The Map Is Not the Pump: How to Prove Marketing ROI for a Small Business

Law Smith

Founder & President, Tocobaga

Law Smith is the Founder and President of Tocobaga, a Tampa-based ROI-focused marketing agency and SMB advisory. He has strategically advised 1,000+ small and medium businesses, executed 600+ integrated marketing campaigns, and written 30+ business plans across industries ranging from professional services to e-commerce to B2B technology. Tocobaga is a Google Partner and Squarespace Gold Partner headquartered in Ybor City.

Henry Whitehead did not believe in the water. In September 1854 he was an assistant curate at St. Luke's church in Soho, London, and like nearly every educated person in the city he believed cholera rode in on bad air.* Soho smelled like a reason. What Whitehead had instead of a theory was a parish and a habit. When a parishioner died, he went to the house, sat in the front room with whoever was left, and asked what had happened in the days before. That month the habit became a full-time job. In the first three days of the outbreak, 127 people died in or around Broad Street, and by September 10 the count of fatal attacks had reached 500.† He had a lot of doors to knock on.

Across town a physician named John Snow had a different theory and no way to prove it from a desk. Snow thought cholera was swallowed, not inhaled, and in 1854 there was one way to test that: walk the streets and write down who died where. He later recorded that "nearly all the deaths had taken place within a short distance" of the Broad Street pump.* Then he kept walking, because the deaths were not the interesting data. The exceptions were. A brewery on Broad Street employed about 70 men and lost none of them; the men drank their daily beer allowance, not the pump water.† The Poland Street workhouse, 530 inmates on its own well, recorded five cases.† A widow in Hampstead, miles from Soho, died on September 2 after having a bottle of Broad Street water delivered to her door because she liked the taste.† On the evening of Thursday, September 7, Snow stood in front of the Board of Guardians of St. James's parish and made the case. The next day the handle came off the pump.*

Here is the part the posters leave out. Whitehead, the skeptic, kept knocking. Snow's map convinced him, and then he did the thing the map could not do: he traced the outbreak backward, house by house, until he reached 40 Broad Street. There, a police constable's wife named Sarah Lewis had been nursing a five-month-old daughter, Frances, who had cholera. Frances died on September 2. Her mother had been soaking the soiled linens and tipping the water into the cesspool in front of the house.‡ That cesspool sat about three feet from the well that fed the pump.* The man who did not believe in the water had found the exact spot where the water went bad, and he put it in writing. By then the handle was already back on. Once the epidemic subsided, officials had replaced it, and the government inquiry that followed ruled that the outbreak had been caused by bad air.*

Law Smith, founder and president of Tocobaga, a marketing agency and SMB advisory headquartered in Ybor City, Tampa, that publishes at SolvingHow.com, has watched that inquiry replay in conference rooms across 1,000+ advisory conversations. An owner spends money on marketing. An agency delivers a map: impressions, reach, clicks, a line that goes up and to the right. Everyone nods. Nobody can point to the pump. The only question that matters in a marketing report is the one Whitehead asked at every door: not whether something happened, but what caused what, and how anyone knows. Proving marketing ROI for a small business is not a math problem. It is a cesspool-finding problem. The arithmetic is the easy part.

What is the one sentence every marketing report has to answer?

The sentence is: this spend produced these leads, which became these customers, who paid this much, and here is how the business knows. A report that cannot be boiled down to that sentence with real numbers in every blank is a map of deaths with no pump on it, however good the chart looks.

The numbers do not have to be flattering. They have to be traceable. "In the third quarter the firm spent $9,000 on marketing. It produced 61 logged consultations. Of those, 14 signed. Those 14 new matters were worth $71,000 in fees. The firm knows because every consultation was recorded with its source the day it came in, and every signed client was matched back to a consultation on the first business day of the month." That is a complete sentence. It can be argued with. It can be wrong, and if it is wrong, the error can be found. "Reach was up 34 percent" cannot be wrong, because it does not claim anything.

The people with the largest analytics budgets in the country struggle with the same sentence. The CMO Survey, run out of Duke University's Fuqua School of Business, polled 281 senior marketing leaders in early 2025, and 64.0 percent named demonstrating marketing's impact on financial outcomes as a challenge, the most-cited answer on the list and up from 61.2 percent two years earlier.§ The same respondents reported pressure to prove marketing's value from the CFO (63 percent), the CEO (61 percent) and the board (50 percent).§ Those are companies with data teams. A two-partner law firm in Tampa has a paralegal and a Google Sheet. The comfort is that the two-partner firm has a far shorter chain of custody between the dollar going out and the dollar coming in, which means the sentence is easier to write at small scale, if anyone bothers.

