The Two-Offer Choice, Scored on Locked Weights to Take, Close, or Probe

You have two written offers, each with a deadline, and you need to choose one without letting the bigger number or the friendlier recruiter choose for you. This guide is for candidates in that position. It gives you a scoring framework that locks your weights before you compare the offers, grades the evidence behind every score, checks the two things new hires most often regret, and ends in one of three answers: Take, Close, or Probe.

What does the two-offer framework decide?

It decides which offer to sign, or which single move you have to make before you can sign. Every run ends in one of three outputs.

• Take. One offer leads on evidence you can name and passes both floor checks. You sign it.
• Close. The two offers are within your tie band and the evidence behind both is solid. You go back to the employer you prefer and ask them to close one named gap, with a specific number.
• Probe. The lead rests on thin evidence. You book one fact-finding call, dated before the earlier deadline, about the dimension that is carrying the result.

Most advice on how to choose between two job offers gives you the same tool. You list factors, pick weights, and score each job. One typical ranking page tells you to "Score each job on each factor (1 - 10, where 10 is perfect)." The weakness is the order you do things in. If you pick the weights after seeing both offers, the offer you already like gets to set its own weights. A user on Wall Street Oasis described what happens next: "Then I tweaked the weights to see how the ranks changed." That is not deciding. It is searching for a set of weights that agrees with the answer you started with.

The framework takes its structure from the Mediating Assessments Protocol (MAP), a decision method from Kahneman, Lovallo and Sibony. MAP has three parts: define the assessments in advance, make each one from facts and separately from the others, and form the overall judgment only when every assessment is finished. Kahneman summed up the point: "One of the essential purposes of MAP is basically to delay intuition."

If you set the weights after reading both offers, the offer you already prefer has written its own scorecard.

Where this framework departs from MAP

The MAP authors write: "Our advice is not to assign explicit weights to assessments." I keep explicit weights anyway, and the reason is specific. The failure this guide guards against is one person quietly adjusting numbers after the fact, and a dated, locked weight sheet stops that. One practitioner source makes the same argument for groups: "agreeing the weights before anyone scores is what stops a team quietly backing its favourite." If you would rather follow MAP strictly, keep everything else here: dimensions defined in advance, each dimension scored separately, evidence grades, and floor checks. Then swap the weighted total for an overall judgment made at the end.

Which dimensions should you score two offers on?

Score five dimensions that you choose before you compare. Two of them, workload and the manager, also work as floor checks: a floor check is a pass/fail test that can rule an offer out whatever its total. If you want to compare two job offers beyond salary, your dimensions have to include what offer letters leave out.

The regret data shows why those two get floors.

In ZipRecruiter's survey of new hires, excessive workload (11.7%) and difficult managers (10.1%) were the top reasons people regretted taking their current job. In the same survey, 69.2% said they had not felt new-job regret, so about 30.8% had. When the same survey asked why people left their previous job, they gave poor management (23.2%) and stress from workload (19.9%). So the reasons people leave a job are the same reasons they regret the next one. Neither shows up in an offer letter. Candidates move on the facts they can check and get caught by the ones they could not. A survey by The Muse found that 72% of respondents had started a job that turned out very different from what they had been told.

You may see a one-year regret figure quoted, often 23%. I could not trace it to an original study, so treat it as not established.

You can rename or replace a dimension if something else really matters to you, such as visa sponsorship or the start date. Make the change before you compare, though, not after. Keep the list to five or six. Every extra dimension is one more place where a vague preference can creep in.

The weights must add up to 100. Do not take them from a statistic. In particular, Gallup's manager finding does not mean the manager should get 70 points (more on this below). Weights describe what you care about. Evidence comes later, in the scores.

Locked on: [date and time]
Second offer details received: [date, fill in later]

Dimension             Weight   Floor rule
Total compensation    __       none
Manager               __       must pass on a named report's account
Workload              __       must pass on at least one peer's account
Role scope            __       none
Location/schedule     __       none
Total                 100

Comp-tie threshold:   __% of the lower offer's annual total
Close band:           weighted totals within __ points (1 to 5 scale)
Evidence pull:        in writing 1.0 / heard from the team 0.75 / inferred 0.25

How do you grade the evidence behind each score?

Tag every score with one of three grades: in writing, heard from the team, or inferred. The grade decides how far the score is pulled toward the midpoint of 3. Strong evidence keeps its score, and weak evidence hardly moves the total.

