Reddit's Ad Revenue Grew 64%. The Stock Fell 21% on 300,000 Users.

Reddit reported Thursday afternoon. Revenue of $805 million, up 61% year over year. Advertising revenue of $762 million, up 64%. Net income of $253 million against $89 million in the year-ago quarter — 31% of revenue. Eighth consecutive quarter of growth above 60%, and third-quarter guidance of $860 to $870 million.

The stock closed Friday down 21%.

The number that did it was United States daily active uniques: 53.2 million, against 53.5 million in the first quarter. Three hundred thousand people. It was the first sequential decline in the five quarters Reddit has charted, and the figure is still up 6% year over year. Management attributed it to choppy search referrals — Google’s AI Overviews answering the query before the click, compressing logged-out traffic.

Every number the revenue organization owns went up. The stock fell on the one it doesn’t.

1. Career Snapshot

Sales Ops and RevOps at the ad platforms is where the movement is this week, so that is where this issue starts. Two prints landed inside 24 hours and they rhyme.

Meta reported Wednesday. Total revenue of $60.8 billion, up 28%. Advertising revenue of $59.36 billion, up 27%, with impressions up 14% and average price per ad up 12% — the ad engine firing on both volume and yield, which is the rare quarter where you cannot argue one is propping up the other. Earnings per share came in at $6.18 against a $7.14 consensus. Full-year capital expenditure guidance moved up to $135–145 billion, with $31.1 billion spent in the quarter, consuming very nearly all operating cash flow. The stock fell about 8% after hours.

Same shape as Reddit. The revenue organization delivered, and the market repriced the company on an input the revenue organization does not control — traffic supply in one case, compute cost in the other.

This is the thing worth internalizing if you are aiming at Head of Revenue Operations, Director of Sales Strategy & Operations, or Head of Analytics at any of these platforms. The job has quietly stopped being report the number and become defend the number against the one next to it. Reddit’s revenue chief cannot walk into the next board meeting with 64% growth and expect the room to be satisfied. The room already knows. The room wants to know what happens to that curve when the top of the funnel stops widening.

The ARPU gap is the actual mandate. Reddit’s US average revenue per daily unique is $11.85. International is $2.26 — 19% of the US rate. Global daily uniques grew 18% to 130.3 million and weekly uniques crossed 500 million for the first time, so the users are arriving. They are arriving in markets that monetize at a fifth of the rate. eMarketer’s read is consistent: international is roughly half the audience and about 17% of revenue.

Read that as a job description rather than a statistic. When US user growth flattens and international user growth does not, every dollar of the next two years has to come from yield — pricing, packaging, advertiser onboarding, sales coverage design, self-serve conversion in markets with no sales team in them. That is the revenue operations function, described exactly. Reddit is already staffing against it, with mid-market client partner roles posted for the UK and an EMEA team run out of Amsterdam.

Still reporting next week: Snap on August 3, guided to $1.52–1.55 billion. Pinterest on August 4, guided to $1.133–1.153 billion against $1,008 million and 631 million monthly actives in the first quarter. If either shows the same divergence — monetization up, engagement flat, multiple compressed — the pattern stops being an anecdote about Reddit and starts being the thesis for the category.

Comp, with a caveat attached. Analytics engineering base bands read $115K–$175K for 2026, with dbt and a cloud warehouse treated as non-negotiable and a four-to-seven-week search to fill a seat well. Director of Analytics is where the aggregators fall apart: Payscale around $152K, Indeed around $162K, Glassdoor around $223K, ZipRecruiter spanning $112K at the 25th percentile to $243K at the 90th. Senior Director figures circulating near $429K total compensation rest on a sample of sixteen profiles, which is not a market — it is a rumor with a decimal point. Treat all of it as directional and check levels.fyi before you quote a number at anyone.

The spread itself is the finding. A title with a $110K interquartile range is a title the market has not defined, and undefined titles are where the negotiating room lives.

2. Tech & Market Trends

The semantic spec became a foundation project. Open Semantic Interchange — the YAML interchange format for datasets, metrics, dimensions, and relationships, launched by Snowflake with dbt Labs, Salesforce, BlackRock, and RelationalAI — shipped v1.0 to GitHub in January under Apache 2.0 and was donated to the Apache Software Foundation, entering the incubator as Apache Ossie in June. Snowflake, dbt, Databricks, and Microsoft Fabric are all behind it.

Why it matters: a vendor consortium is a marketing exercise. A foundation project with a governance model is an actual standard, and the difference decides whether your metric definitions are portable or hostage. For interviews, the useful framing is that OSI is an interchange format rather than a semantic layer product — it is how two layers talk, not a third layer. Say that correctly and you sound like someone who read the spec instead of the press release.

