Why ops hires fail in the first ninety days

An ops hire who leaves or is managed out inside ninety days is rarely a capability problem. The failure modes are boringly consistent, almost all of them are set before the person starts, and most of them are visible weeks before anyone acts.

This post is a diagnostic rather than a list. It goes in the order you will actually encounter it: what the failure looks like at week 4, week 8 and week 12, which failures trace back to the advert rather than to the person, and what to fix in each case.

The short answers

• The mandate is the failure. Three people hold different ideas of what the role owns, nobody wrote it down, and the new hire discovers the disagreement one meeting at a time. Then they are blamed for lack of focus.
• Most of it is decided during hiring. Of the eight failure modes below, six are set before the start date and four are set before the advert goes out.
• The week-4 tell is calendar-full and output-empty. The week-8 tell is a deliverable that keeps moving. The week-12 tell is a quiet job search.
• Access is not an IT detail. Operations work that has to be requested through someone else does not happen, and the request queue becomes the job.
• The first project should be small, visible and finishable. Pointing a new hire at the biggest political problem spends credibility they have not earned.
• Levelling down is a slow failure, not a saving. The research on perceived overqualification finds autonomy and empowerment to be the substantial moderators, which is a design instruction about room rather than an argument for hiring cheap.
• A diagnosis is legitimate first-quarter output. A metric target in month one pushes someone to act before they understand.
• Ask whether it is a system problem or a behaviour problem. If it is behaviour, no hire fixes it and the person you hire will resent being asked to.

The ninety days, as they actually run

	What a working hire is doing	What a failing hire is doing	What to do about it
Week 4	Has named the first deliverable and who signs it off	In every meeting, owns no deliverable, describes discovery	Re-agree the mandate in writing, this week
Week 8	Something small has shipped; definitions are being argued about in public	The deliverable has moved twice; access still pending	Cut scope to something finishable; escalate access
Week 12	A written diagnosis exists and has been challenged by someone senior	Polite, responsive, no longer proposing anything	Ask what they would change if they could. Then act on it

The last cell is the one that matters. The failure almost never announces itself. It looks like a competent person becoming gradually less ambitious, and by the time it is obvious the useful intervention window has closed.

Where each failure actually starts

Failure	Where it starts	First visible
Unwritten mandate	Before the advert	Week 3 to 4
Hired for the wrong role	Before the advert	Week 4 to 6
Levelled down	At the band decision	Month 4 to 6
A people problem sold as a system problem	Before the advert	Week 3
No access	At the offer	Week 1
First project too big	At the start date	Week 6 to 8
Nobody described the current state	Day one	Week 2
Reporting line cannot arbitrate	At the requisition	Week 8 onward

Six of the eight are decided before the person starts. That is the argument of this whole post: ninety-day failure is a hiring artefact with a delayed onset, and the fixes belong to the hiring process rather than to onboarding.

1. The mandate was never written down

The commonest one by a distance. Three people in the business have a different idea of what the role owns, everyone assumed the others agreed, and the new hire spends their first six weeks discovering the disagreement one meeting at a time.

They are then blamed for lack of focus.

What makes this specific to operations is that the role sits between functions by construction. A sales hire who is confused about their remit has one manager to ask. A RevOps hire has three stakeholders with legitimate and incompatible views about what should happen first, and no standing authority to choose between them. Ambiguity that a functional hire could resolve in a one-to-one is, for this role, a negotiation.

The fix: one page, written before the advert, that says what the role is accountable for and what it is not. Circulate it to everyone who will have an opinion later. Their objections are cheaper now than in month two. An objection in week one is a scoping correction, and the same objection in month two is a conversation about a person.

The tell it has already happened: you ask two stakeholders separately what the new hire should have finished by week eight and get two different answers.

2. They were hired for the adjacent role

Sales ops, RevOps, marketing ops and GTM engineering solve different problems, and the titles are used interchangeably. A search that never resolved which one it wanted produces a competent person in a mandate that does not fit the problem, and both sides conclude the other misrepresented the job.

Nobody was lying. The brief was ambiguous and the interviews did not resolve it.

This surfaces later than the others, usually around week 4 to 6, when the first real piece of work makes the mismatch concrete: the person you hired to fix attribution turns out to have been hired by someone who wanted the forecast fixed.

The fix: resolve it at the brief, using the problem rather than the title. The role comparison is in GTM engineer vs RevOps manager and the scoping step is step 2 of the hiring process.

