Track record

Filed before the print.
Scored after it.

Before each Reddit earnings release, Oscylo files a quarterly forecast of revenue and reported users, with uncertainty bands, and freezes it. After the print, the same page scores it against the reported numbers. Once a forecast is filed, it stays up unchanged.

+2.03%Q2 2026 revenue error, while consensus ran 7–9% low
3 of 3banded forecasts landed inside their ±1σ intervals
Q3 2026next filing, before the late-October print

Q2 2026 · scored

The Q2 2026 scorecard.

Filed 2026-07-30 at 19:37 UTC, before Reddit’s release after that day’s market close. The headline call: revenue well above both guidance and Street consensus, carried by monetization per user.

Q2 2026 revenue, $ millionsWhere each estimate sat when Reddit printed $804.9M
$700M$750M$800M$850MActual $804.9MCompany guidance715 – 725Consensus (LSEG)730Consensus (Zacks)744.9Oscylo, filed 07-30821.2 ± 1σ$700M$760M$820M$880MActual $804.9MCompany guidance715 – 725Consensus (LSEG)730Consensus (Zacks)744.9Oscylo, filed 07-30821.2 ± 1σ

Guidance from Reddit’s Q1 2026 shareholder letter. Consensus figures are the pre-print analyst estimates compiled by LSEG (as reported by CNBC) and by Zacks. Actual from the Q2 2026 earnings release. The ±1σ band is one standard deviation of combined model uncertainty; see the note under the table.

Filed estimates against reported actuals, Q2 2026
MetricFiled estimate±1σ bandActualErrorIn bandConsensus (LSEG)Consensus error
Revenue ($M)821.2770.7 – 871.7804.9+2.03%yes730−9.3%
DAUq (M)125.82118.59 – 133.05130.3−3.44%yes130.3≈0.0%
WAUq (M)490.4462.5 – 518.2514.6−4.70%yes——
ARPU ($)6.53not banded6.18+5.66%n/a——
Actuals from Reddit’s Q2 2026 earnings press release (SEC-hosted exhibit). Revenue consensus per LSEG as reported by CNBC; Zacks’ revenue consensus stood at $744.9M (−7.5%). DAUq consensus as recorded in the forecast at filing time. Errors are estimate ÷ actual − 1. ±1σ is one standard deviation of combined model uncertainty: calibration-ratio variance, index dispersion, and DAU/WAU-ratio uncertainty added in quadrature. Under normal errors that is roughly a 68% interval; with two calibration quarters behind it at filing time, treat it as indicative rather than validated coverage.

Reading the quarter

What the print confirmed, and what we learned.

The revenue call came from modeling the drivers instead of the guidance.

Reddit’s management has consistently guided below what it reports, and ARPU has been rising on a steady, modelable trend. By modeling engagement, ad performance, and DAU directly, Oscylo filed $821.2M against a reported $804.9M. That is considerably closer than Street consensus ($730M LSEG, $744.9M Zacks) and the top of guidance ($725M).

Component errors are reported beside every headline, as they will be each quarter: the DAU estimate ran 3.44% low and the ARPU estimate 5.66% high, both inside the filed uncertainty.

The user call measured America, and that became the roadmap.

The Q2 panel was built predominantly from US-centric communities. It implied a slight sequential DAU decline (−0.77%); the print showed US DAUq falling −0.56%, the first sequential US decline Reddit has ever reported and within 0.2 points of the panel. Global DAUq rose +2.76% on international growth the panel wasn’t built to see.

Since Q2, Oscylo has established distinct US and international community groupings and will track the two segments separately alongside total DAU. The US segment carries ARPU, and nobody else is decomposing it.

Evidence chain

How “filed before the print” is established.

  1. Model run completes on a 231-subreddit panel (102,260 readings, 2023-01-01 → 2026-07-30). The forecast file is written and frozen: revenue $821.2M, DAUq 125.82M, WAUq 490.4M, ARPU $6.53, with the bands and the “what would make this wrong” list shown below.

  2. Reddit releases Q2 2026 results and hosts its earnings call at 4:30 p.m. ET. Reported: revenue $804.9M, DAUq 130.3M, WAUq 514.6M.

  3. The scorecard above is published with the forecast unedited. Every future quarter follows the same sequence, and every filed forecast stays on this page.

Filed with the forecast

Where we said this could go wrong.

These come straight from the caveats in the frozen Q2 filing. None were added after the print.

  1. The calibration was thin. Two reported quarters overlapped the index window at filing time, so the bands leaned on limited history. Every print since adds an anchor; the calibration set is already three quarters and grows with each release.

  2. ARPU has to be modeled. The revenue call carried the observed year-over-year ARPU trend forward (+44.1%). A mix shift toward lower-monetizing international users pulls revenue down without touching the user forecast. (ARPU printed 5.66% below the modeled figure; revenue stayed within 2.03%.)

  3. Panel composition could move the index on its own. At 71.1% collection, coverage changes in individual communities can drift the index. That is the failure mode hardest to see from inside, and the reason collection breadth keeps expanding.

Standing rules

The rules this page runs on.

Next filing

Q3 2026, before the late-October print.

The Q3 2026 forecast will be filed and publicly committed before Reddit reports, then scored here the week of the print. Private beta members receive the filing as soon as it is made, before Reddit reports.

Request private beta access