The gap between believing the number and being able to produce it is the entire problem. Nielsen surveyed nearly 2,000 global marketers in December 2023 and found that 84 percent reported high confidence in their ROI measurement, while only 38 percent evaluated holistic ROI by measuring traditional and digital channels together.‖ Confidence at 84, method at 38. That 46-point gap is the Board of Health looking at Snow's map, nodding, and putting the handle back on the pump. The feeling of knowing is not the same as the record of knowing, and a monthly marketing report exists to replace the first with the second.

How does a small business know if its marketing is working?

A business knows its marketing is working when it can follow an unbroken line from a specific activity to a specific lead to a signed customer to a dollar amount, with each link recorded at the moment it happened rather than reconstructed from memory at the end of the quarter. Anything short of that chain is a feeling, and feelings are what the weaker agencies are counting on.

Most small businesses have never built the chain. The owner asks the question at the counter, remembers the answers that confirm what the owner already believes, and the agency sends a PDF nobody reconciles against the bank statement. The dashboards have gotten prettier every year. The habit has not changed.

The chain has four links, and each one needs its own record. The first link is activity: a campaign ran, a post went out, a postcard dropped, with a date and a cost. The second is the lead: a phone call, a form, a walk-in, a direct message, a referral email, logged with the date, the source and a name. The third is the customer: the lead signed, bought or booked, with the date and the amount. The fourth is revenue over time: what that customer paid in month one, month six, month eighteen. Break any link and the one sentence cannot be written. Most small business reporting breaks at the second link, because leads arrive through six doors and nobody is standing at five of them with a clipboard.

Scale changes the doors, not the logic. A PE-backed HVAC platform with eleven locations has eleven phone numbers, eleven Google Business Profiles, a call center and a sponsor who wants cost per booked job by branch on the second Tuesday of the month. A regional 3PL has three salespeople, a trade show calendar and a sales cycle that runs six months, so the activity that produced this quarter's contract happened in a different fiscal year. A Shopify store has the cleanest chain in commerce, a click to a cart to a receipt, and still gets misled by its own dashboards for reasons that take a whole section to explain. Whitehead's method scales to all of them: stand at the door, ask the question, write the answer down, then check the answer against the map.

What is the difference between vanity metrics and decision-grade numbers?

A vanity metric measures what a platform did; a decision-grade number measures what the business got. Impressions, reach, follower counts, likes, comments, shares and video views describe activity, and none of them can be deposited, so none of them can decide a budget.

The useful test is whether a number can change a decision. If reach doubles next month, what does the owner do differently? Nothing, because reach is not connected to anything the owner controls or collects. If cost per signed client doubles next month, the owner pauses the channel, calls the agency and asks for the lead log by Friday. That is a decision-grade number. There are five of them that matter for nearly every small business, and together they take about four minutes to explain.

Cost per lead (CPL) is spend divided by qualified leads from that spend. Cost per acquisition (CPA), often called customer acquisition cost (CAC) once the agency fee, software and staff time are loaded in, is spend divided by customers won. Lifetime value (LTV) is what an average customer pays over the full relationship, net of the cost to serve them, not just the first invoice. Payback period is how many months of a customer's margin it takes to earn back the cost of acquiring that customer. The ratio of LTV to CAC is the single number a private equity associate will ask for in the first ten minutes of diligence, and the single number an owner should be able to say without opening a laptop. Every one of these requires the four-link chain. None of them appears natively in a Meta Ads dashboard, because Meta does not know what a customer is worth to the business; it only knows what a click cost.

Snow's map had both kinds of number available to it, which is why it worked. He could have reported how many streets he walked, how many houses he passed, how many people he spoke to: activity, all real, all useless. What he put on the map was deaths by household, by address, against the location of the pump: outcome, by source.* An agency that reports reach is reporting streets walked.

The comparison below sets the numbers that fill a dashboard next to the numbers that decide a budget.

