Barcelona's tech job market, rebuilt from 2.3 million career histories

Industry labels do most of the work. The VC-backed rule admits any vertical with a funding round and a founding year of 2005 or later — Glovo-type companies whose label says "Consumer Services". The description and team-mix rules are capped at 2,000 listed employees so a bank or a pharma group cannot slip in on the word "software".

Half of the people sit at foreign companies' hubs (49%). The map's "born in Barcelona vs hub" filter is this field.

4,139 of the 5,552 employers have fewer than twenty people in the city; the 43 with five hundred or more hold 27% of the people.

The map's cohort charts work for 79% of the universe; companies founded before 1990 are not shown here.

A company-level flag, set from its own description or from the titles of its Barcelona team, and recorded per company so the client can audit it. Person-level AI titles are a separate measure, in City today.

25% of titles carry no function word the rule knows (42,296); 2.3% have no title at all. Data / ML / AI is drawn in coral because it is the thread the project follows.

"Mid" is the default when a title carries no level word, so it is overstated; the pilot's heat table shows the same split per company. Founder titles: 4,699.

This is the tenure profile of today's workforce, not hiring history: people who joined and left are not here. 14,322 started before 2015, 16,527 have no start date. For hires per year see History.

Median 15 years in the market. The brief's years_experience field is this number per person.

Values are percentages: of the current positions that started in 2025, 2.6% carry an AI/ML word in the title, against 0.6% of those that started in 2019. 19,701 people work at a company with the AI flag — a different, broader measure.

• Software Engineer1,808
• Senior Software Engineer1,455
• Project Manager1,277
• Account Manager830
• Product Manager699
• Account Executive664
• Software Developer655
• Frontend Developer487
• Product Designer472
• Engineering Manager464
• Sales Development Representative442
• Customer Success Manager439

Title normalisation for Table 3 (title_normalized) is a mapping the pilot already applies; the long tail is where the rule set grows.

Four in five Barcelona positions carry their own location; the profile fallback adds the fifth. 57% of all positions in the data carry a location string. The employers are global — most of their positions are elsewhere and are the company's, not the city's.

Past years are undercounted by construction: someone who left Barcelona and has no location on the old position drops out of the fallback rule, and profiles deleted since are gone. Read the trend; the pilot measures the undercount.

2026 is partial and not shown. Departures are undercounted relative to hires in the latest year because an end date is written when the next job appears.

Percent of the year's hires. The rule is a title regex (AI, ML, machine learning, data scientist, computer vision …), the same one used across the page.

What the data cannot show is the share of the real 2019 workforce that is not on a profile at all. The public employee counts the client holds for the pilot companies are the benchmark for that gap — see the pilot table.

Titles read on a systematic 1-in-8 sample of the Barcelona positions (every eighth profile id), scaled back in the last row. The full pull reads every row.

A departure is written when the next position appears, so the latest year under-reports leavers.

The brief's internal_moves column: promotions and team changes that the person recorded as a new position.

This is the pilot's answer to question 3 at person level: of the people we can see in 2019, how many are still visible, and where.

31% of hires with a located previous job came from abroad — the relocation figure the map's "where talent comes from" view needs. Blue marks geography throughout.

"No next position" is the latest leavers and people who left the professional network; both shrink as time passes.

"Located" is the share of hires whose previous position carried a location; the three shares are of that subset.

The Sankey in the brief is this table at the scale of the whole universe: company pairs × year.

2021 starts in July and 2026 ends mid-October — both hatched. A further 21,698 Barcelona postings carry no posted date and sit outside the chart.

Seniority level, employment type and applicant count are present on effectively every posting; a stated salary on almost none until 2025, when pay-transparency rules begin to show. "Still open" is the posting not yet marked closed at the last read — the map's "open roles now".

The values come in a handful of round bands with no currency, period or gross/net on the record, and on 58% of the rows the "min" is above the "max" — the page reads the pair as an unordered band. It is an ordinal signal from a model upstream of our data plane — fit for "which band is more common in engineering than in sales", not for a figure per person. Use posted ranges and an outside benchmark for euros.

Inferred upstream from the first name; the two spellings on the record are merged here. Coverage is 15%, so any gender split is a sample of the workforce, not the workforce.