How we curate.

FlipMedellín scans thousands of active Medellín apartment listings across major portals, boutique brokers, and bank-owned inventory channels. Each is priced against a calibrated reference range for its own barrio, and a listing clears our below-market gate only when its asking price per m² sits at least 10% below the calibrated midpoint for its barrio. Clearing the gate earns a closer human look — condition, red flags, a remodel range, editorial scores — not an automatic verdict. When a listing clears and survives that review, we call it a deal worth a look, and we show every number behind the call.

The gate is one calculation, and we keep it unmistakable: a listing clears only when its asking price per square meter is at least 10% below the calibrated midpoint for its resolved barrio. Clearing the gate earns further review; it does not, by itself, make the property a good investment. The number gets your attention. The investigation tells you whether it is a deal.

Advertised square meters are not always comparable. Portal figures may refer to built, private, or total area, and terraces or other outdoor space can distort a price-per-m² number. We run an area-basis confidence check before treating an unusually low figure as a genuine discount.

We scan major property portals, boutique brokers, and bank-owned inventory channels daily. We dedupe the same unit when it's posted at different prices, normalize barrio names that vary in spelling and accent across sources, and check that a listing is still live before it surfaces.

A source page being live does not guarantee the property is still available. We record when we last checked the source and treat portal availability as evidence of publication — not confirmation from the seller. We do not republish stale inventory, and we retire listings that go dark at the source.

Each barrio's range is an asking-side market reference, built from active asking prices in our own index and adjusted conservatively for the typical gap between asking and negotiated prices. It is a reference, not an appraisal, and not proof of where a specific unit will close. We publish a range, never a single point, because two comparable units in the same barrio routinely sit 10% apart.

Every range carries its sample size and date, so you can see what the comparison stands on. Where a sub-zone's sample is thin, a curator sets the range by hand and it's labeled a curator estimate until the data supports an indexed one. And when a listing can't be resolved confidently to a specific barrio, we either use a clearly disclosed broader reference range or withhold the below-market conclusion — the system never silently guesses a neighborhood.

Pricing variance and the basis for our range

Medellín asking prices vary widely, even among superficially similar apartments. Differences in condition, floor, view, building age, renovation quality and seller expectations all contribute to that spread. Duplicate and inconsistently described portal listings introduce additional noise.

For an indexed asking-price source, we begin with the observed midpoint and apply an 8% downward adjustment for the assumed difference between asking and negotiated prices. We then publish a range extending 6% below and 6% above that adjusted midpoint. When a source is already stored as an adjusted or curator-set reference, we do not apply the adjustment a second time.

This produces a consistent screening range. It is not an appraisal, a prediction of the eventual closing price or proof that comparable properties have transacted inside the range.

Editorial selection over algorithmic ranking. From the inventory that clears the gate, we apply human judgment about condition, building character, and neighborhood liquidity, working from the listing materials: photos, stated specs, building data, and price history. Unless a listing is explicitly marked as visited, nobody from FlipMedellín has been inside it. We weigh elevator access in mid-rise buildings, building age against renovation evidence, HOA fees against area, and how long the unit has sat at the same price.

Editorial scores per listing, 1 to 10: Flip rates the renovate-and-resell math, BnB rates short- and medium-term rental potential, and Home rates the property as an owner-occupier home. Scores are editorial reads on our curation framework, not algorithmic predictions or buy-or-sell calls; we rarely publish below 4, and reserve 10 for the rare listing that clears every test. On short stays specifically: eligibility depends on current regulations, RNT registration, and — critically — the building's propiedad horizontal rules. Verify all three before underwriting short-stay income.

Where we model rental income or a renovation, we publish a range, never a single number. A rental range (in our Pro deal memos) is an estimate to underwrite against, not a promised yield; a remodel range is an envelope based on comparable local work for kitchens, bathrooms, flooring, electrical, and finishes — a starting point, not a contractor's quote. We do not predict appreciation.

Property portals publish asking prices, not closing prices. Colombia does provide some registered transaction information: SNR data offers aggregate market activity, and a property's registry certificate (certificado de tradición y libertad) may reveal prior registered transfers. Coverage and detail vary, so we present sold-side evidence separately and do not treat it as a complete comparable-sales database — and it does not currently feed the below-market gate, which runs on the calibrated asking-side range.

Public listings also carry errors, and sub-barrio tagging is inconsistent across portals — Provenza alone gets tagged El Poblado, Provenza, or Manila depending on the broker. Our normalization catches most but not all, and where a sub-zone's data is thin, we flag it on the listing.

When we say a listing is below market, we mean one specific, checkable thing: its asking price per m² sits at least 10% below the calibrated midpoint of its barrio's range in our data. We do not predict appreciation and we do not promise a rental yield. Markets decide what things are worth; we surface the inputs, flag the gap, and explain the property.

And the gate can't hear the street at 11pm, smell the humidity in the back bedroom, meet the neighbors, or catch the bedroom that is really a closet. Some listings that clear will not survive your visit — that isn't the math failing, it's where the math's job ends and yours begins. What we save you from is spending time on the large majority of listings that never clear the initial price screen. The shortlist still has to be walked, the title studied, and the building's rules (RPH) read, ideally with a Colombian real-estate lawyer. Anyone who tells you a website replaces that is selling something we refuse to sell.