How Freelance Marketers Explain Results They Can't Control

Editorial Note: The individuals and companies in this article are composite characters, assembled from patterns in public reporting, industry research, and practitioner accounts. They are not real people, and no client relationship described here is real. Character names, ages, financial figures, client counts, timelines, and engagement details are illustrative composites. All external statistics, platform policy changes, and market data cited are real and sourced in the references at the end.

Attribution: How Four Millennial Freelancers Explain Results They Can't Control

The Question That Ends Every Quarter

10 a.m. on a Wednesday, screen-shared deck open, the growth lead at a pet-supplements brand asks the question a second time: "How do you explain this quarter's results?"

He's not being difficult. That afternoon he has to stand in front of his board and answer the same question, and the two sets of numbers on his desk flatly contradict each other: conversions as the ad platforms report them, versus revenue attributed on a first-touch basis in the order system — a gap of ten to twenty percent, year after year. He needs a version he can put into the board deck. Any version, so long as it holds.

That same week, the question was asked four times in four cities, in four phrasings, all pointing at the same thing. The people on the receiving end: Nora in Austin, Marcus in Chicago, Jules in Manchester, Dana in Philadelphia. All four are millennials, born between 1984 and 1995. All went independent between 2018 and 2022. All earn primarily through monthly or annual retainers, carry three to eight clients apiece, and work on payment terms of fifteen to forty-five days. They occupy four different links of the digital-marketing chain: search, paid media, social content, and email/CRM.

What makes them worth looking at together is how they file their quarterly reviews. Four documents, four formats — and exactly one column in common: "Attribution." Nora's holds three lines of technical explanation. Marcus's holds a unit-economics paragraph and a test result. Jules's is empty, annotated in the margin: "definitions pending alignment with client." Dana's holds a single number, followed by the test design and the statistical interval.

The answers to these four questions map, in order, onto four of the five positions on the attribution chain: the client, yourself, time, and the way you describe yourself to clients. The fifth — the environment — probably never gets a single answer. Which is why this piece doesn't have a conclusion. The four answers diverge widely, and each carries its own cost — and those costs say more about what this line of work actually looks like than the answers do.

When the Environment Moves: SEO, AI Search, and Who Explains It

Nora Vance, 38, Austin. She went independent in 2019, after four years running in-house SEO at a DTC skincare brand. By her account, she now carries six long-term clients on monthly retainers, with cash flow of $11,000 to $14,000 a month, net-30 terms, and a roster drawn from DTC and B2B software companies doing between $3 million and $20 million in annual revenue.

The client asking the question does about $9 million a year, and its organic traffic had fallen by roughly forty percent over two quarters. Nora opened Search Console: core rankings were essentially intact. What had dropped was impressions and click-through. On the results pages for her target queries, she found AI Overview modules sitting above what used to be the first organic result. She could confirm that both things were happening at once. She could not confirm that one caused the other.

She ran the technical audit first — the part of the job that actually consumes her weeks. Server logs for crawl frequency and crawl allocation. JavaScript rendering checks to see whether some pages were failing to index. Hreflang and canonicalization. Internal links rerouted to push equity from low-intent pages back to commercial ones. Then a keyword inventory re-sorted by intent into informational, comparison, and transactional tiers, with content briefs written for each tier. She ran all of it within two months. The traffic curve did not respond.

Then she did the harder thing. She laid controllability out in three columns: fully controllable (technical fixes, internal linking, content output), partly controllable (rankings, click-through rate), and not controllable (the shape of the SERP, competitor spend, platform crawl policy). At the renewal meeting she presented the columns before she presented the data. The client's reading was immediate: "So the only thing you can guarantee is column one." She said yes.

Then she changed three things. In contracts, deliverables and outcomes were separated: monthly deliverables listed explicitly (technical audits, content briefs, crawl-and-index fixes, internal-link adjustments), while "organic traffic growth" moved out of the deliverables column and into a shared observation metric, with the measurement definition and confidence interval spelled out where applicable. In pricing, she retired the per-keyword monthly package in favor of three tiers — a one-time diagnostic at $3,500, implementation at $2,800 a month, content assets priced per brief. And in reporting, she stopped importing any single institution's industry numbers into client documents: on the question of how AI search affects organic traffic, the estimates published by major research firms vary enormously in both magnitude and method, and writing one into a client report is how a piece of uncertainty becomes your promise.

While repricing, she worked from a set of public benchmarks. The Upwork Research Institute's Future Workforce Index, published April 2025, put the median income of skilled full-time freelancers in the U.S. at roughly $85,000, with 37 percent holding a graduate degree, versus 20 percent among full-time employees. Her conclusion from that data: the income spread inside this population cannot be explained by skill alone, and the way freelancers structure and price their work becomes part of the economic difference.

