Brilliant Earth teardown: 800bps of margin, still a loss

Brilliant Earth (Nasdaq: BRLT) is the rare teardown where the gross-margin line is the good news and the operating line is the warning. Between FY2021 and FY2025 the company added roughly 800 basis points of gross margin, climbing from 49.3% to a 60.3% peak before easing to 57.5%. Over the exact same window, operating margin went the other way: from +10.6% of sales to a -1.2% operating loss, the first loss in its life as a public company. We read this 10-K the way diligence would, because the lesson is the oldest one in direct-to-consumer (DTC) retail: gross margin is not profit, and a brand that wins the margin line while losing the opex line is buying revenue it cannot keep.

The margin scissors

Start with the headline finding, because it sets up everything else. Gross margin and operating margin moved in opposite directions for four straight years. Gross margin improved nearly every year (49.3%, 53.3%, 57.6%, 60.3%, then 57.5% as higher gold and platinum costs bit in 2025). Operating margin fell every single year: 10.6%, 5.3%, 1.0%, 0.8%, and finally -1.2%.

When you chart the two together, they open like a pair of scissors. Each new basis point of gross margin got fully consumed before it reached operating income.

This is the pattern we see again and again when we talk to founders running brands at this scale. They obsess over the gross-margin line because it is the number they can move with sourcing, pricing and product mix, and it feels like progress. But the P&L below the gross-profit line is where the business is actually won or lost. A brand that takes gross margin from 50% to 60% and tells itself the unit economics are fixed has only fixed the top half of the equation.

The 2025 result made the gap concrete: a -$5.4M operating loss and a -$6.4M GAAP net loss, against +$4.0M of net income just one year earlier. Net loss attributable to BRLT was -$3.6M, or -$0.25 per share. Nothing dramatic happened to demand. The cost base simply caught up with a top line that had stopped growing.

Revenue stopped growing in 2022

Here is the structural fact underneath the loss. Net sales were $380.2M (2021), $439.9M (2022), $446.4M (2023), $422.2M (2024), and $437.5M (2025). That is a 3.6% four-year compound annual growth rate, and the 2025 figure still sits 2.0% below the 2023 peak. The business essentially stopped growing after 2022.

Flat revenue changes the math on everything below it. When you are compounding 25% a year, you can grow into a rising cost base and the ratios fix themselves. When revenue is flat, every dollar of new G&A, every new showroom, every public-company compliance cost lands directly on operating margin with no offsetting top-line growth to absorb it. There is no growth to dilute the cost.

Look at the operating-income column on its own and the story is brutal. Operating income fell from $40.1M to $23.3M to $4.5M to $3.4M to negative, while sales went sideways. That is not a demand problem. That is a cost-discipline problem against a top line that quit growing.

Where the gross-margin dollars went

So where did all that gross margin go? Into operating expense, in two buckets.

Marketing and advertising is the larger and the more efficient of the two. It ran $119.3M in 2023 (26.7% of sales), $108.3M in 2024 (25.7%), and $105.9M in 2025 (24.2%). Brilliant Earth actually got more disciplined here: marketing dollars fell 11% off the 2023 peak and improved about 250 basis points as a share of sales. Management even pointed to ~90 basis points of further marketing efficiency heading into 2026. Marketing is not the villain.

The villain is general and administrative cost. G&A grew from $133.2M in 2023 to $150.9M in 2025, up 13.3%, while sales were flat. Add it up and total operating expenses reached $256.9M in 2025, which is larger than the $251.5M of gross profit the company generated. That $5.4M gap is the operating loss, almost to the dollar.

Gross profit held essentially flat near $251-257M for three years while operating expenses kept climbing into it. In 2025 the two lines finally crossed. The operating loss is not a margin story or a demand story. It is the arithmetic of a fixed cost base growing into a flat top line.

When we work with operators at this revenue level, the G&A creep is the hardest thing to get them to confront, because almost every line of it feels justified in isolation. A bit more headcount to staff the new showrooms. A new system. The cost of being public. Each decision is defensible; the sum is an operating loss. The discipline that matters is not killing any single line, it is holding total opex growth below revenue growth, and when revenue growth is zero, that means holding opex flat in absolute dollars. Very few teams are wired to do that, and you can see in this P&L how expensive the gap becomes.

Orders up, ticket down: the lab-grown trap

The other pressure is on the top line itself, and it is specific to this category. Brilliant Earth sold more in 2025 and collected less per sale. Total orders rose 13.0% to 210,158, up from 186,030. But average order value (AOV, the average dollar value of each order) fell 8.2% to $2,082 from $2,269.

The cause sits upstream in the diamond market. A 1ct lab-grown diamond fell from roughly $3,400 in January 2020 to under $900 by late 2024, and toward production-cost levels (some wholesale stones near $80-200 per carat) through 2025. As the centre stone deflates, the natural price point of an engagement ring deflates with it, and the whole bridal category resets lower. Brilliant Earth is not mispricing; it is selling into a market where the anchor price is collapsing.

This is the volume-versus-value trap, and it is the part of the story most relevant to operators outside jewelry. The pattern we see again and again is a brand celebrating order growth while gross-profit dollars quietly erode, because a 13% lift in units against an 8% drop in ticket is a much thinner win than the unit count suggests. If your category is deflating, unit growth is not the scoreboard. Gross-profit dollars per order is the scoreboard, and you have to grow units fast enough to defend it, not just to look busy. When we have worked through this with founders whose AOV was sliding, the move that worked was re-mixing the assortment toward higher-ticket, higher-margin pieces rather than chasing volume at the deflating end of the range.