3. No access

Operations work that has to be requested through somebody else does not happen. A RevOps hire without administrator rights on the CRM and read access to the warehouse spends their quarter filing tickets and building relationships with the people who can action them, which is a real skill and is not the job you hired for.

The compounding cost is that it is invisible in a status update. “Waiting on access” appears once, politely, and then stops appearing while the person quietly rebuilds their plan around what they can reach.

The fix: access on day one, agreed during the offer conversation rather than discovered in week two. Where a security policy genuinely prevents day-one administrator rights, which is common and reasonable in regulated sectors, name the date it arrives and plan the first six weeks around not having it.

It is worth being honest that access is not free to grant. FoundHQ’s analysis of Yelp’s Salesforce team describes 3,000 users supported by a 30-person core team containing five administrators, against the commonly quoted enterprise practice of seven to ten administrators per 1,000 users. Access control is a real discipline. The point is that it is a decision to make deliberately at the offer, not a queue to discover afterwards.

4. The first project was the biggest problem

A new hire pointed at the hardest, most political problem in the business will spend their credibility before they have earned any. The attribution model, the territory redesign, the CRM migration: all of these require agreement from people who have no reason yet to trust the person asking.

The fix: pick something small, visible and finishable in six weeks. The point of the first project is not the project. It is producing evidence that this person makes things better, which is the currency every subsequent argument will be paid in.

Three properties of a good first project: one owner, so it cannot be blocked by committee; a countable before and after, so the result is not a matter of opinion; and an audience beyond the hiring manager, so the evidence lands where it is needed.

5. Nobody told them what the numbers currently are

An ops hire who cannot get a straight answer to “what is our conversion rate from stage 3 to stage 4, and do you trust it” is being asked to improve a system nobody will describe. Often the honest answer is that nobody knows, which is fine and needs saying out loud on day one rather than being discovered in week three.

The scale of that is easy to underestimate. Gartner’s survey work reports fewer than half of sales leaders and sellers holding high confidence in forecast accuracy and a similar share believing their data quality is high. That is a release that discloses no sample size or fieldwork dates, so read it as an indication of how common the complaint is rather than as a measurement.

The publisher-level disagreement is itself instructive. Salesforce’s productivity research, across 7,775 sales professionals with fieldwork in late 2022, reported reps spending 28% of the week actually selling across an average of ten platforms per deal. Its later State of Sales survey, across 4,050 professionals in 22 countries, reported 40% of time selling, 17% on data entry alone, across an average of eight tools. Same publisher, different instruments, materially different answers. Nobody knows this number to a decimal place, including the people selling you software to improve it.

The fix: on day one, hand over the current state including the parts that embarrass you. The bad news is the job. A new hire who finds out in week three that the number they were given in week one was wrong has learned something about the company rather than about the data.

6. They were levelled down

This is the slowest of the failures and the one most often mistaken for a good deal at the time. The band was set to what was affordable rather than to the judgement the role needs, and the person hired into it is capable of considerably more than the seat allows.

The research is more specific than the folk version of it. The Annual Review synthesis of work on perceived overqualification finds it consistently associated with lower job satisfaction and with turnover, and identifies autonomy and empowerment as substantial moderators. The same person, in the same job, with room to act, does not show the same pattern.

Two things follow, and they point in opposite directions to the usual conclusion.

It is not an argument for hiring less capable people. A strong operator with latitude is a good outcome and a common one.

It is an argument about room, not about level. A first ops hire has nobody to ask, so what you are buying is judgement under ambiguity. Buying that and then routing every change through an approval queue is the specific combination the research warns about, and it produces a resignation in month five rather than a performance problem in month two.

That connects directly to the levelling argument in what a first ops hire should look like: the seniority question and the autonomy question are the same question, and answering the first without the second is how a well-paid hire still fails.

7. They were hired to fix a people problem with a system

This one is uncomfortable. A meaningful share of “our data is a mess” briefs are actually “our sales leader will not enforce process”. Systems do not fix that, and an ops hire who works out on week three that they were brought in as a proxy for a conversation nobody wants to have will start looking.

The fix: ask honestly, before the hire, what happens today when someone ignores the process. If the answer is nothing, no system changes that. If the answer is that a manager has a conversation, the system is worth fixing.

The tell: every proposal the new hire makes is accepted, and none of them is enforced.

8. The reporting line cannot settle anything

A role whose function is to arbitrate between sales and marketing, placed inside sales or marketing, will lose every arbitration that matters. It will not lose them loudly. It will lose them by having its recommendations quietly deprioritised whenever they cost the parent function something.