Score each dimension from 1 to 5, where 3 means neutral or unknown. Then adjust it:

Adjusted score = 3 + pull x (raw score - 3)

The pull is 1.0 for in writing, 0.75 for heard from the team, and 0.25 for inferred. Suppose you rate the manager at Offer B a 5 because the recruiter was warm and the reviews look good. That evidence is inferred, so the adjusted score is 3 + 0.25 x 2 = 3.5. If a former report tells you by name the same thing, it becomes 3 + 0.75 x 2 = 4.5. The pull values are my own convention, not a research result. What matters is that they are fixed in advance and the same for both offers.

What each grade proves, and when it lies

• In writing proves a term exists as stated. It lies when the document confirms who but not how good. A manager's name in the letter is graded in writing for the reporting line only, never for quality. It also lies when a bonus is described as "eligible" without any payout history behind it.
• Heard from the team proves that someone with direct experience described the job that way. It lies when the hiring manager picked the only peer you spoke to, or when the person you spoke to has never worked under the manager in question.
• Inferred proves nothing on its own. It is what is left after a guess has been labelled honestly.

An offer letter covers fewer of these facts than most candidates assume. Standard templates list start date, work location, reporting manager, department, base salary, bonus eligibility, and a deadline. SHRM warns that letters are often "unclear about compensation terms, including bonuses, and whether the employee will be eligible to receive the bonus."

Role scope is usually inferred until you have spoken to the team. If you have run both postings through Refolk, its fit score measures your history against each role the same way, so you are not holding two different gut feelings side by side. It is still an inference, because a posting is not the job, so tag it inferred and let the shrink rule handle it.

When is a compensation gap a tie?

A comp gap is a tie when the normalized annual gap is smaller than the threshold you wrote down before comparing. When it is, both offers get the same comp score, whichever is larger. This is the single most useful rule for anyone weighing higher pay against a better manager.

The threshold is a convention, not a finding. The pages that rank for this question disagree, and none of them says where its number comes from.

A range from 3% to 15% with no stated basis is exactly the kind of setting the favoured offer can quietly move. So pick a number, write it on the locked sheet, and treat it as fixed. If you have no view, about 5% is the most common choice. Just know that its origin is not established anywhere public.

Normalize before you measure the gap

1. Start with base salary from the letter.
2. Price the bonus at its real payout history, not its target. If you cannot get the history or the plan document, the bonus figure is inferred, so count it at a discount or leave it out.
3. Value equity at what vests in the first year.
4. Add these up to one annual figure per offer, and express the gap as a percentage of the lower figure.

Score the comp dimension from the gap

• Gap below your threshold: both offers score 3.
• Gap between the threshold and twice the threshold: the higher offer scores 4 and the lower scores 3.
• Gap above twice the threshold: the higher offer scores 5 and the lower scores 3.

So a 4% gap with a 5% threshold produces equal comp scores, and the decision moves to the dimensions that predict regret. That is the intended effect.

How do you check the manager and workload floors?

A floor passes when a named person has described that dimension and nothing they said crosses the line you set in advance. The manager floor needs at least one account from someone who reports, or used to report, to that manager. The workload floor needs at least one peer, not just the hiring manager.

Gallup's research is why the manager gets a floor. In Beck and Harter's wording, "Managers account for at least 70% of the variance in employee engagement scores across business units." Gallup also reports that employees supervised by highly engaged managers are 59% more likely to be engaged themselves. Be precise about what this shows. It shows that the manager is the largest single driver of differences in engagement between teams. It does not show anything about your individual pay, promotion, or career path, and it does not justify a 70% weight. As one explainer puts it, the research captures variance in engagement scores, not absolute levels. It is a good reason to refuse a bad manager. It is not a reason to let the manager decide everything.

Floor outcomes

• Pass: a named person's account describes workload or management you could live with for the length of time you plan to stay.
• Fail: a named person's account crosses a line you wrote down in advance, such as regular weekend on-call, a manager with frequent turnover among their reports, or a team that is short-staffed with no plan to hire.
• Unknown: you have only inferred evidence. An unknown floor cannot produce Take. It produces Probe.

The manager's reports are a short list

Getting a former report on the phone sounds hard. The numbers say it usually is not.

Engineering managers are a small group and easy to find. Any one manager's current and former reports run to dozens of people, not thousands. You can find them before a deadline by searching for the employer, a reporting relationship, and a date window, and Refolk's index will surface current and former engineers by employer and title in a single query. Contact two or three of them. Ask the same questions at both employers so the answers can be compared:

• How many hours does a normal week actually take, and what does a bad week look like?
• How often does on-call or after-hours work happen, and is it shared fairly?
• When something goes wrong, what does this manager do in the first hour?
• Who has left this manager's team, and why?
• Would you work for this manager again?

Step-by-step procedure

Run these seven steps in order. Most of the time goes on steps 4 and 5, so start them as soon as both deadlines are confirmed.