Vanity metrics versus decision-grade numbers: what each one measures and what an owner can do with it

Number	What it measures	Who produces it	Decision it can change
Impressions and reach	How many times a platform showed something	The ad platform	None on its own
Likes, shares, follower counts	Activity on the platform	The social platform	Creative direction at most
Platform-attributed conversions	Observed plus modeled actions the platform credits to itself	The ad platform	Bid and creative tests, not budget
Cost per lead (CPL)	Spend divided by logged, sourced leads	The business's lead sheet	Which channel gets the next dollar
Cost per acquisition (CPA or CAC)	Fully loaded spend divided by customers won	Monthly reconciliation	Whether a channel stays on
Lifetime value (LTV)	What a customer pays over the relationship, net of cost to serve	Invoicing system by cohort	How much is allowed per acquisition
Payback period	Months of margin to recover acquisition cost	Cohort table	How fast the business can scale spend

Why did iPhone privacy changes and cookie rules make attribution messier?

Because the ad platforms lost the ability to watch one person move from an ad to a purchase, and replaced observation with estimation. Since the spring of 2021 a meaningful share of the conversion numbers inside Meta, Google and TikTok dashboards are modeled, which is a polite word for inferred, and an inference does not reconcile to a bank statement.

The first break was Apple's App Tracking Transparency prompt, which arrived with iOS 14.5 in April 2021 and required every app to ask permission before tracking a user across other companies' apps and websites. Most people said no. Flurry Analytics, watching 5.3 million daily users, reported that only about 4 percent of U.S. users opted in during the first weeks, and about 12 percent worldwide.¶ Facebook, Instagram and every app that fed their pixels went mostly blind on the iPhone within a month. By September 2021 Facebook's vice president of product marketing, Graham Mudd, was telling advertisers the company had been underreporting iOS web conversions by roughly 15 percent, with wide variation from one advertiser to the next.** Read that from the owner's chair: the platform's own report of how many sales it produced was wrong by a number the platform itself could only estimate.

The second break never fully happened, which somehow made things worse. Google spent years announcing the death of the third-party cookie in Chrome, the tracking file that lets a retargeting ad follow a visitor around the internet, and the entire ad industry rebuilt its measurement around the funeral. Then in July 2024 Google's vice president for Privacy Sandbox, Anthony Chavez, announced the company would not phase cookies out after all and would instead let Chrome users make an informed choice about them.†† The cookie survived. The confidence did not. A small business now runs campaigns on an infrastructure that was told to prepare for a world that was then canceled, with vendors mid-migration and measurement models built for both.

The third break is the quiet one. When a platform cannot observe a conversion, it fills the hole with a model. Google's own documentation for Google Analytics 4 says the system uses modeling to estimate key events that cannot be observed directly, for reasons including consent, browser limits on cookies, journeys that cross devices and traffic affected by Apple's tracking policy, and that modeled events are only reported when the model's confidence is high; otherwise they go unreported or get attributed to direct.‡‡ That is an honest description of a weather forecast. It is not a receipt. A platform dashboard that says 41 conversions is saying something closer to "some were observed, the rest were inferred, and the inference is a best effort." For a national retailer with a million transactions, modeling averages out. For a two-partner law firm whose month is nine signed clients, one modeled conversion is 11 percent of the truth.

Why does asking customers how they heard about the business work so well?

Because the customer is the only witness who saw the whole journey, and asking is nearly free. A required question on every form and every intake call about how the customer heard of the business is the cheapest attribution layer a business can run, and in head-to-head tests it catches the sources that the software cannot see at all.

The software blind spot has a name, dark social, and it is bigger than most owners assume. In April 2023 Rand Fishkin of SparkToro and Steve Lamar of Really Good Data ran a controlled test: 1,113 visits over ten days to 16 test pages, each page tied to one network so every visit could be traced regardless of what the analytics tool recorded. Every visit that arrived from TikTok, Slack, Discord, Mastodon and WhatsApp showed up in Google Analytics as direct traffic, no source at all. Three-quarters of visits from Facebook Messenger did the same.§§ When a client forwards a firm's name to a friend over a text message and the friend types in the address, the software files that referral under direct, which is analytics for "no idea." A partner's best referral source and the owner's untraceable mystery traffic are frequently the same line on the report.

The gap shows up in revenue, not just visits. Refine Labs, a B2B demand agency, ran a twelve-month comparison on 620 of its own conversions, with HubSpot's software attribution on one side and a free-text field asking how the customer heard about the company on the other. The software credited 78 percent of conversions to web search. The customers themselves credited search with 12 percent and credited social media, podcasts, word of mouth and communities with 85 percent. On closed-won revenue the split was wider still, with customers crediting those untrackable channels 98 percent of the time.‖‖ One company's data, one industry, and the sample is small, so the exact percentages travel poorly. The shape of the finding travels fine: the software over-credits the last click, and the human over-credits the thing that actually persuaded them. The business needs both witnesses.

