management: $890 per channel
- Media fee: a percentage of ad spend that goes down as spend goes up, calculated in the simulator on the page
- Recommended minimum ad budget: $1,500 a month per channel, paid directly to the platform
- [SEO](https://nwlab.com.br/en/services/seo): from $1,140 a month, with no cost per click and a payoff measured in months, tracked with the same discipline
With one channel at the minimum budget, your first month adds up to $440 in setup, $890 in management, the media fee and $1,500 in ad spend; after that, setup drops off. The ad account stays in your company's name: if you change vendors, the history and audiences stay yours.
When you compare proposals, ask what management includes (optimization cadence, reporting, offline tracking), whether ad spend runs through the agency, and who owns the account.
Which mistakes waste the most ad spend?#
- Optimizing for cheap clicks or leads. The algorithm delivers what you ask for. If the event is a form fill, it finds people who fill out forms, not people who buy.
- Links without UTMs and leads without a GCLID. The deal your sales team closes shows up as direct traffic in the report.
- Counting conversations as results. A message or a call is a step. The result is the sale that came out of it.
- Judging a campaign before the window closes. On a 30-day sales cycle, cutting in week two means deciding without data.
What does tracked revenue look like in a real case?#
The public Beetlejuice, The Musical case study applies the method: two seasons in Brazil in 2025, in Rio de Janeiro and São Paulo, with Meta Ads on a conversion objective and purchases tracked end to end through the ticketing flow:
- Ad spend: R$ 93,000 (about US$ 18,200)
- Tracked ticket revenue: R$ 1.68 million (about US$ 329,000)
- Tickets sold (pixel purchases): 6,003
- Cost per ticket: R$ 13.09 in Rio and R$ 17.28 in São Paulo (about US$ 2.56 and US$ 3.38)
- ROAS: 18.06 blended, 20.58 in Rio and 16.63 in São Paulo
The number comes from the Meta pixel, in the standard attribution window, tied to a purchase with a value, so it can be audited against the box office report. The blended number hides the dynamics: event sales pile up at opening and in the final sprint, and budget moved, in each city, to wherever cost per ticket was healthy. One season isn't a promise for the next, but the method carries over to any events operation.
If your dashboard shows one revenue number and your books show another, the first step is connecting the sale event to the ad. After that, CAC, ROAS and LTV stop being opinions.
