How to Plan a 5-Year Career in Data

Planning a 5-year career in data sounds impossible in a field where tools change every year. The trick is not to predict job titles but to choose a direction, set milestones you control, and review the plan regularly against what is actually happening. Two kinds of evidence help with the direction: what employers expect, and what practitioners are moving towards.

What employers expect

The World Economic Forum's Future of Jobs Report 2025 lists technology roles as the fastest-growing jobs in percentage terms, with big data specialists, fintech engineers, and AI and machine learning specialists among them. It also estimates that 39% of workers' existing skills will be transformed or become outdated between 2025 and 2030 (WEF: Future of Jobs 2025). The direction is favourable for data careers, but standing still is not an option: a five-year plan has to include continuous learning.

What data professionals want to learn next

The 2025 Stack Overflow survey asked which technologies people used and which they wanted to work with next year (Stack Overflow survey 2025). Comparing the two among data professionals shows where interest is moving:

import pandas as pd

so = pd.read_csv("results.csv", low_memory=False)          # 2025 Stack Overflow Developer Survey
data = so[so.DevType.isin(["Data engineer", "Data scientist", "Data or business analyst", "AI/ML engineer"])]


def interest(have_col, want_col, top=8, min_n=60):
    """For each technology: share of data professionals using it now, and share who want to work with it next year."""
    answered = data.dropna(subset=[have_col, want_col])
    have = answered[have_col].str.split(";").explode().value_counts() / len(answered)
    want = answered[want_col].str.split(";").explode().value_counts() / len(answered)
    t = pd.DataFrame({"use now": have, "want next year": want}).fillna(0)
    t["net interest"] = t["want next year"] - t["use now"]
    counts = answered[want_col].str.split(";").explode().value_counts()
    t = t[counts.reindex(t.index).fillna(0) >= min_n]
    return t.sort_values("net interest", ascending=False).head(top), len(answered)


for title, have_col, want_col in (("languages", "LanguageHaveWorkedWith", "LanguageWantToWorkWith"),
                                  ("databases and data platforms", "DatabaseHaveWorkedWith", "DatabaseWantToWorkWith"),
                                  ("cloud and platforms", "PlatformHaveWorkedWith", "PlatformWantToWorkWith")):
    t, n = interest(have_col, want_col)
    print(f"F {title}: most wanted relative to current use (data professionals, n={n:,})")
    for name, r in t.iterrows():
        print(f"   {name[:32]:<32} use now {r['use now']:>4.0%} | want next year {r['want next year']:>4.0%} | {r['net interest']:+.0%}")
    print()

core = data.dropna(subset=["LanguageHaveWorkedWith", "LanguageWantToWorkWith"])
for lang in ["SQL", "Python"]:
    keep = (core.LanguageHaveWorkedWith.str.contains(rf"(?:^|;){lang}(?:;|$)") & core.LanguageWantToWorkWith.str.contains(rf"(?:^|;){lang}(?:;|$)")).sum()
    use = core.LanguageHaveWorkedWith.str.contains(rf"(?:^|;){lang}(?:;|$)").sum()
    print(f"F {lang}: {keep / use:.0%} of current users want to keep using it next year")

F languages: most wanted relative to current use (data professionals, n=1,431)
   Rust                             use now  12% | want next year  34% | +22%
   Go                               use now  11% | want next year  23% | +11%
   Zig                              use now   1% | want next year   6% | +5%
   Mojo                             use now   1% | want next year   6% | +5%
   Elixir                           use now   1% | want next year   5% | +3%
   Lisp                             use now   2% | want next year   5% | +3%
   Swift                            use now   2% | want next year   5% | +3%
   MicroPython                      use now   2% | want next year   5% | +3%

