How do you tailor a portfolio for a platform PM role versus an applied one?

Bellcastle Property Group is a fictional real estate firm. Marisol Peralta built Marklane there: a model that estimates a property's market value the moment new listing or inspection data comes in. Absalom Njoroge, a hiring manager, watched her present the exact same write-up to a platform-team panel and an applied-team panel, one week apart.

How to answer this, stage by stage

Nobody is grading whether you can build two different projects. They're grading whether you know which part of one project to put first.

Let's learn

A property valuation model is a tool that reads a listing, some inspection notes, and recent nearby sales, and estimates what the property is actually worth right now.

Marisol used one write-up of Marklane for two different interviews in the same week: one for a platform team building shared valuation infrastructure, one for an applied team building the leasing-agent tool that calls it.

The write-up she used was built around one leasing agent's Tuesday: how long a valuation used to take, and how Marklane cut that down. The platform panel's follow-up questions turned skeptical fast: who else calls this, what happens if the schema changes, how is it versioned. The applied panel's follow-ups stayed curious: what did the agent think, what would she change next.

Here's the turn: the platform panel wasn't harsher by nature. The one artifact underneath was written entirely from a single user's point of view, and any platform reviewer reading that immediately wonders whether the candidate has ever thought past one person's desk.

At its worst: Marisol spends the whole back half of the platform interview defending a systems mindset she genuinely has, purely because her write-up gave the panel no early evidence of it.

What I would leave alone: the underlying model, the eval numbers, the actual screenshots, none of that should ever change between the two versions. Only the order and the opening line change.

The lesson: tailoring a portfolio isn't writing two portfolios. It's knowing which true fact about the same build to say first.

Now here is the same thing as a story

The short version above is what you'd say explaining your tailoring choice out loud. Read this one for how Marisol actually learned it.

Marisol Peralta spent four years pricing commercial properties by hand before she ever wrote a prompt, and she can still eyeball a listing and land within a few percent of its real market value.

She built Marklane around one leasing agent, Delphine, who used to spend forty minutes pulling comps before pricing a new listing. Marklane cut that to under five.

The same valuation number reads two different ways: a dial one person checks, or a pipe three teams already depend on.

The applied-team interview went well. She told Delphine's story, showed the before-and-after screenshot, and the panel asked exactly the questions she expected: what did Delphine think, how did she react the first time it was wrong.

The platform-team interview, four days later, used the identical write-up. The questions turned in three quiet steps. First, a mild "so who else uses this valuation number." Then, a sharper "what happens to the leasing tool if you change how this number is calculated." Then, flatly, "walk me through how you'd version this."

One mismatch cost a single follow-up question. The other cost the rest of the hour.

Marisol didn't have a clean answer, because her write-up had never once mentioned that three other internal tools, a mortgage pre-approval flow and two listing services, already called the same valuation number under the hood.

All four of these facts were true the whole time. None of them were in the version she showed the platform panel.

Rebuilding the platform-facing opening took her about ninety minutes, mostly reordering slides she already had. The valuation model's schema, its versioning plan, the three teams already depending on it, all of that moved to the very first paragraph. Delphine's story moved to a supporting section further down.

Nothing about the build changed. Only the order the same true facts appeared in.

Building one write-up and assuming she'd just verbally reframe it live had felt efficient back when every conversation about Marklane had been casual, over coffee, with room to adjust on the fly. It stopped working the moment a panel asked her to share her screen and walk through the actual page, where the framing was already locked in writing.

The same four pieces of evidence, sorted by which room actually wants to see them first.

The old page asked every reader to find the platform story buried inside a user story. The new one just told each reader the true fact they'd come looking for, first.

I built one page and planned to reframe it live. It took one screen-share and three sharpening questions to see that a written page doesn't get a second chance to lead with the right fact.

PICK, applied to your own portfolioNot two portfolios. PICK is what decides which true fact about one project goes first.

None of these five questions are about the model's accuracy. All five are about what happens when other teams depend on it.

The recap, one line per letter: position is committing to two different openings from one project, impact is naming who gets confused which way, cost asymmetry is the platform mismatch costing the whole interview, and kill criteria is the user-story default when the role is genuinely unclear.

And if you want to be sure it really works, try it somewhere elseSame four letters, a pharmacy chain instead of a real estate firm. A different building, and the platform panel's question changes shape too.

Pinehollow Pharmacy Group is a fictional pharmacy chain. Jonas Eckhart built a drug-interaction flagging tool there, one that checks a new prescription against a patient's existing medications before a pharmacist fills it. Selma Ridgeway reviews his write-up for two different roles.

Mapped onto PICK: his position for an applied pharmacy-tools role opens with one pharmacist, Rosa, and the three minutes she used to spend manually cross-checking a complex prescription. His position for a platform role opens with the interaction-checking service itself, now called by the pharmacy's mobile app, its refill-reminder system, and a nurse-facing tool at a partner clinic. The impact: an applied panel reading the platform opening might wonder if he's ever actually watched a pharmacist work. A platform panel reading the pharmacist story alone might wonder if he understands what happens when three different products depend on the same flagging logic and one of them needs a different alert threshold. The cost asymmetry holds here too: the platform mismatch would cost him defending systems thinking for the whole interview, while the applied mismatch costs one clarifying question about the pharmacist herself. His kill criteria, since the actual req mentioned "cross-product consistency" twice, was to lead with the platform framing this time, correctly guessing which room he was in before the interview even started.

Swap the trigger and it still runs.
Speed: an interviewer caps you at sixty seconds. Say "same build, I open with the contract for a platform room and the user's day for an applied one," and stop.
Cost: there's no time to build two full write-ups before an interview tomorrow. Keep one page, and just rewrite the first paragraph and reorder three sections; that's usually enough.
The model gets better, for real: if Marklane's accuracy improves next quarter, that number strengthens both versions equally. It doesn't change which fact you lead with.

Where people run it wrong.
They build two entirely separate write-ups and end up quietly changing a metric or a detail between them, which is a far bigger problem than a mismatched opening.
They assume a platform room just wants more technical detail, when what it actually wants is evidence of thinking about multiple consumers, not more architecture diagrams.
They guess the room type from the company's size alone, instead of reading the actual responsibilities listed in the req.

How to use it live. When someone asks how to tailor a portfolio, ask yourself one question first: does this room want to hear about one user, or about everyone who depends on this. Lead with the answer to that question, and let the rest of the page follow it.

Flashcards (tap any card to flip it)

Check yourself Score: 0 / 0