General information only. Not financial advice or a recommendation to buy or sell any security. Data as of 17 Nov 2026.

Many marketers know Similarweb from its free tool: type in a website and it estimates how many people visit, where they come from and which rivals they also visit. On 17 November it reported its third quarter of 2026. Revenue growth picked up to 16% from 9% in the second quarter, it reported a GAAP profit per share, which it says is its first as a public company, and it says roughly two-thirds of the new recurring revenue it signed came from AI-related deals. The shares closed 13% higher on about four times their usual volume.

This is the first edition of Below the Ticker. The idea is simple: look past the share price at how a listed company actually earns its money, so builders and creators can borrow what works. Nothing here is a view on the shares.

What Similarweb does and who pays for it

Similarweb estimates traffic and behaviour across websites and apps, then packages those estimates as subscription software and raw data. Its customers fall into a few groups. Marketing and SEO teams use it to see which channels bring a competitor its visitors. Sales teams use it to find companies whose web traffic is growing. Investors use it as an early read on how an online business is doing before its results come out. Retailers use it to compare product and category demand. And, more and more, AI companies use its data as an input for their own products, through an API (a direct machine-to-machine data feed) or MCP, the protocol that lets an AI assistant query an outside data source.

The business is Israeli-incorporated, listed on the NYSE, and had 1,000 employees at the end of 2025, according to its annual report. Pricing starts low and climbs fast. Its pricing page lists Web Intelligence plans from $125 a month (as of 18 Nov 2026) for individuals and small teams, while the API, MCP access, data feeds and data licensing all go through its sales team on custom contracts. That split matters for everything below: a small self-serve front door, and a large enterprise business behind it.

Why it is in the news

The third-quarter release, also filed with the SEC as Exhibit 99.1 to a Form 6-K, reported revenue of $83.0 million, up 16% on the same quarter of 2025. GAAP earnings per share were $0.04, against a loss of $0.05 a year earlier. Non-GAAP operating profit, which strips out share-based pay and some other items, was $10.1 million, or 12% of revenue, against 6% a year before. Net revenue retention rose to 104% from 98%. Remaining performance obligations, the contracted revenue not yet recognised, rose 25% to $334.2 million. The company raised its full-year revenue guidance to $318 million to $320 million, the third increase this year.

The context makes the quarter more striking. On 17 February the company reported fourth-quarter 2025 revenue of $72.8 million, up 11%, and gave 2026 guidance of $305 million to $315 million. Then-CEO Or Offer blamed "longer sales cycles" on larger multi-year deals. Chartmill reported that revenue missed the roughly $78.0 million analysts expected and that the guidance sat below their $331.1 million estimate. The next day, 18 February, the shares fell 34.6% to $2.55 on about 12 million shares traded, according to stockanalysis.com price data. Nine months later the quarter those longer sales cycles were supposed to delay has arrived. There is also a new person in charge: Michael Akkerman became CEO on 2 November, succeeding founder Or Offer.

Line chart of Similarweb's daily closing share price from 17 Nov 2025 to 17 Nov 2026 with 50-day and 200-day moving averages

SMWB daily close with 50-day and 200-day moving averages. Source: stockanalysis.com, data as of 17 Nov 2026.

The read below the ticker

How it goes to market

Similarweb runs two motions at once. The self-serve plans pull in individuals, freelancers and small teams who find the free tool through search. The money is made further up. Customers paying $100,000 or more a year now account for 70% of annual recurring revenue (ARR), up from 63% a year ago, and there are 474 of them, 6% more than a year earlier. Two-thirds of ARR is now on multi-year contracts, up from 58%. On the earnings call, Offer said the AI deals are annual commitments and "most of them, if not all of them, are multi-year commitment," and that when customers start with the API or MCP, "usually the deals are bigger from day one" (Q3 call transcript).

Retention tells you whether that motion works. Net revenue retention compares what last year's customers pay today with what they paid a year ago, after upgrades, downgrades and cancellations. At 98%, the customer base shrank slightly on its own. At 104% it grows, and among the largest customers the figure is 114%, up from 105%. Growth is coming from customers expanding, which costs less than finding new ones.

The weaker spot is the middle. CFO Ran Vered said there were 1,790 customers paying more than $25,000 a year, against 1,828 a year earlier. The company is adding large accounts and losing a few mid-sized ones, and management's stated priority is moving more "six-figure customers to be a seven-figure customers."