F databases and data platforms: most wanted relative to current use (data professionals, n=1,153)
   DuckDB                           use now  14% | want next year  27% | +13%
   Snowflake                        use now  16% | want next year  22% | +6%
   Databricks SQL                   use now  17% | want next year  22% | +5%
   Redis                            use now  16% | want next year  19% | +3%
   Clickhouse                       use now   5% | want next year   8% | +3%
   Neo4J                            use now   7% | want next year  10% | +3%
   Cassandra                        use now   4% | want next year   6% | +3%
   Supabase                         use now   5% | want next year   7% | +2%

F cloud and platforms: most wanted relative to current use (data professionals, n=1,051)
   Kubernetes                       use now  29% | want next year  37% | +8%
   Terraform                        use now  20% | want next year  25% | +5%
   Cargo                            use now  11% | want next year  14% | +3%
   Prometheus                       use now   8% | want next year   9% | +2%
   Supabase                         use now   5% | want next year   6% | +1%
   Vercel                           use now   5% | want next year   7% | +1%
   Podman                           use now   9% | want next year  10% | +1%
   Firebase                         use now   7% | want next year   8% | +0%

F SQL: 70% of current users want to keep using it next year
F Python: 77% of current users want to keep using it next year

Languages. Rust had the biggest gap between current use (12%) and interest for next year (34%), followed by Go. Interest is not the same as job demand, so treat it as a signal of curiosity about where tooling may go, not as a hiring requirement.

Data platforms. DuckDB showed the strongest rise in interest, from 14% using it to 27% wanting to, followed by Snowflake and Databricks SQL. Interest is moving towards fast analytical engines and managed cloud platforms.

Infrastructure. Kubernetes and Terraform led among platforms, a sign that data work is moving closer to engineering and deployment.

The core stays. 70% of data professionals who used SQL and 77% who used Python wanted to keep using them next year. Whatever else changes in five years, those foundations remain.

How to build the plan

Step 1: choose a direction, not a title

Pick one of a few broad directions for the next five years: deeper technical specialism (for example, data platform engineering or machine learning), broader business impact (analytics leadership, product analytics), or people leadership. You can change direction later; you need one now to decide what to learn.

Step 2: work backwards in yearly milestones

Milestones should be things you can mostly control: skills, projects, responsibilities. Titles and pay follow from them, though not always on schedule. How to get promoted in a data team covers turning milestones into a promotion case.

Step 3: keep a T-shaped skill plan

Keep the core broad, SQL, Python, data modelling, statistics and communication, and go deep in one or two areas that fit your direction. Use the survey signals to choose experiments, not to chase every trend: try DuckDB on a project, or learn enough infrastructure to deploy your own work.

Step 4: build evidence each year

Every year should produce something you can point to: a project, a system, a published analysis, a talk, a team you helped grow. Evidence is what lets you change direction if you need to.

Step 5: review twice a year

Every six months, ask: Is my direction still right? Am I on track for this year's milestone? What has changed in the field or my company? Adjust the next year, not the whole plan. Upskilling vs reskilling helps when the review suggests a bigger change.

Signals to watch at each review

At each six-monthly review, look at evidence rather than headlines. Read twenty current job postings for the role you want in two years and note which skills appear repeatedly; that is the most direct measure of demand. Notice what your own team is adopting and what it is retiring. Look at what the senior people you admire learned in the last year. And check your own energy: which work did you enjoy, and which did you avoid? A plan that ignores what you enjoy rarely survives five years. If three or more of these signals point the same way, adjust your next year's milestone; if they disagree, run a small experiment, such as a side project or a short course, before making a larger change.

Common planning mistakes

• Planning titles instead of skills and evidence.
• Chasing every new tool instead of going deep in one area.
• Never reviewing the plan, so it becomes a document nobody reads.
• Ignoring personal priorities such as location, family and the kind of work you enjoy.

For the specific path from analyst to scientist to lead, see analyst to scientist to lead.

Foundations that stay useful for the next five years are taught in the Data Engineering course.

Sources

• Future of Jobs Report 2025 (World Economic Forum) — World Economic Forum
• 2025 Stack Overflow Developer Survey — Stack Overflow