Self-reported attribution is not a clean confession either. People forget, people guess, people name the last thing they saw rather than the first, and a meaningful share will skip the question when it is optional. The fix is to make it required, keep the answer list short (eight to twelve options plus "other" with a text box), train whoever answers the phone to ask it in the first thirty seconds, and record it in the same row as the lead's name and date. For the HVAC platform, that means the call center script. For the law firm, the intake paralegal. For the Shopify store, a post-purchase survey on the order confirmation page, which gets a response rate the pre-purchase form never will. Whitehead was running self-reported attribution in 1854, one front room at a time, and it found the cesspool the map could not.

What tracking setup can a small business implement this week?

The setup is six pieces, none of which requires a developer, and most of which an office manager can finish in a few afternoons: call tracking numbers, a required source question on every form and call, source tagging in the CRM or a spreadsheet, UTM tags on every link the business controls, a monthly lead-to-client reconciliation, and cohorts by month. Tocobaga has installed some version of this chain for clients across 600+ campaigns, and the pattern is always the same: the pieces are simple, and the discipline is the whole product.

Step one: put a unique phone number on every channel

Each marketing channel gets its own tracking phone number that forwards to the main line. The Google Business Profile gets one, the website gets one, the paid search ads get one, the postcard gets one, the truck wrap gets one. When the phone rings, the number that was dialed is the source, no memory required. Google Ads has call reporting built in: it swaps a Google forwarding number into the ad, records how long the call lasted and when it started, and lets a call over a set length count as a conversion. A dedicated call tracking service does the same for every other channel, costs less than a lunch per number per month, and records the calls so the owner can hear what a $90 lead sounds like.

Step two: make the source question required

Every form, every intake call, every walk-in asks the same question: how did this customer hear about the business? Required, not optional. Keep the list short and plain: Google search, Google Maps, a friend or family member, another professional, social media (which one), our newsletter, a sign or vehicle, a mailer, an event, other. Pick the list once and never change the wording, because a question that changes every quarter cannot be compared across quarters.

Step three: tag every lead with a source in one place

A CRM if the business has one, a Google Sheet if it does not. The spreadsheet version has columns for date, name, contact, source (from the phone number or the required field), channel detail, status, date signed, and first-year value. Nothing fancy. The rule is that no lead exists until it has a row, and no row is complete without a source. A lead with no source is written down as "unknown," never as a guess, and the share of unknowns is itself a metric: above 15 percent means somebody is skipping the question.

Step four: put UTM tags on every link the business controls

A UTM tag is a short code added to the end of a link that tells analytics where the click came from. Google's own URL builder documentation defines the core parameters: utm_source identifies the referrer (a newsletter, a site), utm_medium identifies the medium (email, cpc, social), utm_campaign identifies the specific promotion, and utm_term and utm_content split keywords and creative variations.¶¶ A link in the newsletter, a link in the Google Business Profile, a link in an Instagram bio, a QR code on a postcard: each gets its own tag. Ten minutes of setup, done once, and the "direct" bucket shrinks every month afterward.

Step five: reconcile leads to clients once a month

On the first business day of the month, someone takes the lead sheet and the list of new customers from the invoicing or practice management system and matches them, one to one. Every new customer gets matched back to a lead row and inherits its source. Every lead that signed gets its first invoice amount filled in. This is the step nearly everyone skips and the only step that produces the one sentence. It takes an hour for a firm doing forty leads a month. It is the meeting Snow had with the Board of Guardians: the point where the walking becomes a decision.

Step six: group customers by the month they became leads

A cohort is a group of customers sorted by the month they first appeared, tracked forward over time. March's leads become March's cohort. Six months later the sheet shows what March's cohort cost, what it has paid so far, and how that compares to February's. This is how a six-month sales cycle stops looking like a wasted quarter: the 3PL's February trade show looks like a loss in March and a win in August, and only a cohort view shows both. Payback period lives here, and so does the honest answer to "is this channel getting better or worse."

That is the full system. Written as a checklist it fits on an index card:

• One tracking phone number per channel, all forwarding to the main line.

• The source question, required, same wording forever.

• One lead sheet or CRM; no lead without a row, no row without a source.

• UTM tags on every link and QR code the business controls.

• First business day of the month: match new customers to lead rows.

• Cohorts by lead month; report cost, revenue and payback per cohort.

How do cost per lead, cost per acquisition and lifetime value connect to revenue?

