Attribution inside a single P&L has to net to the reported total, so if the factors Carlex credited with $10 million of positive EBITDA were in fact costing it $2 million [4], either the consolidated number was overstated by that $12 million [13] or the margin was being earned somewhere the original analysis never looked. The case study as described does not say which, and it does not name the factors, the period covered, or the division's total EBITDA [17].
The generous reading is that the Aftermarket Division ran an allocation convention that flattered a set of products or customers and Sapien changed the convention, in which case neither figure is true in any deep sense and the useful question is which one prices better. The ungenerous reading is that a spreadsheet was wrong, which makes this one division's bookkeeping failure rather than evidence about anyone else's. The third reading, or rather the more interesting one, leans on the smaller number: the $1.5 million Sapien says it found in a customer-channel pattern nobody was looking for [5] is incremental rather than reallocated, and incremental findings are the ones that eventually show up in cash.
Jason Waltz's two weeks against 20 minutes [6][5] works out at roughly 240 times faster on a 40-hour week [15], though the honest version of that arithmetic is not that Carlex banked a fortnight of analyst time, because Waltz also said his team probably would not have reached that level [6]. What the division was giving up was never the hours. It was the questions it had stopped asking because each one cost two weeks.
On price: the round values a two-year-old company at $180 million [1] with no dollar amount reported [16], against an $8.7 million seed led by General Catalyst in 2024 [7] and headcount up fivefold in a year [8]. That puts the equity at about fifteen times the swing Sapien found in one division of one customer [18], which is either a bet that the finding repeats across Bayer, Cooper Standard, Blink Charging and Westgate Resorts [3] or a lot of money for one case study. Ron Nachum's own framing names the constraint on the first outcome: large companies are data rich and analysis poor, and the hard part is getting finance teams to trust the answer [10]. A vendor that refuses to be an Excel copilot [11] cannot sell on hours saved; it has to sell on decisions changed, and that requires a buyer to concede in advance that its existing answer is wrong.
Which is where the risk turns around. Carlex has extended the platform into pricing, inventory, customer orders, OEM quoting, supply chain and operations [9], so the rebuilt attribution is now upstream of a good deal more than one margin report, and an error in the rebuild would propagate through those decisions considerably faster than the one it replaced did. Sapien's other published results so far are stated as hours or days compressed into minutes [12], not as signs flipped, and speed is the easier claim to make.
