A recruiter just emailed you a Data Scientist offer at $138,000 base. You have 48 hours. You don't know if that's strong, weak, or insulting — and the BLS median says one thing, Levels.fyi says another, and your last manager said "the market is soft right now."
The market isn't soft. The market is split. Knowing which side of the split your offer sits on is the entire negotiation.
The 2026 data scientist market, in one paragraph
Two roles live under the same job title now. Analytics-DS — dashboards, A/B tests, SQL-heavy product analytics — has flattened. ML/AI-DS — production models, LLM fine-tuning, eval pipelines — is paying premiums that didn't exist 18 months ago. The BLS reports a $112,590 median annual wage for data scientists as of May 2024¹. Levels.fyi puts median total comp at $176,250 in May 2026². That gap is not noise. It is the bifurcation.
Before you counter, you need to know:
- Which role your offer actually is (titles lie).
- Which side of the bifurcation the company sits on.
- What the company can actually pay right now.
Then you counter.
The honest part
Headline numbers about AI premiums are real but narrow. PwC's 2025 Global AI Jobs Barometer found workers with genuine AI skills earn a 56% wage premium over same-role peers without them³. That sounds like leverage. It usually isn't yours.
Robert Half's 2026 Salary Guide shows tech salary growth decelerated to 1.6% year-over-year. Series A/B startups have cut analytics-titled data science postings by 15–20%. If your last three years were dashboards, SQL, and the occasional regression, the AI-premium headlines do not apply to you. You are negotiating against a slack market.
If your last three years included shipping production models, fine-tuning open-weight LLMs, building eval harnesses, or owning an inference budget — the premium applies. Aggressively.
Most candidates sit in the middle. Some Python, some ML coursework, one production model that mostly worked. Middle of the road is where most counters get botched, because the candidate either over-claims the premium or under-claims it. Neither works.
The 2026 anchor points you actually negotiate against
Stop using one number. Use three.
Floor — BLS $112,590. The official national median. Useful for one thing only: making sure you're not being lowballed against the public-record number. If an offer is below this, it's not a negotiation — it's a rejection waiting to happen.
Tech-sector base — $118,000–$122,000. Cadence's 2026 projection, derived from BLS plus ~4% annual tech comp drift over two years⁴. This is your base-salary anchor for non-FAANG tech roles, mid-level, U.S. metro.
Tech-sector total comp — $176,250. Levels.fyi median total compensation across tech in May 2026². This is the ceiling for the average tech data scientist — base, equity, target bonus, all in. Anchor your TC counter here, not at the base number.
If the role is AI/ML-engineering-adjacent, add a fourth anchor: Robert Half's 2026 mid-band for AI/ML Engineers at $170,750⁵. That's the bridge benchmark for data scientists pivoting toward production ML.
The BLS number is, in the words of one analyst, "technically accurate and practically useless" for tech-sector negotiators because it aggregates ML research scientists at frontier AI labs with analytics-titled roles at regional banks. Don't anchor your tech counter to BLS. Use it as the floor, not the target.
Move 1: Identify which role you actually have
Read the job description, then read it again. Look for these signals.
Analytics-DS signals: "partner with product," "cross-functional communication," "A/B testing framework," "dashboards," "SQL," "experimentation."
ML/AI-DS signals: "production model," "MLOps," "inference," "fine-tuning," "eval," "embeddings," "vector," "LLM," "training pipeline," "model serving."
Hybrid signals: "model deployment," "Python in production," "feature engineering at scale," "collaborate with ML engineers."
Your counter calibrates against the role you actually have, not the one on the offer letter. If the JD says "Senior Data Scientist" but the responsibilities are SQL and dashboards, anchor low and negotiate hard on title. Title is comp two years from now.
Move 2: Read the company before you read the offer
This is the move most candidates skip and the one that decides the counter. The company's funding stage, hiring temperature, and recent news tell you what they can pay and how badly they need you.
Public signals you can pull in 10 minutes:
- Funding stage and last raise. Series A is tight. Series C with a recent raise is loose. Public and profitable is loose-but-disciplined.
- Layoffs in the last 12 months. WARN Act filings, TechCrunch, LinkedIn. A company that just cut 8% will not pay top of band.
- Hiring volume. Are they posting 3 DS roles or 30? Thirty means they will move on the next candidate. Three means you have leverage.
- Recent product launches. A company that just shipped an AI feature needs people who can ship more. Time your counter accordingly.
- Open source / GitHub activity. Engineering-heavy orgs that publish models or libraries pay engineering bands, not analytics bands.
Pull the company brief before you respond to the recruiter. Funding stage, hiring temperature, layoff signals, and recent news from SEC, WARN, GitHub, TechCrunch, and YC — anchored to the company across the table from you. Walking into a negotiation without this is walking in blind.
Move 3: Calculate your counter, three ways
You will give the recruiter one number. You will arrive at it by computing three and picking the right one for the situation.
The base-salary counter. Anchor at $118,000–$122,000 if the role is tech, U.S. metro, mid-level. Adjust up for senior (+15–25%), down for early-career (−15%), up for high-cost metros (SF, NYC, Seattle: +10–15%), down for fully remote with no metro adjustment (−5–10%). If the offer base is at or below this anchor, counter base. If the offer base is well above, leave base alone and counter the equity or sign-on.
The total-comp counter. Anchor at $176,250 for a standard mid-level tech DS. Compute the offer's total comp over the equity vesting period (typically 4 years). RSUs at private companies need a haircut — use last-round valuation, not the strike-price fantasy. If TC is below the Levels anchor by more than 10%, you have room to push on equity and sign-on.