They connect in one straight line of division, and the arithmetic is simple enough that the owner should be able to do it on a napkin while the agency is still loading its slides. The worked example below uses a two-partner estate planning and probate firm, because a law firm's numbers are clean and its margins are high enough that the lesson is unmistakable.

The firm spends $4,000 a month on marketing: $2,500 on Google Ads and $1,500 on an agency fee to manage the ads, the Google Business Profile and a monthly email to past clients. In March the lead sheet shows 38 logged consultations with a source attached. Cost per lead is $4,000 divided by 38, or about $105. Of those 38, nine sign a fee agreement. Cost per acquisition is $4,000 divided by nine, or about $444 per new client. The firm's average first-year value per client, pulled from last year's invoices, is $2,900. Nine clients at $2,900 is $26,100 in first-year revenue against $4,000 in spend. Spend returns about 6.5 times in gross revenue, and the ratio of first-year value to acquisition cost is $2,900 to $444, roughly 6.5 to 1 before the cost of delivering the work.

Now the sources, because the blended number hides the decision. The sheet says 22 of the 38 consultations came from Google Ads, 11 from the Google Business Profile and referrals, and five from the email to past clients. Of the nine signed, four came from the ads, three from the profile and referrals, and two from the email. The ads cost $2,500 (plus most of the agency's attention) and produced four clients: about $625 per client. The email cost almost nothing beyond the fee and produced two: the cheapest clients the firm signed all month. The profile and referrals produced three at a cost that rounds to the agency's time. The blended $444 was true and useless. The source-level numbers tell the firm to keep the ads (a $625 client worth $2,900 in year one is a fine trade), to send the email twice a month instead of once, and to ask every signed client for a Google review, because the profile is outperforming its budget.

Then the cohort view, six months later. The March cohort of nine clients has paid $31,400 so far, because two probate matters ran past the first-year average. Payback on the $4,000 was reached inside the first month. February's cohort, by contrast, had 41 leads and seven clients, a $571 acquisition cost, and $19,300 collected at the same point. Same spend, same firm, same agency, a 29 percent swing in acquisition cost. The answer is in the February sheet: a tracking number was left off the new landing page for two weeks, 14 leads were logged as "unknown," and the cheapest channel got no credit. The data did not lie. The clipboard was missing from one door.

One paragraph on the number the owner should also run: adjusted ROI, with the agency's fee counted as a cost, not as part of the spend that produces results. The firm's $4,000 included $1,500 in fees. On the ad spend alone, $2,500 producing $26,100 reads like a 10.4 times return. On the full $4,000 it is 6.5 times. The second number is the true one, because the fee does not come back, and a report that quotes the first number without the second is already juking. Tocobaga quotes the second number, which is one reason it charges a flat fee rather than a percentage of ad spend: a fee that grows with the budget has an incentive to grow the budget.

Why do most agency marketing reports fail this test?

Because they measure what the platform produced instead of what the client collected, and because the incentives of the person writing the report point away from the one sentence. An agency paid as a percentage of ad spend earns more when the spend grows, not when the acquisition cost falls, and a report built on reach, impressions and platform-attributed conversions will always look better than a report built on signed clients reconciled to invoices.

The failure has a famous dramatization. In the third-season premiere of The Wire, which aired on HBO in September 2004, the Baltimore police commissioner tells his district commanders at a ComStat meeting to cut the felony rate to the mayor's target. Major Howard Colvin, watching his peers water down their figures, asks how exactly a man is supposed to juke the stats on a murder victim.*** The show's phrase for the practice, juking the stats, entered the language because everyone who has ever sat through a quarterly review recognized it: the number went up because the definition changed, not because the thing changed. A marketing report that reclassifies "leads" to include every form fill, every chatbot ping and every newsletter signup is a district commander turning robberies into larcenies. The count improved. The city did not.

There are four tells that a report is juking. The first is that it reports platform conversions without a reconciliation to the client's own customer list; Meta and Google will both count a conversion that the invoicing system never saw. The second is that it reports blended numbers only, never by source, which hides the channel that is losing money inside the one that is making it. The third is that there is no cohort view, so the lifetime value and payback period that justify the spend are asserted rather than shown. The fourth is the vocabulary: when the first page is about reach and the last page is about revenue, the order is the confession. The bigger the ad budget being managed, the more often these tells appear together, which is not a coincidence. Everyone wants more budget, and the report is the argument for it. A report that cannot prove the current budget worked makes a weak case for the next one, so the report finds something that went up.

