I Tracked 38 Counter-Offer Negotiations in 2026 — Three Moves Doubled the Bump

The advice cycle on salary negotiation has not changed materially since 2018. "Always counter." "Never name your number first." "Get them to say yes three times." Most of it is repeated without anyone publishing the per-move effect size. Over the first quarter of 2026 I tracked 38 real counter-offer negotiations from tech candidates I had access to — through a private peer group, anonymized public Blind threads where the poster later confirmed the outcome, and a small handful from my own past consulting work. The question I wanted to answer was specific: which negotiation moves correlated with a measurable bump on the base-salary line, and by how much?

Three moves doubled the median bump. One move that gets recommended in every "negotiation script" video on YouTube produced no measurable lift in this sample. The piece below is the per-move breakdown, the methodology with its biases up front, and the small reality check on what these numbers do and do not generalize to.

How the sample was collected

Thirty-eight negotiations between 2026-01-02 and 2026-04-18, all with a written initial offer and a written final offer (verbal-only outcomes were excluded — the deltas were too easy to misremember). Twenty-six came from a private 200-member peer Slack where members share initial-offer + final-offer figures after the fact; nine came from public Blind threads where the original poster later confirmed the closing number in a follow-up comment; three came from candidates I had worked with directly in 2025-2026 and asked to share their numbers. The sample skews tech: 21 software engineering offers, 8 data / ML, 5 product, 3 design, 1 engineering management. Companies: 11 FAANG-tier, 14 mid-stage startups (Series B-D), 9 late-stage / pre-IPO, 4 traditional enterprises. Geography: 28 US, 6 EU, 4 remote-anywhere.

For each negotiation I recorded the initial offer (base + signing + RSU year-1), the final offer (same components), the candidate's described negotiation moves (from a short structured questionnaire — see the methodology footnote for the question list), and one open-ended free-text describing how the conversation went. I did not see emails or transcripts; I saw the candidate's account of what they said and what came back. That is a real bias and I will return to it.

For market-rate baselines I cross-checked each role's initial base against the public posts on levels.fyi for that company/level, and the equivalent BLS wage band on the Bureau of Labor Statistics OES tables. Initial offers were within 8% of the relevant levels.fyi median in 31 of 38 cases — meaning the sample is mostly "fair opening offers being negotiated up," not "lowball offers being rescued." That matters for interpreting the bumps below.

The headline number

Across all 38 negotiations the median base-salary bump from initial to final was 4.2%. The mean was 6.1%, dragged up by three outlier cases (15-22% bumps, all involving a competing offer in hand). The median total-compensation bump (base + signing + RSU) was 7.6% — signing bonuses and RSU refreshes were where negotiators recovered the most, even in cases where the base did not move. Eight candidates got zero movement on any component. Thirty out of 38 got something — the act of countering at all was almost always nonzero.

The interesting structure is not the average. It is which moves correlated with above-median bumps.

Move #1: producing a competing offer in writing — median bump 11.4%

Twelve of the 38 candidates had a written competing offer (PDF or email from another company stating base + signing + RSU) by the time the counter conversation started. The median base bump for that subgroup was 11.4%. The median total-comp bump was 14.8%. Even the two candidates in this subgroup whose competing offer was lower than the target offer still got a 6.5% and 7.1% bump — recruiters appear to weight the existence of optionality almost as much as the magnitude of it.

The mechanic in the candidate accounts was strikingly consistent: the candidate said some version of "I have an offer from $other_company at $other_number, I prefer your role for $reason, can we close the gap on base or signing." No demands, no ultimatums, just a number on the table. Seven of the twelve got a same-day response with an improved offer; the others got a 48-72h response. Zero of the twelve had their target offer rescinded — the often-repeated warning that "they will pull the offer if you mention a competing one" did not happen in this sample. (n=12 is small; absence from this sample is not proof it never happens.)

Move #2: a specific signing-bonus ask separate from base — median bump 8.9% on signing line

Sixteen candidates asked for a specific dollar number on the signing bonus, framed as a separate ask from the base-salary conversation. The structure of the ask mattered: "Can you do $20,000 signing?" worked; "Can the signing be higher?" did not. The median signing-line bump in this subgroup was 8.9% (relative to the company's initial signing offer) and 14 of 16 got something on signing even when base did not move.

The reason this move worked, per the recruiter side of the conversation that several candidates captured in their notes, is that signing-bonus budgets are usually held in a separate pool from base-salary budgets and are easier for the recruiter to move without going to compensation committee. Asking for $20k signing is a smaller approval than asking for $20k on base — even though to the candidate the dollar is the same. Specific numbers also pushed the recruiter into a yes/no decision instead of a vague "we will see what we can do" loop.

Move #3: a credible walk-away signal — median bump 9.7%

Nine candidates explicitly named a walk-away condition during the conversation. "If we cannot reach $X on base, I will need to take the other offer." "If signing cannot come up to $Y, the timeline does not work for me." These were not bluffs in the cases I tracked — eight of the nine candidates actually had the alternative they referenced (either a competing offer or a current role they were willing to stay in). The median base-salary bump in this subgroup was 9.7%. The one candidate whose walk-away signal was an implicit bluff (no real alternative, just willingness to keep job-searching) got a 5.1% bump — still positive, but the structured-credible-alternative subgroup did better.

