Actionable insights — Uber: AV Anxiety

1. Before believing a revenue miss, check whether the accounting changed

1. When reported revenue misses but volume metrics don't, look for a revenue-recognition change — most commonly a shift between the merchant/gross model (book the whole transaction, expense the supplier's cut) and the agency/net model (book only the commission). The cash is identical; the reported line halves.
2. Find management's own quantification of the impact in points of growth, and add it back before comparing to consensus.
3. Re-anchor on the metric the change does not touch — gross bookings, gross merchandise value, total payment volume, units — and judge the quarter there.
4. Note whether the change is confined to one geography or product; a narrow scope confirms it is presentation, not deterioration.

• The quarter the change laps (growth optically re-accelerates for no operating reason); any second change in another geography; and whether take-rate metrics get restated alongside it.

2. Locate a deceleration on the map, then ask if it is supply or demand

1. When a volume growth rate slows a point or two, refuse the company-level explanation and ask which market caused it. Management usually names it when the answer is flattering.
2. Classify the cause as demand (fewer customers wanting the service) or supply (not enough drivers/inventory/capacity to serve the demand). Supply constraints are competitive-spend problems with a known fix; demand loss is structural.
3. Check the rest of the portfolio for the opposite signal — a market accelerating while another drags proves the slowdown isn't systemic.
4. Where price moved, test elasticity directly: compare volume growth in the markets that cut price most against those that didn't.

• Whether the constrained market's growth recovers once driver incentives normalise; whether the price-cut markets keep converting fares into trips (or the elasticity fades); and the cost of the incentives showing up in segment margin.

3. Test a disruption narrative twice — against the multiple, and against share in the disrupted markets

1. Write down what the market is already assuming: express the valuation as a forward multiple on the metric management guides to, and compare it to the current growth rate. A high-growth business on a low multiple means the feared outcome is priced.
2. Separate the two questions the narrative conflates: does the incumbent's capability get replaced, or its demand aggregation? Name the specific functions the challenger would still have to build — dispatch, fleet operations, insurance, regulatory relations, customer acquisition.
3. Find the natural experiment: the markets where the disruptive technology is furthest along. Track the incumbent's category share there versus a year ago — the only live evidence available.
4. Watch the incumbent's optionality budget (free cash flow available to buy into, or partner with, the disruptor) rather than assuming a fixed position.

• The first mature-AV market where category share falls; the seven→fifteen city count actually landing; and whether the multiple re-rates before or after the share data turns — the sequencing tells you what the market is really trading.

4. In a console/hardware transition, read the previous generation's software line

1. Don't judge a platform changeover on new-unit sell-through alone; launch quarters are supply- and price-determined, and comparisons to the prior generation's launch are noisy.
2. Compare installed base at the equivalent point in the cycle instead — cumulative machines is what software attaches to.
3. Then read the old generation's software units. Where backward compatibility exists, a rising old-catalogue number means the legacy installed base is still monetising while hardware migrates — high-margin revenue bridging the trough.
4. Check mix quality alongside it (digital share of software revenue, licensing/IP revenue) and strip any one-time item — tariff refunds, legal reversals — before crediting the margin.

• The September price rise to $500 landing into the holiday quarter — a hardware price increase at the seasonally decisive moment; whether old-generation software units roll over once the new library fills out; and whether FY27's unchanged targets survive the first post-price-rise print.

5. Measure whether AI is a distribution channel or a disintermediation risk — with the conversion delta

1. Ask the platform for two separate AI metrics: AI-referred traffic and AI-attributed orders. Traffic alone proves curiosity; orders prove the channel monetises.
2. Find the conversion differential between agents reading the platform's structured product feed and agents scraping ordinary web pages. That gap is the measurable value of being machine-readable — and the moat, if it persists.
3. Check the distribution of AI-driven orders across categories. Concentration in head categories means AI favours incumbent brands; a long-tail skew means AI discovery is redistributing demand toward smaller sellers.
4. Map that distribution onto the platform's own customer mix to decide whether AI is additive or corrosive to this business specifically.

• Whether the 2x structured-vs-scraped conversion gap widens or compresses as agents improve at reading raw pages; whether AI orders keep skewing long-tail once large brands build their own agent feeds; and any quarter where AI traffic grows but AI orders don't.