None of this requires a villain. The agency usually believes its numbers, the same way the Board of Health believed in bad air. Snow's theory was rejected after the outbreak because miasma was the professional consensus, the vocabulary everyone had trained in, and the map was asking the whole establishment to admit it had been looking at the wrong thing.* Platform attribution is the miasma of 2026. It is what the tools report by default, what the certifications teach, what the dashboards are built to show. The owner who insists on the reconciled number is Whitehead with a notebook, and will get the same reception from some vendors that Whitehead got from the inquiry.

How long does it take to prove whether a marketing channel works?

One full sales cycle plus one reconciliation, which for most local service businesses is about 90 days, for a law firm 90 to 180 days depending on practice area, and for a B2B company with a long consideration period six to twelve months. The error most owners make is judging a channel on its first thirty days, which is like pulling the pump handle on a Thursday and judging the outcome by Friday lunch.

Snow's story carries its own warning. The handle came off on September 8 and the deaths kept falling, but the outbreak had already been subsiding for several days before, and Snow himself acknowledged as much.† The lesson for a marketing report is that a number moving in the right direction after a change is not proof that the change caused it. A law firm that launches ads in January and sees more clients in February has also hit tax season, the month everyone remembers they do not have a will. The cohort view is the defense: compare January's cohort against last January's, not against December's.

The sales cycle also determines where the lead sheet needs patience built in. An HVAC platform's residential service lead converts or dies within a week; a report can close out each month cleanly. A 3PL pitching a regional retailer on a warehousing contract may log the lead in February, present in May, and sign in August, and the February cost sits in the sheet as a loss for six months. The reconciliation step handles this if the lead row is never deleted, only updated. Lead rows are evidence. Evidence stays in the file.

Speed matters on the front end too, and it is measurable. A study published in Harvard Business Review in 2011 by James Oldroyd, Kristina McElheran and David Elkington audited 2,241 U.S. companies by submitting a web lead and timing the response. The average response, among companies that answered within 30 days, was 42 hours; 37 percent responded within an hour, and 23 percent never responded at all. Firms that made contact within an hour were nearly seven times as likely to qualify the lead as firms that waited even an hour longer, and more than 60 times as likely as those that waited a day.††† A tracking system that logs when the lead arrived and when the business responded will often find that the "underperforming channel" is a perfectly good channel feeding a slow inbox. Attribution cannot credit a channel for a lead nobody called back.

What should a small business marketing report actually include?

One page, in this order: spend by channel, leads by channel, customers by channel, revenue by cohort, the five decision-grade numbers, and a short list of what changed and what will change next month. A report that needs more than one page to answer the one sentence is hiding something in the appendix.

The first block is the four-link chain by channel, in a plain table: channel, spend, leads, customers, first-year revenue, CPL, CPA. The second block is the cohort table: lead month down the side, months since lead across the top, revenue collected in the cells, with payback month marked. The third block is the five numbers for the whole business: CPL, CPA (fully loaded, agency fee included), LTV, LTV to CAC ratio, payback period. The fourth block is the unknowns: how many leads arrived with no source, and what is being done about it. The fifth is a paragraph of decisions, written in the plain voice of a person who is accountable for them: we are moving $500 from the display campaign to the email program because the email cohort paid back in month one and display has not paid back since November.

Who the report is for changes the emphasis, not the structure. A managing partner wants the practice-area split, because estate planning leads and litigation leads have different values and different cycles. A private equity operating partner wants the same table by branch and the LTV to CAC ratio trended over eight quarters, because that ratio is a line item in the next valuation. An e-commerce founder wants contribution margin after shipping and returns next to CPA, because a $38 customer acquisition cost on a product with $31 of margin is a losing business with a beautiful dashboard. In every case the report answers the one sentence first and decorates afterward.

The hard part is not the template. It is the two hours a month of reconciliation that fills it honestly, and the willingness to publish a bad month without redefining what a lead is. Across 1,000+ advisory conversations the owners who end up with the cleanest marketing numbers are rarely the most analytical. They are the ones who treated the first business day of the month like a court date.

Is it worth hiring an agency to measure marketing ROI?

Only if the agency's report is built on the client's reconciled customer data rather than the platform's attributed conversions, and only if the client owns every account, number and spreadsheet in the system. An agency that measures its own work on the platform's numbers is grading its own exam with a pencil the platform supplied.

