From losing $2.9M to cash positive: a turnaround in three acts.

Act IThe call

I first worked with this brand in 2022, when they were small and hiring a financial modeler on Upwork. I built their model, we finished the project, and that model quietly ran the company's planning for the next two years. The brand grew and got acquired. Eventually the model broke, which is what models do once a business outgrows the assumptions underneath them.

When the founder came back to me in 2024, everything was bigger, including the problems. The company was doing $7.69M in revenue and losing $2.88M a year doing it, a negative 37% EBIT margin. There was a board now, with questions he struggled to answer. And what struck me most in those early conversations wasn't in the numbers at all. It was his mood. He'd been carrying the story that the company might be failing for so long that he'd started to believe it.

Act I · continuedWhat the data actually said

Before forming any opinion I went into the data. Raw transaction exports, cohort tables, subscription behavior, acquisition costs by channel, fulfillment invoices. The underlying records, not the dashboards.

The picture that emerged did not match the mood in the room.

Underneath the ugly EBIT line, the machinery was working. Cohorts were retaining. Subscriber economics were sound. On corrected projections, the company was on a clear path to cash positive, provided it kept its nerve and managed to the right numbers. So my first deliverable was strange for a turnaround: an argument, backed by data, that the doom was wrong.

The same dig produced a smaller finding that shows why I insist on raw data. The 3PL's invoices were structured in a way that understated international fulfillment costs. Everyone believed shipping an order to Canada cost under $40. The real figure was closer to $70, and Canada wasn't the only country affected. Nobody had lied and nothing was broken. The cost was simply invisible in how the paperwork flowed, which meant international unit economics, and every decision built on them, were fiction.

Act IIThe reframe

The deeper problem was the lens. The brand sells a physical product online, so everyone measured it like an e-commerce company: revenue, ROAS, blended margins, month over month. But most of its revenue came from subscribers. Structurally this was a subscription business, a SaaS company wearing an e-commerce costume, and subscription businesses are governed by different numbers: cohort retention, lifetime value, payback periods.

We rebuilt the company's financial lens around that reality. New KPIs tracked cohort behavior and retention instead of just monthly revenue. A blended LTV-to-CAC framework changed how the media budget was deployed, because once you know what a customer is truly worth over their life, you know what you can afford to pay for one, channel by channel. CAC came down roughly 25%. And retention became strategy rather than an afterthought: the loyalty program and the money-back guarantee both came straight out of the cohort math.

Numbers don't defend themselves, though. The founder had a board expecting answers, so we built the story together and I walked into the board presentation beside him, with corrected projections and the reasoning under every line. He stopped playing defense with numbers he didn't trust and started making a case he could argue himself. A founder who can defend his own projections is a different founder.

Act III2025: the year the dashboards lied

Then the ad platform changed the rules.

New restrictions on health-related advertisers landed in 2025, and this brand's customers sit squarely in that category. The platform stopped reporting conversion metrics for them, targeting degraded, and attribution broke. The brand's analytics stack began reporting a paid-social CAC of $900, three times the historical $300. Taken at face value, the data said the channel had collapsed and the company should pull out entirely.

I didn't take it at face value. Post-purchase survey data told a different story: customers were still finding the brand through that channel. My read was that the channel was wounded but working, and that attribution was what had actually broken. Blended CAC, measured the attribution-proof way as total spend divided by new customers, averaged $472 through the restriction period. Painful, but a completely different decision than $900.

Acting on a hunch is cheap talk, so we tested it. I argued for bringing in a specialized MMM agency to measure the channel's true CAC without relying on broken attribution. In the meantime we cut ad spend rather than pour money into a channel we couldn't measure, replaced a media buyer who wasn't performing, and I backed the founder's push to scale Amazon as an alternative channel.

In Q3 2025 the restrictions lifted. The changes we'd made started compounding immediately. Average CAC fell to $301 in the testing window and kept improving into the low $200s, better than before the crisis, and we scaled spend back up into a machine that now worked properly.

The books tell the year plainly. Revenue dipped 11% to $6.82M because we deliberately pulled spend rather than pour money into a channel we couldn't measure. And the EBIT loss shrank from $2.88M to $0.73M. Three quarters of the loss, cut in the middle of a measurement crisis.

Where it standsThe scoreboard

For 2026 we set the plan the previous two years had been building toward: $10.11M in revenue and $1.06M in EBIT, cash positive at a 10% margin. Six months in, the company sits at 95% of the revenue target and is beating the profit target, because CAC keeps coming in under plan. Two years earlier it lost $2.88M and its founder was rehearsing the failure story.

The strategy barely changed through any of this. What changed was the quality of the numbers it rested on, and the discipline to hold course when the dashboards were screaming nonsense. My current argument with the company is one you don't expect from a finance person: six months into 2026 they've hit 95% of their revenue target while underspending marketing, and I'm pushing them to spend more. The cohort math says every correctly priced customer acquired today compounds for years. Most CFOs exist to say no. The job is knowing when the numbers say yes, and saying it just as loudly.

The takeawayWhat this story is about

Nothing here came from cost cutting or growth hacking. The turnaround came from replacing numbers that had drifted away from reality with numbers that hadn't, then managing to them through a board meeting, a platform crisis, and a season when every dashboard in the stack was wrong.

Every engagement starts the way this one did: in the raw data, with the question nobody was asking. Are the numbers we're steering by actually true?