Do the AI claims show up in the numbers?

Partly, and the part that shows is meaningful. Revenue growth went from 9% in the second quarter ($77.2 million, per Subscription Insider) to 16% in the third. ARR passed $300 million in June. The company says AI-related business drove about two-thirds of the ARR added in the quarter, and Offer said on the call that "almost 20% of Similarweb revenue is AI-related."

What the release does not give is a separate, audited AI revenue line, so "AI-related" is the company's own label and its definition is not published. The test that matters is the one already visible in the totals: if AI deals were only relabelled old deals, total growth and retention would not have moved. Both did. The open question is how durable AI-company budgets are, since that customer group is young and changes fast.

Early signals

Pricing: the self-serve entry price is published, while everything AI-facing (API, MCP, data licensing) is priced by contract. That keeps the large AI deals off the public price list. Hiring: Vered said "headcount is staying flat," so the margin gain came from revenue growing over a fixed cost base. Launches: the company released AI Ads in August, which tracks advertising inside ChatGPT and Google's AI answers, and announced a collaboration with NIQ in September on measuring "agentic commerce" (shopping done by AI agents), with a joint product due around December. Insiders: four directors, including Offer, made open-market purchases between 18 May and 1 June at $3.20 to $4.15 a share, according to their Form 4 filings.

The numbers

Measure

Value

As of / source

Share price

$8.84 (+13.0% on the day)

Close 17 Nov 2026, stockanalysis.com

Market capitalisation

$780.1 million

17 Nov 2026, stockanalysis.com

52-week range

$2.22 to $9.50 (intraday)

17 Nov 2026, stockanalysis.com

P/E (trailing)

Not meaningful (trailing 12-month GAAP loss of $14.0 million)

17 Nov 2026, stockanalysis.com

Price to sales (trailing)

2.54

17 Nov 2026, stockanalysis.com

Revenue, trailing 12 months

$306.8 million (+11.4%)

To 30 Sep 2026, stockanalysis.com

Revenue growth, latest quarter

+16% ($83.0 million)

Q3 2026, company release

Debt to equity

1.31 (the $43.2 million is lease liabilities; no borrowings on the balance sheet)

30 Sep 2026, stockanalysis.com and 6-K balance sheet

Cash and cash equivalents

$80.2 million

30 Sep 2026, company release

50-day moving average

$8.38 (price 5.5% above)

17 Nov 2026, computed from stockanalysis.com closes

200-day moving average

$5.67 (price 56% above)

17 Nov 2026, computed from stockanalysis.com closes

The debt-to-equity ratio looks high because shareholders' equity is small ($32.8 million) after years of losses, and the "debt" in that ratio is made up of office lease obligations; there are no loans. And the moving averages show the shape of the year in the chart above: the price fell below both averages after February, crossed back above the 200-day line in late June, and has stayed above it since.

Ownership

Holder

Stake

As of / source

MIH E-Commerce Holdings (Prosus and Naspers)

11.17 million shares, 12.8%

14 Feb 2026, 2025 Form 20-F

Viola Group (venture fund; partner Harel Beit-On is a director)

10.81 million shares, 12.4%

14 Feb 2026, 2025 Form 20-F

Anglo-Peacock Nominees (for Joshua Alliance)

8.88 million shares, 10.2%

14 Feb 2026, 2025 Form 20-F

Menora Mivtachim Holdings

4.02 million shares, about 4.6%

13F for 31 Mar 2026, filed 14 May 2026

Wellington Management

2.97 million shares, about 3.4% (down 1.2% in the quarter)

13F for 31 Mar 2026, filed 15 May 2026

Directors and executive officers (10 people)

29.4%

14 Feb 2026, 2025 Form 20-F

Insiders / institutions (data-provider estimate)

26.7% / 39.7%

17 Nov 2026, stockanalysis.com

A note on dates. 13F filings, the quarterly reports large US investment managers file, arrive up to 45 days after the quarter they describe, and the most recent quarter in the dataset used here is 31 March 2026. Those positions are seven and a half months old and predate the rally. In that quarter, according to 13F data compiled by Fiscal.ai, several funds increased small positions by more than 100%: S Squared Technology (to 1.6 million shares), Millennium Management and Renaissance Technologies.