The AI-premium counter. Only if you can defend it with shipped work. The 56% premium attaches to demonstrable AI/ML production experience³. If you have it, anchor against the Robert Half AI/ML Engineer mid-band at $170,750⁵ for base, and target $220,000–$260,000 TC at a Series C+ tech company. If you don't have it, do not invoke it. Recruiters can tell.
Pick the counter that matches your leverage. Then add 8–12% to the number you actually want. They will negotiate you down. Plan for it.
Move 4: Frame the counter so they can say yes
The number is half the work. How you frame it is the other half. Three rules.
Specific dollar amounts, not ranges. "I was thinking $155,000 base, $40,000 sign-on, and the senior title" lands. "Somewhere in the 150s" leaks.
Anchor to market data, not to your needs. "Levels.fyi median total comp for this role is $176,250; my counter brings the package to $172,000 TC over four years" is a negotiation. "I have student loans" is a request for charity.
Give the recruiter a path. Recruiters have approval bands. If you ask for one thing they can't grant, you lose. If you ask for three things and rank them, they can come back with two-of-three and feel like they won. Rank: base, then sign-on, then equity, then title. Most counters resolve on base + sign-on.
Move 5: Counter the counter-objection
They will say one of four things. Have a response for each.
"This is at the top of our band." Translation: it isn't, but we hope you'll believe it. Response: "I understand bands exist. The Levels.fyi median for this role and metro is $176,250 total comp. Can you check whether there's flexibility on sign-on or first-year equity refresh?"
"We can't move on base, but we can do more equity." Translation: cash is constrained, equity is cheap to us. Response depends on the company. Public company with liquid stock — accept and negotiate vest acceleration. Private company pre-Series C — push back, take cash. Equity at most startups is a lottery ticket priced as currency.
"Our offer reflects the current market." Translation: we're hoping you don't have other data. Response: "The current market for this specific role profile is the Robert Half 2026 mid-band at $170,750 for AI/ML-adjacent work. My counter sits inside that band."
"We need an answer by Friday." Translation: we want to prevent you from running a process. Response: "I'd like to give you a real yes. I need 5 business days to align on the package." Almost always granted. If not, the company has shown you something about how they treat employees.
The loadout you need before the call
Before the call, you need five things. The grade — is this offer actually competitive against the market for your specific role and metro? The intel — what does the counter look like with real numbers attached? The war room — what's the script, what are the objections, what's the read on the counterparty? The scout — what other roles should you be running in parallel to manufacture leverage? The case files — what wins, metrics, and callouts from your last three years justify the number you're about to ask for?
AMMO has all of them. Seven instruments. One pocket.
The most consequential one for this moment is the War Room. Three questions in. A negotiation script out — including the counter-objection bank and the read on the company across the table from you. That's the difference between negotiating and asking.
The counter-view, because the numbers don't lie evenly
Here is what the bullish takes leave out. Tech salary growth in 2026 is 1.6% YoY — that's deceleration, not growth. Analytics-DS postings at Series A/B startups are down 15–20%. The 56% AI premium³ applies to a narrow cohort — not the median data scientist. If you're negotiating from an analytics-DS resume into an analytics-DS role, you are in a soft market and pushing too hard will cost you the offer.
The 66% of people who negotiate succeed¹ is real. The 30% who don't ask is also real. But "succeed" in 2026 sometimes means moving from $138,000 to $144,000, not from $138,000 to $175,000. Know the market, know your leverage, and counter with a number you can defend in one sentence.
What "fair counter" actually looks like in 2026
For a mid-level data scientist at a Series C+ tech company, U.S. metro, with 2–4 years of experience and some production ML work:
- Base: $145,000–$165,000
- Sign-on: $20,000–$40,000
- Equity: $80,000–$140,000/year (vesting)
- Target bonus: 10–15%
- TC: $185,000–$240,000
For an analytics-titled DS at a non-tech enterprise, 2–4 years, U.S. metro:
- Base: $125,000–$145,000
- Sign-on: $10,000–$20,000
- No equity, or de minimis RSU
- Target bonus: 8–12%
- TC: $140,000–$170,000
For a senior data scientist with shipped production ML at a FAANG-tier employer:
- Base: $185,000–$215,000
- Sign-on: $50,000–$100,000
- Equity: $200,000–$400,000/year
- TC: $400,000–$650,000
These are anchors. Your offer either clears them or it doesn't. If it doesn't, you have a counter to make.
Before you send the email
Stop reading. Grade the offer first.
Grade your offer free — paste the numbers, get the verdict in 60 seconds.
Compare two offers side-by-side if you're running parallel processes.
Come to the table loaded.
¹ U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Data Scientists, May 2024 OEWS data, published August 2025. https://www.bls.gov/ooh/math/data-scientists.htm
² Levels.fyi, Data Scientist Salary, May 2026. https://www.levels.fyi/t/data-scientist
³ PwC 2025 Global AI Jobs Barometer, via The 56% Premium: What AI Skills Actually Pay in 2026, March 2026. https://letsdatascience.com/blog/the-56-premium-what-ai-skills-actually-pay-in-2026
⁴ Cadence / Remote.ai, Data Scientist Salary 2026, May 2026. https://cadence.withremote.ai/blog/data-scientist-salary-2026
⁵ Robert Half 2026 Salary Guide via Pin, AI Compensation Benchmarks 2026: The AI Hiring Bubble, May 2026. https://www.pin.com/blog/ai-compensation-salary-guide/