The Good, The Bad, and The Ugly: Reading Q2 GDP From Outer Space
Atlas Analytics' Weekly Subscriber Update
This past Wednesday, the Bureau of Economic Analysis released its much-anticipated first estimate of U.S. GDP growth for Q2 2025: +3.0% annualized.
At Atlas Analytics, we’ve been calling for a big quarter since May. On July 17, we put a precise number to it: +4.3%.
Clearly, the initial print came in below our forecast. But as we’ve said many times before, our ROY model is trained not on the first release but on the final number, the one that actually sticks.
That final Q2 GDP figure will be published on September 25, and we’ll reserve our formal concession (or victory lap) until then.
In the meantime, here’s our post-mortem on the advance estimate, structured in a format that might feel familiar to fans of spaghetti westerns: The Good, The Bad, and The Ugly.
The Good
At first glance, +4.3% may seem far from +3.0%. But dig into the components, and the picture gets a lot more interesting, and encouraging.
Atlas Analytics constructs its headline GDP call by modeling three core components:
Core GDP (Private Consumption, Government Expenditures, and Fixed Investment)
Private Inventories
Net Exports
Here’s how we stacked up against the BEA’s first release:
As the chart above shows, we were very close on Inventories and Trade, the two most volatile components. The shortfall came in Core GDP, the economic “heartbeat,” where we saw a late-quarter surge that likely wasn’t fully captured in the BEA’s advance read.
We expect that late surge, especially in June, will show up in future revisions, a point we’ll return to in a moment.
Also worth noting: market consensus for Q2 was around +2.3%. The Atlanta Fed’s GDPNow sat at just +2.4% as late as July 28. Relative to those benchmarks, Atlas Analytics wasn’t just early: we were directionally right and bold.
The Bad
We missed the first release number, and that’s always disappointing.
But here’s the key context: this number is not final.
According to the BEA’s own historical data, the mean average revision from the advance release to the third release is +0.7 percentage points, and from advance to final (after annual updates) is +1.2pp.
That means revisions of a full percentage point are not the exception. They’re the rule.
So when we forecast +4.3%, and the initial print comes in at +3.0%, we don’t panic. ROY isn’t trained to match a preliminary estimate. It’s trained to match reality.
And reality tends to reveal itself over time.
As a gentle reminder, in Q1 2025, Atlas forecasted GDP growth of +0.6%. The final number? -0.5%. Sure, a miss, but much closer than the legacy forecasters. The Atlanta Fed's GDPNow called for -2.7% while the New York Fed Nowcast projected +2.6%. In a quarter full of whiplash due to trade upheavals, we were the only model in the ballpark.
The Ugly
Now for the spiciest part of the post: model comparisons.
Let’s talk about the two highest-profile forecasting models: the Atlanta Fed’s GDPNow and the New York Fed’s Nowcast.
Here’s the headline comparison:
On July 17, the same date as our final public forecast, GDPNow was calling for +2.4%, while Nowcast was at +1.7% (as of July 25).
Then, on July 29, just one day before the BEA release, the Atlanta Fed’s GDPNow jumped to +2.9%, nearly a bullseye. Impressive? Maybe. But this is the third time in four quarters that GDPNow has converged with the BEA’s advance estimate immediately after the embargo window opens.
We’re not suggesting anything nefarious. But when a model suddenly aligns with the final number only after the official data pipeline opens, you have to ask: is this signal or confirmation?
At Atlas Analytics, we’re building something different.
No opaque adjustments at the 11th hour. No black-box recalibrations based on embargoed indicators.
Just data from space: satellites, land use, and economic activity signals. Consistent. Explainable. Updated weekly.
We’ll miss sometimes. But when we do, we’ll tell you why and show you how we got there.
That’s our promise.
Looking Ahead
We’ll be watching for the next two revisions on August 28 and September 25. Based on our satellite signal and internal validation, we’re still leaning toward an upward revision to Q2 GDP, especially in Core GDP.
We’ll publish another full breakdown after the second estimate.
Until then, keep watching the skies.
Want the Edge Before the First Print?
Atlas Analytics delivers real-time macro insights from outer space.
Subscribe for early GDP forecasts and economic nowcasts powered by satellite data.
Share this post if you know a portfolio manager, economist, or policy pro who needs sharper signals.
Inquire about enterprise access or custom data packages: atlasanalytics.com
We're building the Bloomberg Terminal of the stratosphere — and you're early.