Debt-free, but liquidity tightened

There is real good news on the balance sheet, but it comes with a catch the headline cash number hides. The genuine positives: Brilliant Earth retired its entire term loan in 2025 (down from $55.7M the prior year) and ended the year with zero long-term debt, it generated positive free cash flow of about $5.8M even through the operating loss, and it initiated a dividend. A debt-free, FCF-positive business is not in distress.

But the cash position deteriorated sharply, largely because of those same moves. Cash fell roughly $83M to $79.1M as the company spent down its balance to repay debt and fund distributions. Inventory built up about $15M to $53.2M against $186.0M of cost of sales, around 3.5 inventory turns, slower than the year before. And the 10-K is explicit that the company ended 2025 with negative working capital of -$24.5M, with a current ratio of roughly 1.6x. So the honest read is debt-free and cash-generative, but with a tighter liquidity cushion than the prior year, not the cash fortress the cash line alone might suggest.

That is why the GAAP loss should not be read as distress. The company stayed positive on Adjusted EBITDA (its 17th consecutive positive quarter as of Q3 2025) and positive on free cash flow. The loss is driven by fixed and non-cash costs (depreciation, stock-based compensation, public-company overhead) sitting on top of a flat top line, not by the business burning cash. The debt-free, FCF-positive model still works. But with cash down to $79.1M and working capital negative, the balance sheet is no longer a free pass: the cost-discipline problem above the cash line now has less of a cushion beneath it than it did a year ago.

What operators should take from this

Strip the jewelry specifics away and Brilliant Earth is a clinic in three things every ecom operator should internalize.

First, gross margin is not profit. You can run a 57.5% gross margin and still lose money at the operating line if marketing and G&A eat more than the gross profit you generate. Track operating margin, not just contribution margin, and watch the gap between them.

Second, flat revenue makes opex discipline existential. Growth forgives a lot of cost sloppiness because the top line dilutes it. The moment growth stops, every new dollar of overhead lands on the bottom line at full weight. When we talk to founders whose revenue has gone sideways, the single most powerful move is freezing absolute opex, not trimming a percentage of it, until the top line moves again.

Third, defend gross-profit dollars, not units. If your category or your price point is deflating, unit growth can mask a shrinking business. Brilliant Earth grew orders 13% and still saw gross profit slip. Measure the dollars per order, and grow volume fast enough to defend them or re-mix toward higher-ticket product. This is exactly the kind of read an experienced finance partner brings, and it is the core of how we think about interim CFO work for brands at this stage.

For more category context, our jewelry financial benchmark sets the margin and AOV ranges this teardown should be read against, and the sibling 1stDibs teardown shows the same flat-revenue, cost-discipline tension in another luxury-adjacent DTC name.

Sources and methodology

The primary source for this teardown is SEC EDGAR, Brilliant Earth Group, Inc. (CIK 0001866757, ticker BRLT, Nasdaq, SIC 3910 Jewelry). The five-year P&L series is drawn from the consolidated statements of operations in the FY2021 through FY2025 Form 10-K filings, with the FY2025 10-K (accession 0001628280-26-018794, filed 2026-03-17, period ending 2025-12-31) as the anchor for the current year.

Income-statement figures (net sales, cost of sales, gross profit, marketing and advertising, general and administrative, total operating expenses, operating income, and net income) come from the rendered XBRL statement of operations in each filing. The marketing-versus-G&A split exists only from FY2023 forward; FY2021 and FY2022 reported a single combined selling, general and administrative line of $147.3M and $211.0M respectively, which is why those years show "n/a" in the marketing column of Table A.

Operating key performance indicators (total orders of 210,158, average order value of $2,082, and 42 showrooms as of 2025-12-31) are taken from the FY2025 10-K MD&A key operating metrics narrative. Balance-sheet and cash-flow figures (cash, inventory, long-term debt, operating cash flow, and capital expenditure) are read from the consolidated balance sheet and cash-flow statement in the same filings, taking the 2025-12-31 column rather than the prior-year column: FY2025 cash of $79.1M (the MD&A states $79.1 million excluding restricted cash), inventory of $53.2M, and the company's stated negative working capital of -$24.5M. Free cash flow is computed as operating cash flow less capital expenditure. Margins are computed from reported line items, and the revenue CAGR is calculated as (437.483/380.189)^(1/4)-1 = 3.6%.

Lab-grown diamond price context was triangulated from public trend reporting (Draco, Goodstone citing Edahn Golan, and The Knot data via Money), which place the 1ct lab-grown price down roughly 74% from January 2020 to December 2024 and approaching a production-cost floor through 2025. Storeleads channel data confirms brilliantearth.com runs on Shopify with a broad paid-media stack (Attentive, Klaviyo, Criteo, Global-e, plus Google, Bing, TikTok, Snap, Pinterest and Reddit pixels), consistent with the ~24% marketing-of-sales figure, though Storeleads "estimated sales" are modeled proxies and the 10-K is the only authority used for revenue. Operator interpretation throughout is anonymized and reflects patterns across founder conversations, with no client named.

Frequently asked questions