Typical effort: about 30 minutes to lock the sheet, an hour to normalize comp, one to two days of back-and-forth with recruiters for written facts, two to three days of team conversations, and an hour to score. Some ranking pages put negotiation before the matrix is finished. I leave Close until the end, because you cannot name the right gap until the scores show which gap matters.

How do you read the result: Take, Close, or Probe?

Apply the floor rules first. Then look at two things together: how far apart the weighted totals are, and how strong the evidence behind that gap is. The answer comes from the combination, never from the total alone.

Decision rules, in order

1. A failed floor rules the offer out. If one offer fails a floor, the other wins as long as it passes. If both fail, the framework has no winner. Probe the offer that failed by less, or keep searching.
2. An unknown floor means Probe. You cannot take an offer whose workload or manager you have only guessed at.
3. Take when the totals differ by more than your close band and the lead still holds after you set its single biggest inferred score to 3.
4. Close when the totals are within the close band and the main scores on both sides are graded in writing or heard from the team.
5. Probe when the lead disappears once you neutralize one inferred score. That dimension is the subject of your call.

The close band works like the comp threshold: a number you lock in advance. Mine is 0.3 points on the 1 to 5 scale. Use it if you have no view of your own, but treat it as a convention.

Why Close works when you hold two offers

When you hold a second offer, negotiating is close to a sure thing. Across survey periods, most new hires do not negotiate, yet most of those who do come away with something.

ZipRecruiter also found that negotiators had a median of 2 offers, against 1 for those who did not negotiate. That is the clearest sign that a second offer is what makes asking pay. Results vary, though. In Pew's sample of 5,188 US workers, 28% of those who asked got exactly what they requested, 38% got more than the first offer but less than they asked for, and 35% got only what was first offered. In ZipRecruiter's data, 69% of negotiators got better base pay. These rates move from one survey wave to the next, so check the latest ZipRecruiter Survey of New Hires before relying on a single figure.

Hi [recruiter name],

Thank you again for the offer for [role]. I want to join [company], and I am ready to sign this week.

I have another written offer with a deadline of [date]. The one thing standing between me and a yes is [the gap, e.g. base salary / start date / remote days]. If you can move [term] to [specific number or condition], I will sign by [date before the earlier deadline].

Can we confirm that in writing by [date]?

Thanks,
[your name]

How does the framework go wrong?

It goes wrong when a step looks finished but the evidence behind it is weaker than the tag claims. These are the seven failure modes I check for every time, each with its test.

1. Weights locked after a recruiter call. The pre-commitment is tainted before you start. Check: the sheet is dated before the second offer's details arrived.
2. A comp gap under the threshold still decides the result. For example, a 4% gap scored 9 against 6 instead of being called a tie. Check: any gap under the locked threshold produces equal comp scores.
3. The manager's name in the letter graded as evidence of quality. The letter tells you who, not how good. Check: an in-writing tag on the manager dimension covers only the reporting line.
4. The bonus counted at target. SHRM documents disputes over bonus eligibility. Check: you have the payout history or the plan document, or the bonus is tagged inferred.
5. The workload floor passed on the hiring manager's word alone. The hiring manager has a reason to describe the pace kindly. Check: at least one peer conversation, tagged heard from the team.
6. Close used without leverage. Only 69% of negotiators got better base pay, and a vague request rarely moves anything. Check: you have named one gap and the exact number you will ask for.
7. Probe booked after the earlier deadline. A call after the deadline is not a probe. Check: the call date falls before the earlier of the two deadlines.

Signals that look strong but are not

• The friendlier recruiter. Warmth during hiring tells you about the recruiter, not your future manager. Grade it inferred.
• A peer the manager chose for you. That conversation is useful, but it is one hand-picked account. Look for one more person, ideally a former report, before you grade the manager floor as passed.
• The bigger headline number. Before normalizing, a target bonus or a four-year equity figure can make a smaller offer look larger. Compare only normalized annual totals.
• A score you adjusted to fit how you feel. If you notice yourself wanting to change a raw score after seeing the total, stop. That urge is the intuition MAP exists to delay.

Before you sign, and how to keep the sheet current

Before you sign anything, every item below should be true. If one is false, your answer is Probe, even if the totals look settled.

Keep the filled-in sheet. Over the first month, most new hires settle whether the job was the right fit, and that is when you can check your evidence grades against reality. Note which heard-from-the-team accounts held up and which inferred scores were wrong. Next time you hold two offers, adjust your pull values and close band based on that record, before the first offer arrives, and date the new sheet. The published numbers this guide relies on, such as negotiation rates and comp-tie conventions, change from source to source and from survey wave to survey wave. Check them when you lock the sheet, not when you are scoring.