Q1 renewals: five of six clients renewed. One converted its monthly retainer into a one-time audit at a third of the fee. Her attribution column: "Impression decline coincided with SERP feature changes; causality unconfirmed. Deliverable definitions and billing structure adjusted."

The same week, Marcus Reyes was working a structurally similar problem. The difference: the system he faced didn't even offer the chance to explain.

Half the Data: Attribution Inside Platform Black Boxes

Marcus Reyes, 34, Chicago. Independent since 2021, after eight years at a performance agency, the last two running a team. Five clients, mostly VC-backed SaaS and DTC, with cash flow of $16,000 to $20,000 a month by his calculation — monthly fees plus a performance component based on contribution margin, never on revenue.

In April 2025, Google announced that Chrome would maintain its current approach to third-party cookie choice rather than introduce a new standalone prompt. On October 17, Google followed with an update to its Privacy Sandbox roadmap, announcing that several technologies — including Topics, Protected Audience, and Attribution Reporting — would be retired. The same day, the UK's Competition and Markets Authority closed its related case and released Google from its prior commitments.

Marcus had spent five years preparing for the "death of the cookie." He had migrated server-side tagging, configured Consent Mode v2, built first-party data pipelines for two clients, and paid out of pocket for a course on aggregated attribution. That morning he messaged his clients, to the effect of: everything we built for third-party cookie deprecation still applies to first-party data, but the aggregated-attribution track no longer needs budget — cut this quarter's technical spend accordingly. One client wrote back: "So what did all that money buy us?"

He didn't hand the question back to the platform. He started building his own evidence. Clients with budget got geo holdouts: selected states or regions held out as a control group for a period, then compared against the exposed regions. Clients without budget got time-sliced before-and-after comparisons, with seasonality and promotion variables controlled. He ran open-source marketing-mix models — Meridian and Robyn — on three years of history for two clients, and put the model output next to the platform dashboards in his monthly reports, with the discrepancy written up in plain terms. Page one of every monthly report was fixed: a section titled "What we can prove this period, and what we can't," three to five bullets, each annotated with method, sample, and confidence interval.

The question he put back to one client — "Do you want explainable, or do you want reproducible? Those two requirements are an order of magnitude apart in cost" — was later quoted by two clients in their own internal reporting.

Jules Okafor, 31, Manchester, working remote. Independent since 2022, after two and a half years doing social execution at a consumer brand. Eight small clients by her account, cash flow of £5,000 to £7,000 a month, mostly project work with two small retainers. Her black box is a different one from Marcus's: the platform gives her reach, engagement, and completion rates. It does not give her sales. Half her clients have no conversion tracking installed; the half that do have attribution windows that don't line up with anything.

Her day looks more like a production line than the job's public image suggests. Content calendars built a week out. Asset libraries organized by format — static, short-form video, customer testimonials, UGC reposts. UGC used only with signed releases. Comment response under an internal SLA: first reply within four hours on business days. Of her metrics, she trusts saves and shares most, because in her own client sample their correlation with downstream conversion has been steadier than likes. Reach she uses only as a diagnostic — a way to tell whether the distribution environment has shifted.

Same problem, two completely different toolkits. Marcus has budget, access, and history; he can manufacture causal evidence. Jules has none of those, so she redefined the deliverable instead: "content performance" is split into what she controls (publishing cadence, format mix, number of creative iterations, response time) and what she doesn't (distribution, algorithm changes), and the monthly report promises only the first list.

Her UK status imposes two specific constraints. The first is VAT: once turnover crosses the registration threshold, charging becomes mandatory at the standard rate, and her pricing had to absorb a visible jump; two price-sensitive clients didn't renew afterward. The second is contract architecture: every contract she signs carries substitution clauses, equipment provisions, and self-directed-schedule language, and she keeps her client list above three names — all of it aimed at staying clearly outside "providing services like an employee." Her take on the topic is strictly practical: it's a tax risk question.

Look at Q2 and Q3 separately. All five of Marcus's clients renewed; one had its media budget cut by 30 percent, driven by the client's internal financing schedule and unrelated to performance. Jules's client count went from eight to five; her per-client rate went up; total cash flow stayed roughly flat; her working hours dropped by about a third.

Marcus's attribution column: "Platform-reported attribution diverges from incrementality testing; divergence quantified. This period's conclusions based on incrementality testing."

Both of them spend their days explaining things to clients. Meanwhile, the clients have a boss to explain to as well.

The Client's Own Attribution Problem

Ray, 35, growth lead at a consumer brand, managing four outside contractors.