The owner's checklist for any agency, or for an in-house marketer, is short. Does the report show customers, not just leads, by source? Is there a monthly reconciliation against the client's invoicing or practice management system, and who does it? Are the tracking numbers, the analytics property, the ad accounts and the lead sheet registered in the client's name, so the data stays if the agency goes? Is the fee flat, or does it rise with the ad budget? Does the contract end when it ends, or does it renew unless someone remembers to cancel? Is the agency willing to say a channel is not working, or that the business is not a fit for its services?

This is the set of questions Tocobaga built itself to answer in the affirmative, and it is why the agency bills a flat fee rather than a percentage of ad spend, never auto-renews a contract, puts every account and asset in the client's name from day one, and keeps its own branding off client property. It is also why Tocobaga will tell a business when the fit is wrong and point it somewhere better; a tracking system that proves a channel does not work for a particular business is doing its job, and so is the advisor who says so. An owner who wants a second set of eyes on the current report, or help installing the six-piece system before the next quarter, can book an intro call at (813) 534-1261 or through SolvingHow.com. Bring the lead sheet, however ugly. Ugly lead sheets are the ones with the cesspool in them.

Whitehead kept knocking. The pump handle went back on, the inquiry blamed the air, and the curate who had not believed in the water is the one history credits with finding the source: the house, the cesspool, the baby, because he had walked to the door and written down what he found.‡ Nobody remembers the inquiry. Everybody remembers the map, and the map only mattered because somebody checked it against the front rooms. A marketing report is a map. It proves nothing until the owner, or someone the owner trusts, goes door to door and asks the question at every one.

How do small businesses prove marketing ROI?

A small business proves marketing ROI by connecting each dollar of spend to logged leads, then to signed customers, then to revenue collected over time, with every link recorded when it happens. The practical tools are a tracking phone number per channel, a required question about how each customer heard of the business, one lead sheet with a source on every row, UTM tags on owned links, and a monthly match of new customers back to their lead rows. The result is one sentence: this spend produced these customers, worth this much.

What is a good marketing ROI for a small business?

There is no universal benchmark, because margins differ by industry, but the useful test is whether first-year customer value covers fully loaded acquisition cost several times over and pays back within a few months. A law firm signing a $2,900 client for $444 in total cost, agency fee included, is in healthy territory. A product business with $31 of margin and a $38 acquisition cost is losing money regardless of what the dashboard reports. Judge the ratio and the payback period, not the raw percentage.

What is the difference between CPL, CPA and LTV?

Cost per lead (CPL) is marketing spend divided by the qualified leads it produced. Cost per acquisition (CPA), also called customer acquisition cost, is the fully loaded spend, including agency fees and tools, divided by customers won. Lifetime value (LTV) is the revenue a customer pays over the whole relationship, net of the cost to serve them. CPL measures the top of the funnel, CPA measures the close, and LTV decides how much acquisition the business can afford.

Why does marketing attribution not match actual sales?

Platform attribution counts conversions the platform observed or modeled, not customers the business invoiced. Since Apple's App Tracking Transparency prompt in 2021 and years of cookie uncertainty in Chrome, a growing share of reported conversions are estimated by machine learning rather than observed, and referrals that arrive by text message or word of mouth show up as direct traffic with no source. The only reliable fix is a monthly reconciliation of platform numbers against the business's own customer list.

Is asking customers how they heard about the business accurate enough to rely on?

It is accurate enough to be the cheapest and most revealing layer of attribution a small business has, as long as it is required rather than optional, asked the same way every time, and recorded in the same row as the lead. Customers misremember and name the most recent touch, so the answer should be checked against tracking numbers and UTM data rather than used alone. In controlled comparisons the self-reported answer consistently catches referral, podcast and social sources that software files as direct.

What should a monthly marketing report for a small business include?

One page: spend, leads, customers and first-year revenue by channel; a cohort table showing revenue collected by the month leads arrived; the five decision-grade numbers (cost per lead, fully loaded cost per acquisition, lifetime value, the ratio of lifetime value to acquisition cost, and payback period); the count of leads with no known source; and a short paragraph of decisions for next month. Impressions, reach and follower counts belong in an appendix, if anywhere.

