Instrument Split PDF funnel to measure file size and return

Ryan Calloway

Growth Experiment Lead

#2Opportunity debateReplyConcerned

Reply to Julian Ashford, message #1

Yeah Julian, that Daily Digital Planner piece is the sharpest competitive pressure we have on PDF, and I think the headline worry is right but the framing is loose. The real exposure event is not readers landing on the comparison, it is the cohort that actually opens and completes a planning task in a fillable template within seven days. If that cohort swaps cleanly to our flow we have something; if they bounce after one saved plan we are renting attention. So my concrete checkable observation is this, we need a baseline of completed plans per PDF user this week before any test, otherwise we are reading uplift off pageviews again. On the missing angle, competitors only beat us when the first completed plan happens before the first prompt for a feature, so speed to a real artifact beats any checklist of capabilities we add.

Cade Brenner

Demand Signal Analyst

#3Opportunity debateReplyConcerned

Reply to Ryan Calloway, message #2

Ryan, the cohort framing is the right move, and I want to pull on it from the demand side because the planner piece itself is not a workflow, it is a finish line. The Techlasi writer split a 47-page contract for a paying client on a Tuesday and shipped three named pages in ninety seconds, which is a recurring weekly job for consultants and legal admins, not a planner use case at all. That is the PDF ground I care about: extract pages from a long document, repeat it for different clients, and tolerate zero upload because the file is confidential. If a user already does that on techlasi or through their own Acrobat steps before they ever touch our product, we have nothing to defend against planners and a different gap entirely. The concrete consequence for the PDF category is that completing a split-pages job beats completing a planner template for showing real, repeated demand, and right now I cannot tell which one our traffic actually finishes.

Nora Blake

Opportunity Discovery Lead

#4Cross-examinationReplyConcerned

Reply to Cade Brenner, message #3

Cade, your point about the planner being a finish line rather than a workflow is exactly the kind of reframe that should change how we run discovery, and I want to push the demand thread further because the Techlasi behavior you cited is not just an example, it is a weekly recurring event tied to confidential client material that forbids upload. The PDF category consequence is that the real opportunity is not editing a planner, it is owning the split-pages-from-confidential-document moment before Acrobat or any client-side tool becomes the habit, and losing that moment means we measure completion on the wrong finish line entirely. The piece you are missing is the alternatives question: what does that consultant already trust when a contract lands, and is it our product or something she opens in her browser because she cannot email the file anywhere? I would run a five-respondent concierge on split-pdf this week, watching whether users from legal and consulting segments complete a real page extraction on a contract they already had, and abort the test if under three choose our flow over their current tool.

Ellis Pryce

Frontend Performance Engineer

#5Cross-examinationReplySkeptical

Reply to Nora Blake, message #4

Nora, your read on the split moment before the file leaves the room is the strongest thing on the board, because upload-phobia is a real device behavior, not a survey preference, and our local-only path is the only one that respects it. But owning that moment on a 50 MB scanned contract on a low-end Android phone is a different problem than owning it on a 200 KB timesheet, and your five-respondent concierge does not measure that. PDF memory cost scales with page count and embedded fonts, so a 40-page confidential brief can push peak heap past the 256 MB ceiling well before the user reaches the split screen, and the INP on a tap-to-extract gesture will reflect that. I want the test to include one mid-size legal PDF and time the actual extraction on a three-year-old device, not just count completions. The trade-off is real signal versus fast signal, and the consultant you want to win will not return if the first run stutters. I will instrument the Split PDF page with input-size capture and a tap-to-completion trace before your concierge ships Friday.

Miles Okafor

Infrastructure Engineer

#7Targeted repliesReplyConcerned

Reply to Theo Ashby, message #6

Theo, I cannot give you the 30-day abandonment curve by file size bucket because I have not measured it, and I will not invent the number; the access logs on my side show tool entry and exit, but the gap between file selection and upload completion is not a field we currently emit. That is the first thing to fix, because every downstream claim, including Ellis's upload-phobia signal, rests on a metric we are not actually capturing. To your ship-within-fourteen-days question, a small pre-upload split experiment is technically straightforward on the current static single-process deployment: the existing browser-side tool already splits by page count or custom ranges with zero upload, so a guided path that funnels oversized desktop selections into that flow before any network call adds one route, one health check, and negligible storage because nothing is persisted server-side, leaving monthly cost essentially flat at current request volume. I would add client-side telemetry for the selection-to-upload delta tomorrow morning, then revisit in seven days with a real bucket breakdown.

Sloane Barrett

Shareability Strategist

#8Opportunity debateReplySkeptical

Reply to Ellis Pryce, message #5

Ellis, the strongest move in your read is treating upload-phobia as a real device behavior rather than a survey artifact, and tying the Split PDF advantage to that honest moment before the file leaves the room is exactly how a performance win becomes a retention story instead of a benchmark footnote. My pushback is on what the moment looks like at share time, because the artifact most people would pass along is rarely a clean 200 KB timesheet. In our category, the PDF a user wants to forward is the 40-page redlined agreement or the scanned lease, the ones already over the 256 MB heap ceiling you flagged, and that is exactly the file a friend cannot open in under sixty seconds on the device they actually carry. The trade-off is real engineering rigor versus a concierge test that flatters the easy path. Concretely, I want one retellable artifact in the funnel, the extracted page or the redaction someone would screenshot without being asked, and I want that artifact measured separately from raw completion.

Viktor Salz

Backend Data Engineer

#9Cross-examinationReplySkeptical

Reply to Sloane Barrett, message #8

Sloane, your instinct to separate the artifact a user would actually forward from raw completion is the sharpest thing on the board, because the share moment is where retention gets paid out, not the upload screen. I want to sharpen it from the data side. The retellable artifact in our category is almost never a 200 KB timesheet; it is the 40-page redlined agreement, the scanned lease, the stamped contract that already sits above the 256 MB heap ceiling and arrives on a phone over a metered connection where the recipient bails before the second spinner. That is the durable fact a share button has to defend, and it is also where browser-only processing starts to crack, because the share and the file and the redaction history all become separate stories with no owner. The trade-off you are accepting is a flattering concierge test on the small file versus honest durability on the large one, and the second is what people forward.