ServiceTitan, Inc. [TTAN]

The Tier-1 screen recorded TTAN at a $2.775bn market cap on 35,397,085 shares. That share count is a PRE-IPO cover-page figure last tagged in the 10-Q filed 2025-01-14, six weeks before the December 2024 IPO. The verified count from the 10-Q cover is 95,397,579 (82,746,425 Class A + 12,651,154 Class B) - 2.695x higher. Market cap is $7.479bn, not $2.775bn; EV/Sales is 6.96x, not 2.42x. Net cash is $421.5m, not $317.5m, because the scanner subtracted a $104.0m term loan that was voluntarily repaid in full in January 2026. Re-run on the screen's OWN assumptions with the corrected share count, the required 5-year CAGR is 33.6% and the margin is -9.5pp - a FAIL. The screen's +18.9pp margin was produced entirely by the share-count error. On my own bottom-up terminal margin of 20% (built from a disclosed 78.7% platform gross margin, versus the screen's generic 11.2% placeholder) and a growth-matched exit multiple of 22.7x, the price requires a 19.8% CAGR against 24.1% demonstrated: PASS, margin +4.3pp, falling to +2.1pp at a 12% WACC. The 12-month multiple anchor is UNIDENTIFIED - 1.31 years of usable own-multiple history spanning a monotonic post-IPO de-rate from 14.67x to 5.03x - and no peer median is substituted.