Forrester reported centralised revenue operations rising from 15% to 40% of organisations across two years, across 316 respondents, which measures centralisation rather than reporting lines, and is a description of what companies are doing rather than evidence that it works better. ICONIQ’s 2025 survey of 205 B2B software go-to-market executives puts revenue operations at about 6% of go-to-market headcount, rising to 7% at high-growth companies, which tells you the function is usually small enough that its reporting line determines its influence entirely.

The fix: the role reports to whoever can settle a disagreement between two functions without escalating. At most companies that is a revenue leader or a COO. If it has to sit inside a function, give it a standing forum where the other function is present and a written escalation path, and expect to use both.

What good looks like at ninety days

Not a transformed function. Realistically:

• A written diagnosis of what is broken and in what order it should be fixed
• One visible thing shipped and working
• Definitions agreed for the three or four metrics people argue about
• A relationship with sales and marketing leadership that is not purely reactive

Anyone promising more than that in a first quarter is either exceptional or telling you what you want to hear, and the base rate favours the second.

Note what is not on that list. Not a rebuilt CRM, not a new attribution model, not a forecast that is suddenly accurate. Those are second- and third-quarter outputs, and a company that expects them in the first will read a well-run first quarter as underperformance.

The onboarding plan that prevents most of this

Twelve weeks, four artefacts, three checkpoints. It takes about two hours to write.

Before day one. The one-page mandate from failure 1, circulated and objected to. Access agreed in the offer conversation. The first project chosen and named. The current state written down, including the embarrassing parts.

Week 1 to 2. Introductions with a purpose: every stakeholder is asked the same two questions, what do you think this role owns and what would make this a success for you, and the answers are written down. The disagreements in that document are the real map of the job.

Week 4 checkpoint. The question is not how are you finding it. It is: has anything you have learned changed what this role should own? A mandate correction here is free.

Week 6 to 8. The first project ships. Small, visible, finished.

Week 8 checkpoint. The draft diagnosis, presented to the people whose numbers it concerns. It should be uncomfortable in at least one place; a diagnosis that offends nobody has not looked anywhere interesting.

Week 12 checkpoint. Against the scorecard written before the advert, not against an impression. If there was no scorecard, this checkpoint is a conversation about feelings and it will reach the wrong conclusion.

Telling a hiring failure from an onboarding one

Worth doing honestly, because the two have opposite remedies and the default assumption is almost always the expensive one.

Ask whether a different person would have succeeded in the same seat, under the same conditions, with the same mandate, the same access and the same first project. If the answer is no, it was not a hiring failure, and replacing the person reproduces it at the cost of another search.

The related question is whether the process even measured the right thing. Screening selects hard for capability and almost never for the conditions the person will work under, which is why the failure lands where the attention was not. That is also why the evidence on conventional hiring practice is unflattering: Peter Cappelli’s survey of the field is the standard reference and it is largely a catalogue of processes that measure what is easy to measure.

What it costs

Gallup’s estimate for replacing an employee is one half to two times annual salary, which the firm itself calls conservative. That is a US general-population figure rather than an operations-specific one, so use it for order of magnitude: on a $120,000 US hire, the low end is $60,000 before the lost quarter.

Then add the search. Employ Inc’s 2026 benchmarks, across 6,640 companies, report 63.5 days to fill and 46.2 days to hire on the same hires. The UK market gives you no relief on the second attempt either: vacancies stood at 707,000 in May to July 2026, with about 2.5 unemployed people per vacancy, which produces more applications without producing more of the people you want.

The uncosted part is the one that actually hurts. A first ops hire who fails takes the case for the function with them, and the next request for that headcount is answered with “we tried that”. The scorecard, the mandate page and the access conversation exist to protect that argument, not just the individual.

The part that starts before day one

Every fix above happens during hiring, not after it. A brief that says what the role owns, an honest account of the current state, a first project chosen before the offer goes out, access agreed in the offer conversation, and a band set to the judgement required rather than to what was convenient.

That is also why the Rareix process produces a written scorecard before sourcing starts. It is not a formality. It is the artefact that prevents failure mode one, which is the one that causes most of the others.

Related reading

• How to hire a RevOps manager, for the process that produces the mandate and the scorecard.
• What a first ops hire should look like, for the levelling argument in full.
• GTM engineer vs RevOps manager, if failure mode 2 sounds familiar.
• Why your RevOps job description attracts the wrong people, for where the mismatch is usually introduced.
• RevOps or another AE, if the headcount case is still being argued.
• Work sample tests vs interviews, for what a hiring process can and cannot measure.