Ray's quarterly deck has an attribution section too. His problem is the same two sets of numbers that don't reconcile: conversions as reported by the ad platforms, versus revenue attributed on first touch in the order system — a gap of ten to twenty percent, persistent year after year. He has tried unifying definitions. He has tried MMM. He has tried making every contractor bring their own evidence. The result: every party can prove their own case. No party can prove anyone else's.

He replaced three vendors this year. The first walked over price at contract renewal. The second took on a category he didn't want his brand adjacent to. The third was a casualty of his own internal channel reprioritization — the channel got demoted wholesale. Not one of the three reasons was "they weren't good." The sentence he says most often: "I'm not trying to assign blame. I need a number that can go into a board deck."

Dana Whitfield, 42, Philadelphia. Independent since 2018, after seven years running CRM at a retail group with 4 million subscription records. Three clients, all annual contracts, cash flow of $22,000 to $26,000 a month by her account. Her work is email and lifecycle marketing: segmentation, automation, repeat purchase and winback, deliverability and domain health.

The technical side of the job is considerably more granular than "sending newsletters." RFM and behavioral tagging for segmentation. Welcome, browse-abandonment, post-purchase, and re-engagement flows built and maintained. Frequency caps so no user gets hit by three flows inside seven days. SPF, DKIM, and DMARC maintained, domain warm-up managed on schedule, invalid addresses scrubbed, dormant-list policies enforced. Order data synced in from the client's commerce system, because only then can email revenue be computed as revenue rather than as clicks.

Dana is the only one of the four whose current client setup gives her a relatively direct path to causal evidence. Email's data lives inside the client's own systems — send, open, click, and conversion all inside one pipe — so she can run A/B tests, and she can run holdout groups. When she explains the holdout to a client, she leads with the cost: the users in the control group receive fewer marketing touches for those ninety days, which means voluntarily giving up a slice of short-term revenue in exchange for an incrementality number that will stand up. That trade requires the client's signature, and she has watched two clients take two weeks to sign.

Her monthly report's attribution column is usually a single number, followed by test design and sample size.

The market conditions line up behind her position. Klaviyo's fiscal 2025 results, announced February 10, 2026, reported revenue of $1.234 billion, up 32 percent year over year, net revenue retention of 110 percent, and more than 193,000 customers. Dana cites those figures for one narrow purpose: the category has a large and expanding commercial ecosystem, while Klaviyo's reported net revenue retention indicates continued expansion among its existing customer base. For an independent consultant living on retainers, that makes the broader market context relevant. She doesn't forecast. She cites published results.

Ray's request — "a number that can go into a board deck" — is easiest to satisfy in Dana's world. One of her clients, somewhere in the second year of the engagement, watched the renewal-meeting question change from "how do you explain this quarter" to "can you take over winback next quarter too."

Her Q3 attribution column: "+18.4% incremental revenue (holdout control, n=42,000, 90-day interval)."

The person who can prove results gets a different question. At the next node, the blame swings back to themselves.

Scope, Rates, and the Parts That Actually Were Their Fault

Some of Jules's problems are ones she built herself.

One B2B SaaS client signed a contract reading "twelve LinkedIn posts per month plus comment response." By the third month, "just one more revision" had happened seven times. On revision seven, she spent four hours on the wording of a single post — a post that, once published, performed indistinguishably from the first draft. She calculated her effective hourly rate for that month. It came to £22.

She did something laborious and effective: for four straight weeks, she logged the difference between what clients asked for and what contracts said. After a month, the gap concentrated in three places — extra content beyond the contracted count, urgent schedule insertions, and off-hours comment response. She turned those three categories directly into clauses: a revision cap (two per deliverable, beyond that billed hourly), a materials responsibility split (if the client misses the agreed delivery window for assets, the schedule slips; if they insist it doesn't slip, the time counts as extra), and a response window (9 a.m. to 6 p.m. business days, no weekend commitment).

At the same time she narrowed her scope to two segments: B2B LinkedIn content operations, and UGC licensing and asset management. The narrowing cost her three clients. Two of the remaining clients raised their budgets. Her self-assessment is blunt: the first two years of low-rate work were her own choice — she needed cash flow and case studies — and the price anchored there. It took eighteen months to move it.

Nora keeps a similar account against herself. In 2024 she took an SEO engagement with an industrial-parts client — a category she didn't know, with long decision chains and scattered keyword intent. She took it for cash flow. Five months in, she delivered everything the contract specified; the client didn't renew, with the reason recorded as "no business-level momentum observed." Her own post-mortem: that client was lost on her category judgment, not on the search environment. Two hard conditions went into her intake criteria afterward — she must have served at least one comparable client in the category, or the client must agree to a paid diagnostic before any long-term commitment.

