An Atlas of Ignition Points: Where the Blitz Actually Caught Fire

Fourteen winners, one fake — a descriptive atlas of decisive windows

Every company we score is, in some sense, mid-blitz. But the winners all share a moment — usually a one-to-three-year window — when the business caught fire. We built an atlas of those windows for the fourteen companies in our 65+ Club, plus the most important counter-example in our coverage: WeWork.

The ground rules are the same as always: figures come from SEC filings and XBRL, score trajectories from our estimated history, and every claim below is a description of the past, not a forecast of the future. Where the data cannot support a verdict, we say so.

The question

Two research questions converged into this atlas. First: among our highest-scored companies, where — in which window — did the blitz actually ignite? Second: can the decisive windows be typed, so that we recognize the shapes when they appear? The answer turned out to be yes for the first and partially for the second.

Here are the verdicts, company by company.

Tour one: the demand explosion

NVIDIA, 2023–24. Revenue went from $26.97B to $60.92B (+125.9%) to $130.50B (+114.2%). Gross margin rose from 56.9% to 75.0% — an 18.1-point expansion — while operating margin went from 15.7% to 62.4%. What makes this window singular in the atlas is the margin story: the explosion strengthened the margin structure instead of diluting it. Our N5 test found that margin dilution during scale-up is the normal state of successful blitzscalers — NVIDIA skipped the dilution. Data-center revenue overtaking gaming adds a second-curve element to the same window.

DoorDash, 2020–21 and Snowflake, 2019–21 are the purer specimens of the type. DoorDash revenue grew +226.1% then +69.4% while operating margin climbed from −69.6% to −9.2%. Snowflake compounded +173.7%, +123.7%, +106.0% while gross margin rose from 46.4% to 62.4%. Neither company had a platform change to explain it — demand simply arrived, and the cost structure absorbed it.

Tour two: the platform transition

AppLovin, 2023–24 is the most dramatic margin conversion in the atlas. Revenue grew +75.1% while gross margin leapt from 55.4% to 80.6% and operating margin from −1.7% to 41.9% — a 61-point swing. The mechanism is legible: the AXON 2.0 AI advertising engine, plus the divestiture of the game studios, turned the company into a pure advertising platform. The explosion was not more of the same business. It was a different business, proven by the margin structure.

Palantir is the same shape, still in progress. U.S. commercial revenue — the AIP story — grew +149%, total revenue accelerated from +28.8% to +56.1%, and operating margin expanded from 5.4% to 31.6%. MongoDB is the completed, quieter version: revenue grew +60.9% and +57.9% while Atlas, the consumption-based database-as-a-service, rose from 4% to 38% of total revenue. In all three, the explosion is inseparable from a margin-structure conversion.

Tour three: the second curve, and the fake

Uber, 2022–23. Revenue grew +82.7% then +17.0%, but the number that defines the window is the margin: operating margin went from −5.7% to +3.0%, the company's first annual GAAP operating profit. Delivery had become countable ($17.25B), and advertising — a $2.5B annualized second curve — was emerging. This is what a completed blitz looks like at the moment it becomes a harvest.

WeWork, 2016–18. By the numbers alone, this was the finest explosion in the atlas: revenue $436M → $886M (+103%) → $1,822M (+106%), triple-digit growth three years running. But operating margin in 2018 was −93% — the company spent roughly $1.90 for every $1 of revenue — and lease obligations stood at $17.9B, nearly ten times revenue, with maturity mismatches baked in. We scored it blind at 29/D. The discriminator between real and fake explosions is not the growth rate. It is the direction of the margin plus the structure of the commitments.

Signature one: acceleration, plus the margin's verdict

In 12 of the 13 determinable windows, revenue acceleration arrived together with margin expansion. NVIDIA (+18.1pp GM), AppLovin (+28.5pp GM, +61pp OM), Airbnb (OM −106.3 to +21.5), Uber (first profit), MercadoLibre (turned profitable), Arista (OM +7pp with GM stable), Palantir (OM +26pp), DoorDash (OM +60pp), Snowflake (GM +16pp).

The single exception is MercadoLibre: gross margin fell from 48.0% to 42.5% during its explosion, and the explosion sustained anyway. The margin fell because the company chose to invest — logistics, credit — not because unit economics broke. So the precise statement is this: acceleration defines the explosion; the margin's direction judges whether it is real. WeWork's margins collapsed and it died. MELI's margins diluted on purpose and it thrived.

Signature two: explosions arrive in clusters

The decisive windows are not scattered randomly through time. They cluster into three eras: the COVID demand explosion of 2020–21 (Shopify, DoorDash, MercadoLibre, Airbnb, Alphabet), the efficiency turnarounds of 2022–23 (Meta, Uber — cost discipline and leverage), and the AI capex cycle of 2023–25 (NVIDIA, Palantir, Arista). Each company's window is the intersection of a macro trigger and company readiness. The atlas does not claim these windows are reproducible on demand — it records that prepared companies exploded when the trigger arrived.

Signature three: platform transitions are proven by margins

All three platform-transition cases — AppLovin (AXON), Palantir (AIP), MongoDB (Atlas) — are accompanied by a qualitative margin-structure conversion: AppLovin GM +28.5pp, Palantir OM +26pp, MongoDB's Atlas share from 4% to 38%. We are registering this as a hypothesis candidate (N11): platform-transition explosions are accompanied by margin-structure conversion of GM +15pp or more. It is an observation now, a test later.

The honest one: Salesforce graduated

Salesforce refuses a verdict, and we refused to force one. Its explosion — the cloud-SaaS replacement of on-premise CRM — ran from 1999 to 2016, outside our financials and estimated-score range. What we can see, 2016 to 2026, is a graduation chart: the score drifted from 71 to 65 while speed fell and the moat rose. The first blitz is finished; the company is in harvest, with a possible second blitz (Agentforce) being prepared. Where the data ends, the verdict stops.

A related honesty note: in 7 of the 9 companies with estimated score histories, the financial explosion overlaps a rising knee in the score — NVIDIA 79→87, MercadoLibre 63→71, Arista 61→65 (crossing the line), AppLovin 68→73, Palantir 66→74, MongoDB 56→61, Shopify 75→79. This reads as "the score reacted during the explosion," not "the score predicted it" — every estimated history carries hindsight bias, a lesson our H3 test forced us to internalize.

The boundary: a business explosion is not a price explosion

This atlas maps the trajectory of businesses. Our own tests showed that business-ignition timing is not price timing: after 65-point crossings, the one-year median return was −5.9% (H1, rejected); score momentum's correlation with forward returns was −0.091 (H3, rejected). Entering at the explosion window usually means buying the explosion after it is priced in.

So do not read this atlas as "if you had bought then." It is a descriptive study with no verified forward predictive power. What it does support is narrower: the type of an explosion and its margin structure are legitimate evidence for the model's moat and creative-destruction components — and margin direction remains our sharpest real-vs-fake discriminator.

What comes next

The 33 companies outside the 65+ Club are the natural second volume — especially the 60–64 borderline, where unexploded and in-progress cases will enrich the control group. As quarterly re-scores accumulate, the in-progress verdicts (Palantir, AppLovin's second curve) will be updated. And N11 — the platform-transition margin signature — will be registered for forward testing.

BlitzLab scores businesses against a scholar's standard — not investment advice. No price targets, no buy/sell recommendations. Scores cited: 65+ Club as of 2026Q4, model v1.0 frozen. Financial figures are from SEC filings (XBRL) as labeled; the estimated score history carries hindsight bias, disclosed in the atlas. Past patterns are reported as technical statistics, not as signals.