Key findings

• THE SCREEN'S SHARE COUNT IS A PRE-IPO FIGURE AND IT IS WRONG BY 2.695x. TTAN's XBRL CommonStockSharesOutstanding tag has exactly two observations, both from the 10-Q filed 2025-01-14 covering the quarter ended 2024-10-31 - six weeks BEFORE the December 2024 IPO: 34,185,388 and 35,397,085. The company has never re-tagged the concept. The scanner took the newest observation of the tag it wanted and got a private-company share count with no staleness signal. Verified from the 10-Q cover page: 82,746,425 Class A + 12,651,154 Class B = 95,397,579. Market cap $7.479bn, not $2.775bn. EV/Sales 6.96x, not 2.42x. The scale cross-check ties exactly: Q1 FY27 net loss $(22,818)k / 95,003,992 weighted shares = $(0.24), which is the filed EPS; on the screen's count the same loss implies $(0.64), a figure that appears in no filing.
• THE NET-CASH ERROR IS SEPARATE AND ADDITIVE. The scanner used current cash ($421.5m at 2026-04-30) less a STALE debt balance ($104.0m at 2025-01-31). TTAN voluntarily repaid the ~$107.0m term loan in full in January 2026; the 10-Q states 'As of April 30, 2026, no loans were outstanding under the Amended Credit Agreement.' TTAN is debt-free with a $250m undrawn revolver to 2031.
• THE SCREEN'S PASS WAS THE ERROR, NOT AN ASSUMPTION CALL. Re-running the screen's OWN parameters (terminal margin 11.2%, exit 23.5x, 5y, WACC 10%) on the corrected share count gives a required CAGR of 33.6% and a margin of -9.5pp: a FAIL. Its stated +18.9pp margin came entirely from the 2.695x share-count error.
• PAYMENTS REVENUE IS BOOKED NET OF INTERCHANGE - WHICH INVERTS THE TERMINAL-MARGIN FEAR. The 10-Q states usage revenue is recognised 'net of interchange and other direct expenses which are passed onto the customer'. So the ~22% payments mix arrives at software gross margin, not payments gross margin. PLATFORM GROSS MARGIN IS 78.7% AND RISING (73% FY25, 77% FY26), with the usage mix flat at 22.8% / 22.5% / 22.2% / 21.8%. The blended-margin drag is entirely professional services - a 3.1%-of-revenue line running at MINUS 136.4% gross margin, a deliberate onboarding subsidy that has shrunk from 5.3% of revenue in two years. TTAN must be modelled as a high-gross-margin software company, not a payments hybrid.
• THE MECHANISM IS GTV, AND IT IS MOSTLY NOT SERVICETITAN'S. Subscription pricing is disclosed as 'linked to the size of the customer's business, generally based on the number of field technicians'. Usage is payment processing on customer invoicing. Management states plainly: 'we focus on increasing the GTV on our platform, rather than new customer count.' GTV grew +19.9% in FY26 against revenue +24.5%, and +22.6% in Q1 FY27 against revenue +24.6%. THE COMPANY-CONTROLLED RESIDUAL - attach, price, net new logos - was +4.6pp in FY26 and only +2.0pp in the most recent quarter. Roughly 80-90% of revenue growth is the customers' own business growth, i.e. US residential trades spend. This is a cyclical exposure dressed as a software compounder.
• IS GROWTH SEATS OR PRICING? NEITHER, IN THE USUAL SENSE. There is NO disclosed list-price increase in any post-IPO filing, no disclosed ARPU and no disclosed seat or technician count. Because subscription pricing is a function of the customer's technician headcount, expansion is volume-linked seat/size growth inside existing accounts - which the company measures only through NRR. The usage take rate on GTV is rising slowly: 0.254% (FY25) to 0.260% (FY26) to 0.270% (Q1 FY27).
• TWO GENUINE DISCLOSURE DEFECTS, BOTH OF THE 'FLOOR, NOT A NUMBER' TYPE. Net dollar retention has been disclosed as '>110%' for FOUR consecutive periods (FY24, FY25, FY26, Q1 FY27) and gross dollar retention as '>95%' for THREE consecutive fiscal years - never as values. A fall from 118% to 111% would produce IDENTICAL disclosure. The two metrics an investor would use to detect the thesis breaking are censored from above, so deterioration is structurally undetectable until the floor breaks, by which point it is several quarters old. Active Customers (~10,800) is annual-only, so logo growth cannot be tracked intra-year. The observable substitutes are quarterly GTV growth and the derivable usage take rate, and the monitoring plan keys on those.
• GROWTH IS ORGANIC AND THE MARGIN IMPROVEMENT IS REAL. Acquisition spend was $1.184m (FY25) and $19.781m (FY26) - 2.1% of revenue, the only FY26 deal being Conduit Tech at $6.97m of identifiable intangibles. No settlement income, no milestone revenue, no distributor or customer concentration. GAAP operating margin improved +12.2pp FY25-to-FY26 while SBC/revenue improved only 0.7pp, so ~11.5pp is genuine operating leverage, not an SBC artifact - and the TTM measure confirms it independently at +13.3pp, past the FY2025 IPO-cost comp base. Cash R&D grew +18.1% against revenue +24.6%.
• $53.6m/YEAR OF EXPENSE FOR AWARDS STRUCK 79% ABOVE SPOT. In October 2024 the company granted each Co-Founder 3,241,544 performance RSUs - 6,483,088 total, 6.8% of shares - vesting on VWAP hurdles from $140 to $440 per share, expiring 2034-10-21. The stock is $78.40; the LOWEST hurdle is 78.6% away. Under ASC 718 the market-condition expense runs regardless and is not reversed if never met, so it inflates the GAAP loss by 5.6pp of revenue for awards that will probably deliver nothing - ex this charge, FY26 operating margin is -12.0%, not -17.6%. But it cuts both ways: above $140 the 6.48m shares issue, so ANY target above $140 must be struck on ~101.9m shares. The share count is discontinuous at $140 and most screens will miss it for exactly the reason they missed the base count.
• SIX FOR SIX ON GUIDANCE, IN A 2.1-POINT BAND. Every quarter since the IPO: +4.64%, +3.70%, +5.73%, +4.69%, +3.67%, +5.01%. Mean +4.57%. At the full-year level FY26 was guided initially at $895-905m and printed $960.965m, +6.8%. This is not a beat, it is a policy - and it is the entire basis of the house FY27 estimate of $1,190m (+23.8%) against a guide of $1,135m (+18.1%), which three independent methods place within $5m of each other. Non-GAAP operating income is sandbagged far harder still: FY26 guided at $48-53m initially, printed $94.058m.
• THE VERDICT IS A TERMINAL-MARGIN CALL AND NOTHING ELSE. At the base 22.7x exit multiple, the sign of the valuation answer flips between a 16% terminal margin (-1.2pp, FAIL) and a 20% terminal margin (+4.3pp, PASS). Growth and the multiple both matter less. Anyone consuming this analysis should interrogate the terminal-margin build and ignore everything else.