Both women's moves converged, in the same period, on the same act: saying no. Jules turns down vague-scope inquiries. Nora turns down unfamiliar categories and budgets below her threshold. Both report the same discovery — cash flow didn't fall after the refusals. It got less volatile.

Nora's Q2 attribution column, for the first time, contained a line about herself: "One client lost this quarter, attributable to category fit. Intake criteria updated."

Scope and pricing are variables you can change. The next one you can't: when the money arrives, and how much of it sits in one place.

Cash Flow, Concentration, and the Three-Year Question

Dana's largest client represents 60 percent of her revenue. In the fall of 2025, that client was acquired by a larger group — one with its own martech team and standing vendor contracts. The engagement terminated within 60 days. Final invoices were paid on time. No dispute.

In those 60 days she did three things. She staggered the renewal dates of her two remaining clients so no two contracts expire in the same month. She subcontracted two execution-level pieces of work — template development and some flow configuration — to a long-standing designer and an automation engineer, keeping strategy, test design, and data interpretation in her own hands. And she packaged seven years of methodology into a set of deliverable documents — segmentation frameworks, test-design templates, a deliverability troubleshooting checklist — as the onboarding asset for new clients. Her target is written down and specific: no single client above 40 percent of revenue.

Marcus runs a different number: if receipts went to zero for two consecutive months, how long could he last. The answer, by his calculation, is four months and two weeks. That number governs which engagements he accepts. A deal came in at three times his average monthly cash flow, but net-90 terms with a three-month test period demanded upfront. He took it — on the condition of a 30 percent deposit, and with the test-period success criteria written into the contract.

Around the same time, a former colleague of his went back in-house. The colleague had freelanced for three and a half years with a stable client base and income above his old salary. The reasons given were two specific things: cash-flow volatility made long-range planning impossible, and health insurance. He didn't say the industry was broken on his way out. What he said was: "I need something I can run the numbers on out to 2030."

Q4 renewals: her two remaining clients renewed. One switched from annual payment to quarterly prepayment; another added SMS budget. A new client brought her back to three active accounts. Her attribution column: "Client concentration reduced from 60% to 34%; incremental revenue +11.2% (holdout control)."

The Same Column — Your Turn

Four people answered the same column with four different kinds of answers, each carrying a stated cost.

Nora's answer: move the uncontrollable parts out of the promise. The cost — some clients conclude "you only guarantee the technical work," and she has lost bigger-budget opportunities to that reading.

Marcus's answer: manufacture your own evidence. The cost — time and money. A proper geo holdout burns weeks to months of a client's budget, and the answer at the end may simply be "no significant incrementality."

Jules's answer: redefine the deliverable. The cost — a smaller client count in the short term, and a standing obligation to explain to every new client why the contract caps revisions at two.

Dana's answer: hold the line with a reproducible number. The cost — she must first talk a client into accepting the short-term revenue loss of a holdout group, and that conversation alone takes two weeks.

Behind all four answers sits the same judgment: the fewer positions on the attribution chain you control, the more clearly you need to define the ones you do — and price them separately. That sentence isn't advice. It's the place all four arrived at after paying tuition in their own currencies.

Now for the part of the column that's still blank. The four questions below are the ones each of these four people was forced to answer in their first three years independent. There is no standard form for the answers.

The last client you lost — who did you attribute it to? If the client retold the same story from their side, would the account differ?
Does your contract separate deliverables from outcomes, or run them together? For the last contract where they ran together — when it broke, who was holding the pieces?
What share of your revenue does your largest client represent? If that client were acquired tomorrow, how many months does your cash flow hold?
When did you last raise your rates, and what reason did you give? If the reason was "I've gotten better at this" — what exactly is the client paying for?

The answers to these four questions map, in order, onto four of the five positions on the attribution chain: the client, yourself, time, and the boundary between what you promise and what you control. The fifth — the environment — probably never gets a single answer. Which is why this piece doesn't have a conclusion.

If you work in one of these four channels, you're welcome to write your own most recent attribution column in the comments. We'll compile answers by channel as source material for follow-up pieces.

References

Google. 2025. Next steps for Privacy Sandbox and tracking protections in Chrome. https://privacysandbox.google.com/blog/privacy-sandbox-next-steps
Google. 2025. Update on Plans for Privacy Sandbox Technologies. https://privacysandbox.google.com/blog/update-on-plans-for-privacy-sandbox-technologies
Competition and Markets Authority (UK). 2025. Investigation into Google’s ‘Privacy Sandbox’ browser changes. https://www.gov.uk/cma-cases/investigation-into-googles-privacy-sandbox-browser-changes
Klaviyo, Inc. 2026. Fourth Quarter and Full Year 2025 Results. https://investors.klaviyo.com/

Menu