Databricks is raising at a $188 billion valuation. The July 16 round, led by existing investor Coatue, points capital at Unity AI Gateway, Genie, and Lakebase.

Why it matters: follow where the money is aimed. All three are governance-and-context plumbing rather than modeling tools. The platform vendors have concluded the scarce thing is a trustworthy layer between the warehouse and the agent. They are correct, and they are building the pipe rather than the judgment about what flows through it.

Power BI’s July release is a governance release wearing a features hat. TMDL View on the web, Model Options in the Service, and new Fabric REST APIs for org apps, audiences, and paginated reports, with semantic model settings moving to a pane in August.

Why it matters: TMDL on the web plus REST APIs for app and audience management means the Power BI estate is becoming something you administer as code and diff in a pull request. If your Fabric experience is entirely click-path, that is a gap with a name now. A Director candidate who can describe a reviewable, source-controlled semantic model deployment is arguing at a different altitude than one who can describe a workspace.

The analyst seat got more strategic, not less necessary. The consistent read across market surveys is that the 2026 analyst spends less time on query syntax than the 2022 analyst and more on scoping, framing, and stakeholder translation.

Why it matters: this is the most abused sentence in data career writing, so be precise about it in interviews. It does not mean SQL stopped mattering. It means SQL stopped differentiating. The thing being priced is the judgment about which question to answer, and nobody has found a way to put that on a résumé other than showing the work.

3. Next Action Plan

The portfolio move: build the monetization gap model, on Reddit’s public numbers.

Everything you need is in the shareholder letter and the 8-K. You are going to build the analysis a Head of Revenue Operations would present at the next planning cycle, using a company that just made the problem legible for you at no cost.

• The shape. One model, one page of commentary.
• The time budget. Three hours.
• The artifact. Decompose revenue into daily uniques × ARPU, split US and international. Hold US uniques flat — the quarter says that is now the honest base case — and grow international at the trailing rate. Then answer one question: what does international ARPU have to reach for total revenue growth to stay above 40% through 2027? Compute the required rate. Compare it to the $2.26 you are starting from and the $11.85 ceiling the US suggests. Write down what has to be true operationally for that number to happen — sales coverage in which markets, self-serve conversion at what rate, which advertiser segment — and what breaks if search referrals stay choppy.
• What success looks like. You end up holding a defensible view on whether a public company’s growth story survives its own user curve, expressed as an operating plan rather than an opinion. That is a first-round conversation you can steer, and it works whether you are interviewing at Reddit, at one of its competitors, or at a company that has never heard of any of this and has the identical structural problem in a different market.

The reason this beats a dashboard is that a dashboard shows what happened. This one commits to a number and states its conditions. A revenue organization that can only explain the revenue is a reporting function with a quota attached.

The outreach move. Post the numerator-denominator observation, short, without a lesson bolted onto the end. Reddit grew advertising revenue 64% and lost 21% of its market value on a 300,000-user sequential dip that is still positive year over year. One paragraph on what that implies about which metrics a revenue team is now accountable for, and one honest question: when the market reprices you on an input you don’t control, what does the operating review actually change? People who run these functions will answer that, because it is the thing keeping them up. Do not append “thoughts?” and do not explain the joke.

For a warmer version, take it to a specific person — someone in RevOps or analytics leadership at a platform reporting next week — and ask what they are watching in the Snap and Pinterest prints. You are asking about their week, three days before their week happens. That is a better opener than anything with the word “connect” in it.

The 7-day micro-skill: write one metric in OSI-shaped YAML.

Not the whole layer. One metric — the ARPU definition from the model above. Express it in the Open Semantic Interchange format: the dataset, the dimensions it slices by, the relationships it depends on, the context it needs to be read correctly. Then write the same metric in dbt’s MetricFlow spec and in a Snowflake semantic view, and note where they disagree.

Seven days gets you something most senior candidates cannot do: describe, from having done it, what actually survives translation between semantic layers and what quietly does not. Given that four competing vendors just agreed to a shared format precisely because that translation keeps failing, this is a small investment sitting directly on top of where the platforms are spending.

---

One closing calibration. The $135–145 billion Meta is spending on capital expenditure this year is roughly 180 times Reddit’s entire quarterly revenue. Both companies were punished this week regardless. Scale is not the protection it is assumed to be, and a market that repriced the largest advertising business on earth over a cost line will not be gentler with your quarterly business review. Know which of your numbers you control. Be able to say the sentence out loud before someone asks you for it.