Sections

Disclosed limitations

• NO CONSENSUS ESTIMATE WAS RETRIEVED. Alpha Vantage's EARNINGS_ESTIMATES endpoint returned the 25-requests/day quota-exhaustion message. Every forward number in this memo is either company guidance or my own build off the disclosed guidance-beat pattern. THERE IS NO STREET COMPARISON ANYWHERE IN THIS MEMO, and Consensus Criteria is INDETERMINATE.
• THE 12-MONTH TARGET HAS NO IDENTIFIED MULTIPLE ANCHOR and is reported as a conditional band with the multiple held explicitly free, not as a point estimate. Issuing a point target would mean inventing the one parameter that determines it. Resolution: roughly eight more months of trading takes the sample past two years and past the first full post-lock-up year.
• TRANSCRIPT MENTION-FREQUENCY RUNS ON ~550-WORD PRESS RELEASES, NOT CALL TRANSCRIPTS - LOW CONFIDENCE. Alpha Vantage EARNINGS_CALL_TRANSCRIPT was attempted first and returned the same quota message, so no call transcript is in the corpus. The corpus is all seven 8-K Ex-99.1 earnings releases that exist post-IPO (Q3 FY25 to Q1 FY27), narrative section only, single source, no mixing; word counts 603/785/538/527/506/624/544 are near constant so counts are comparable. The finding - that Max, agentic, Agentic Operating System and velocity ALL entered from zero simultaneously in Q4 FY26 and persisted, while the entire IPO-era Pro-product vocabulary (Sales Pro, Pro products, Contact Center) went to zero and stayed there - is a HYPOTHESIS for a proper re-run, not a result. The trend cannot be measured quantitatively on a 550-word corpus and no claim is made that it can. No 'first ever' claim is made about any term.
• THE PEER EV/EBIT SET USED FOR THE EXIT MULTIPLE COMES FROM THE SAME TIER-1 SCANNER WHOSE SHARE COUNT ON TTAN WAS WRONG BY 2.695x. TTAN's own inputs were verified against primary filings; the seven comparators' share counts were NOT. Their market caps are individually plausible against known scale (VEEV $32.7bn, PAYC $7.7bn, MANH $9.9bn, SSNC $18.8bn, TYL $14.1bn), and the TTAN defect was specific to a recent IPO carrying a stale pre-IPO tag - but the exit multiple is MEDIUM confidence, not verified, and the sensitivity grid is the correct way to consume it.
• MAX ADOPTION IS NOT SIZEABLE. 'More than doubled the number of locations on Max' is disclosed off an UNDISCLOSED base. No revenue, ARR or location count has ever been given for Max or Atlas. It is a direction, not a number, and it is treated as such throughout.
• SEAT COUNT, ARPU AND PRICING ARE NOT DISCLOSED IN ANY FILING. The seat-versus-price decomposition is inferred from the disclosed pricing architecture and the GTV/revenue gap; it is not directly observable.
• NRR AND GDR ARE FLOORS, NOT VALUES. Their trend is unobservable by construction. See key findings.
• THE Q2 FY27 EARNINGS DATE IS ESTIMATED from a six-observation filing-lag record (34-41 days after quarter end, always a Thursday), NOT announced by the company. Every estimated date in the calendar is marked ESTIMATED with its derivation shown; no date is fabricated.
• WACC OF 10% IS THE FRAMEWORK DEFAULT, NOT A CAPM BUILD. At 12% - arguably more appropriate for a 54%-realised-vol, zero-debt, GAAP-unprofitable name - the required CAGR is 22.0% and the margin is +2.1pp. Beta was not estimated: 1.31 years of post-IPO history spanning a de-rating regime is unidentified for the same reason the multiple anchor is.
• THE 12-1 MOMENTUM PERCENTILE IS INDETERMINATE. The raw 12-1 return of -33.6% is reported; no cross-sectional universe ranking was computed in this run, and criteria.md requires momentum to be assessed cross-sectionally rather than by absolute rules.
• NO CHART PACK WAS PRODUCED in this run. The mention-frequency exhibit that references/mention-frequency.md requires is present as a table but not as a chart.