Sources

*

Wikipedia. "1854 Broad Street cholera outbreak." Snow's map and quote, the Board of Guardians meeting of September 7 and removal of the handle the following day, the cesspit three feet from the well, Whitehead's conversion, and the official attribution of the outbreak to miasma. https://en.wikipedia.org/wiki/1854_Broad_Street_cholera_outbreak

†

UCLA Fielding School of Public Health, John Snow site. "Broad Street Pump Outbreak." Excerpt from Judith Summers, Soho: A History of London's Most Colourful Neighborhood (Bloomsbury, 1989), pp. 113-117. 127 deaths in the first three days, 500 fatal attacks by September 10, 616 by the end of September; the brewery, the Poland Street workhouse, the Hampstead widow. https://www.ph.ucla.edu/epi/snow/broadstreetpump.html

‡

Cholera and the Thames (Westminster Archives project). "Cholera in Soho: Dr John Snow and Reverend Whitehead." Frances Lewis, five months old, daughter of Thomas and Sarah Lewis of 40 Broad Street, died September 2, 1854; wash water emptied into the cesspool. https://www.choleraandthethames.co.uk/cholera-in-london/cholera-in-soho/

§

The CMO Survey (Duke University, Fuqua School of Business). "The CMO Survey: Highlights and Insights Report." March 2025. Fielded January 21 to February 12, 2025; 281 marketing leaders. 64.0 percent cite demonstrating impact on financial outcomes as a challenge; pressure from CFO 63 percent, CEO 61 percent, board 50 percent. https://cmosurvey.org/wp-content/uploads/2025/03/The_CMO_Survey-Highlights_and_Insights_Report-2025.pdf

‖

Nielsen. "Nielsen Releases Its 2024 Annual Marketing Report Surveying Global Marketers on ROI Strategies." April 25, 2024. Nearly 2,000 global marketers surveyed December 2023; 84 percent confident in ROI measurement, 38 percent measure holistic ROI. https://www.nielsen.com/news-center/2024/nielsen-releases-its-2024-annual-marketing-report-surveying-global-marketers-on-roi-strategies/

¶

Lovejoy, Ben. "Unsurprisingly, Almost No Americans Are Opting In to App Tracking." 9to5Mac, May 7, 2021. Flurry Analytics data on 5.3 million daily users: about 4 percent U.S. opt-in, about 12 percent worldwide. https://9to5mac.com/2021/05/07/opting-in-to-app-tracking/

**

CNBC. "Facebook Says It Underreported Ad Performance After iPhone Privacy Update." September 22, 2021. Graham Mudd's estimate of roughly 15 percent underreporting of iOS web conversions. https://www.cnbc.com/2021/09/22/facebook-says-it-underreported-ad-performance-after-iphone-privacy-update.html

††

Chavez, Anthony. "A New Path for Privacy Sandbox on the Web." Google Privacy Sandbox blog, July 22, 2024. Google will not deprecate third-party cookies in Chrome and will instead offer users an informed choice. https://privacysandbox.google.com/blog/privacy-sandbox-update

‡‡

Google Analytics Help. "[GA4] About Modeled Key Events." Modeling estimates key events that cannot be observed directly because of consent, cookie limits, cross-device journeys and Apple's tracking policy; modeled events reported only at high confidence. https://support.google.com/analytics/answer/10710245?hl=en

§§

Fishkin, Rand. "New Research: Dark Social Falsely Attributes Significant Percentages of Web Traffic as 'Direct.'" SparkToro, April 27, 2023. 1,113 visits over ten days to 16 test pages; 100 percent of visits from TikTok, Slack, Discord, Mastodon and WhatsApp recorded as direct; 75 percent from Facebook Messenger. https://sparktoro.com/blog/new-research-dark-social-falsely-attributes-significant-percentages-of-web-traffic-as-direct

‖‖

Bowen, Megan. "The Attribution Mirage." Refine Labs, October 15, 2024. Twelve-month comparison of 620 conversions: software credited web search with 78 percent, customers credited it with 12 percent and dark social with 85 percent; 98 percent of closed-won revenue self-attributed to dark social. https://www.refinelabs.com/blog/attribution-mirage

¶¶

Google Analytics Help. "URL Builders: Collect Campaign Data with Custom URLs." Definitions of utm_source, utm_medium, utm_campaign, utm_term and utm_content. https://support.google.com/analytics/answer/10917952

***

Wikipedia. "Time After Time (The Wire)." Season 3, episode 1, aired September 19, 2004, on HBO; the ComStat meeting, commanders watering down figures, and Colvin's "juke the stats" objection. https://en.wikipedia.org/wiki/Time_After_Time_(The_Wire)

†††

Oldroyd, James B., Kristina McElheran and David Elkington. "The Short Life of Online Sales Leads." Harvard Business Review, March 2011. Audit of 2,241 U.S. companies: 42-hour average response, 37 percent within an hour, 23 percent never; firms responding within an hour nearly seven times as likely to qualify the lead. https://hbr.org/2011/03/the-short-life-of-online-sales-leads