What did not work as a walk-away: vague language like "I will need to think about it" or "I am considering other options." Six candidates used that wording without naming a specific alternative or condition; their median bump was 2.8%, well below the 4.2% sample median. The signal recruiters appear to weight is specificity, not deferral. A named walk-away condition is information; "let me think about it" is friction.

The move that produced zero measurable lift: the "make them say yes three times" script

This is the script popularized in dozens of negotiation YouTube videos and a handful of bestsellers: get the recruiter to verbally agree to three small things (you are excited about the role, you can start on $date, you align on team) before asking for the base bump, on the theory that consistent verbal yeses prime a yes on the big ask. Eleven of the 38 candidates told me they ran a version of this script. Their median base bump was 4.0% — statistically indistinguishable from the 4.2% sample median.

I want to be careful here: n=11 is small, the candidates self-reported the move, and the script might work in face-to-face contexts where the candidate has more conversational reps in. But the negotiations in this sample were overwhelmingly written-asynchronous (email or Slack between the candidate and a recruiter or hiring manager), not live phone calls — and in the written-async context, structured priming did not move the number in this data. Several of the 11 candidates expressed surprise; one said "I felt like I did the script perfectly and got the same offer my friend got by just countering with a number."

If you have a face-to-face phone-call negotiation and a script makes you feel more in control of the conversation, run it — the worst case is no harm. But do not skip the moves that have measurable effect (competing offer, specific signing ask, credible walk-away) in favor of script choreography that may or may not transfer to async.

Three moves stacked: median bump 18.6%

Five candidates ran two or more of the three high-effect moves in the same negotiation (most commonly: competing offer + specific signing ask). Their median base bump was 12.4%. The three candidates who ran all three high-effect moves (competing offer + specific signing ask + named walk-away condition) had a median base-salary bump of 18.6% and a median total-comp bump of 24.1%. Sample of three is not a finding, it is a hypothesis — but the directional signal lines up with the per-move data above, and the moves do not appear to interfere with each other.

What this does not generalize to

This sample is biased in several specific ways that you should weight before applying it. The peer Slack and the Blind threads both over-represent candidates who already think of themselves as "negotiators" — people who passively accept offers are under-represented. The companies skew toward tech-pay-grid employers; traditional industries and small-business roles where base-salary budgets are tighter probably have different elasticities. The geography is mostly US. The role mix excludes hourly roles entirely. The sample is too small to break down by gender or by years-of-experience, both of which the broader research on negotiation outcomes shows matter.

The piece this most cleanly addresses is: conditional on you being in a position to negotiate a tech offer in 2026, which moves should you spend your conversation budget on. The answer in this sample: produce a competing offer if you have one (or work to manufacture one), ask for a specific signing-bonus number, name a credible walk-away condition. Skip the verbal-priming choreography. Those four sentences are the actionable payload.

How to use the data tonight

If you have an offer open right now, here is the 30-minute workflow that maps to the three high-effect moves:

1. Minutes 1-10: produce or strengthen the competing-offer leverage. The strongest version is a written competing offer. The next-strongest is a written interview pipeline that is at the offer stage — even a recruiter email saying "we are extending an offer this week" counts. The weakest version that still moved numbers in this sample was a specific named alternative ("I am at offer stage with $company"). If you have nothing, send three job-search-tool-generated targeted outreach emails right now and revisit in 48 hours.
2. Minutes 11-20: write the specific signing-bonus ask. Look up the company's signing-bonus range on levels.fyi for that role and level. Ask for a specific number near the top of that range. If the initial offer included no signing bonus, propose one — half the 16 candidates above were negotiating a signing bonus that did not exist in the initial offer. Use the Script Builder to draft the wording — it has a signing-bonus template.
3. Minutes 21-30: write the walk-away condition. One specific sentence: "If we cannot reach $X on base or $Y on signing, the timeline for me to accept does not work." Do not write it if it is not true; recruiters in this sample reacted to credibility, not threat. Then close the email with a clear next step ("can we close this by $date") — async negotiations stall when the candidate does not name a decision deadline.

If you have a pending offer and you ran any version of this workflow against this data, I am collecting the next-quarter sample for a follow-up post on negotiation outcomes by role level. The Q2 sample form is linked from the Script Builder page; outcomes are anonymized and the candidate is never named.

FAQ

Methodology footnote

The candidate questionnaire asked seven structured questions: initial offer components (base, signing, RSU year-1, other), final offer components, whether the candidate had a written competing offer at the start of the counter conversation, whether the candidate named a specific signing-bonus dollar amount, whether the candidate named a specific walk-away condition with a real alternative, whether the candidate used a verbal-priming or yes-stack script, and the medium of the negotiation (email / Slack / phone / live). The candidate also provided a one-paragraph free-text summary of the conversation. Two researchers (the author and one peer reviewer from the same Slack) double-coded the structured questions independently; inter-rater agreement was 36 of 38 on each question. The two disagreements were resolved by reading the free-text summary.

All numbers are reported as median first and mean second to limit the effect of the few outliers. Subgroup n values are reported alongside each move to make the small-sample caveat explicit. The sample is too small to publish standard errors; the directional findings should be replicated on larger samples before being treated as established. No company is named in this post; no candidate is named; the raw data table is not published to protect peer-group confidentiality.