Similarweb is a foreign private issuer, and until this year its insiders did not file Form 4 trade reports. Its annual report says that under the Holding Foreign Insiders Accountable Act, its directors and officers have had to file Forms 3, 4 and 5 since 18 March 2026, while large shareholders who are not directors or officers remain exempt. The SEC confirms the Act took effect on 18 March 2026 for directors and officers of foreign private issuers. The Form 4 filings on EDGAR show both directions. Besides the May purchases, Offer disposed of about 1.76 million shares between 13 August and 24 September at weighted average prices of $7.69 to $9.01, worth about $15.2 million at those prices, around 0.95 million of them from exercised options, and reported 3,930,481 shares afterwards. The filings for the September disposals state they were made under a Rule 10b5-1 trading plan adopted on 1 June 2026; the filings give no other reason.

The case others make

Analysts who cover the company read the quarter as evidence that the AI demand is real and profitable. Investing.com reported that Oppenheimer pointed to "continued demand for AI-related data and solutions," the 114% retention among large customers, record normalised free cash flow and positive GAAP net income. Northland, per the same outlet, said the improvement "reflects higher value and higher margin deals" with "continued momentum with AI and enterprise deals." Management's own case, in the release, is that contracted backlog growing 25% (faster than revenue) gave it the confidence to raise guidance.

Main risks

  • Lumpy enterprise deals. February showed what happens when large contracts slip: one quarter of slower growth and a 34.6% fall in a day. A business built on fewer, bigger contracts has fewer, bigger surprises.

  • A young customer group. AI companies now drive most new ARR. Their budgets, data needs and willingness to license third-party data could change quickly, and the company does not publish how it defines "AI-related."

  • Data access. The company's own risk language cites "our reliance on certain third party platforms and sources for the collection of data." Its estimates depend on data it does not fully control.

  • A shrinking middle. Customers above $25,000 fell to 1,790 from 1,828, so growth leans on the largest accounts.

  • Leadership change and dilution. A new CEO started two weeks before the results, and shares outstanding rose 4.8% in a year.

  • Thin balance sheet cushion. Equity is $32.8 million and working capital is negative ($43.3 million short, per stockanalysis.com), which is common for subscription companies that bill in advance and still worth knowing. The shekel's strength against the dollar is a cost headwind, Vered said, since about half of the company's employees are in Israel.

  • Volatility and concentrated ownership. Directors, officers and three large holders own a big share of the stock, and the free float is about 42.7 million shares, so the price can move sharply on modest volume.

One lesson for builders and creators

Similarweb did not invent a new product for the AI boom. It took the asset it already had, years of web traffic estimates, and offered access to it in the form the new customer wanted: an API and an MCP server alongside the dashboard. According to Offer, deals that start this way are bigger from the first day. The new customer was software reading data, and the packaging followed the customer.

You may have the same kind of asset. A creator with five years of tested prompts, a developer with a cleaned dataset, a newsletter with a structured archive, a consultant with a library of templates: each can be packaged for an AI agent as well as for a human. If you are already thinking about how agents discover tools, our piece on listing your tool where AI agents shop goes deeper, and our note on pricing by outcome instead of seats covers how to charge when the user is not a person. A practical way to start:

  1. List what you own that is hard to rebuild. Data, archives, tested workflows, labelled examples. If someone could recreate it in a weekend, skip it.

  2. Ask who would read it by machine. An AI assistant, an agent, another developer's app. Write down one named customer type.

  3. Ship the thinnest machine-readable version. A JSON file, a simple API endpoint or an MCP server with one or two tools. Keep the human dashboard as it is.

  4. Price it separately and by contract at first. Similarweb keeps its AI-facing prices off the public page. Talk to the first five customers before you publish a number.

  5. Track retention from day one. The figure that changed Similarweb's year was retention moving from 98% to 104%. Measure what each customer pays you this month against a year ago, even if the answer is small. If you want a simple place to keep revenue and traffic side by side, Raah.dev puts both in one dashboard.

The decision rule: if a second kind of customer can use your existing asset with a change of format, try the format before you build anything new.

General information only. Not financial advice or a recommendation to buy or sell any security. Data as of 17 Nov 2026.

Sources