Most succession plans get built the same way: a board meeting is coming, someone asks "what's our succession plan?", and HR produces a spreadsheet over a long weekend. Names go in boxes. The deck gets presented. Everyone exhales.
Then a VP resigns in March and nobody opens the spreadsheet, because everyone already knows it was a guess.
If that describes your organization, you have company: a SHRM survey found only 21% of organizations have a formal succession plan, another 24% run an informal one, and 56% have nothing at all.
This guide is about building the other kind of plan: one with criteria you can defend, readiness calls you actually believe, and a review rhythm that keeps it matched to the org as people move. None of it requires software, though the last section covers where software earns its keep.
Start with the roles that would actually hurt
The first mistake is scope. A plan that tries to cover every manager becomes a data-entry project, stalls around the 60% mark, and dies with its credibility. A plan that covers only the CEO misses where most of the damage actually happens.
The filter is cost of vacancy, not seniority. Ask three questions about each role:
- If this seat went empty tomorrow, what stops moving? Revenue, releases, regulatory filings, a key account relationship?
- How long would an external search take, realistically? Six months for a niche VP is normal. What breaks in those six months?
- Does critical knowledge live only in this person's head? A principal engineer who owns the billing system can out-rank a VP on this list.
For a 300 to 1,000 person company, this usually produces 15 to 30 roles: the executive team, a slice of the level below it, and a handful of single points of failure that no org chart flags. Write down why each role made the list. That sentence becomes useful every time someone asks why their role isn't on it.
Decide what ready means before naming names
Here is where most plans quietly go wrong. If you ask a room of leaders "who could take over for Maria?", you will get names in thirty seconds. What you won't get is agreement on what the names mean. One leader is answering "who could survive the first month," another is answering "who deserves it," and a third is answering "who I'd pick if I had to decide today."
So before any names: define what success in the role requires. Keep it to five to eight dimensions per role. A workable set for a sales VP might be: builds and retains a leadership team, forecasts within a defensible range, handles board exposure, manages cross-functional conflict, and knows the market cold. For a head of engineering the list looks different, and it should.
Two properties matter more than the specific labels:
- Each dimension is observable. "Executive presence" is a horoscope. "Has run a QBR in front of customers without an executive in the room" is evidence.
- The dimensions are weighted. Every role needs judgment; not every role needs board exposure. If everything counts equally, the scores flatten into mush.
This step feels bureaucratic and slow. It is also the entire difference between a succession plan and a popularity list. Budget a working session per role family and get the incumbent's input, because they know which parts of the job the org can't see.
Score candidates against the role, not each other
With a profile in hand, identify candidates. Cast wider than instinct suggests: the obvious successor is sometimes the incumbent's mirror image, and the role may need the opposite. Peers in adjacent functions, strong leaders a level down, and the person running the biggest cross-functional program all belong on the first-pass list.
Then score each candidate against each dimension of the role profile. Not against each other. Ranking people head-to-head produces a horse race and a lot of politics; scoring against fixed criteria produces gaps you can act on. "Elena is a 2 of 4 on board exposure" leads directly to "get Elena in front of the board twice this year." A ranking leads nowhere except to a bruised ego when it leaks.
Practical rules that keep scoring honest:
- Use a small scale. Four points is plenty. Ten-point scales generate false precision and hour-long debates about the difference between a 6 and a 7.
- Require a sentence of evidence per score. Scores without receipts revert to reputation.
- Record who scored, and when. Assessments age. A score from two years ago describes a different person.
- Calibrate across raters before you trust cross-team comparisons. One manager's 3 is another's 4; a calibration pass in a talent review is how you find out.
Put readiness windows on every candidate
A list of scored candidates still isn't a plan. The plan emerges when you translate scores into readiness: who could take the role now, who needs about a year, who is two to three years out.
The standard windows are Ready Now, Ready in 1 Year, and Ready in 2 to 3 Years, and the discipline is refusing to inflate them. A Ready Now call means this: if the incumbent resigned this afternoon, you would give this person the job rather than open an external search. If the honest answer is "we'd probably run a search anyway," they are not Ready Now, and pretending otherwise is how boards lose faith in the whole exercise.
Seen across all critical roles, the windows become your bench strength picture: which roles have real cover today, which have cover coming, and which are exposed. Expect the first pass to be uncomfortable, and know that the discomfort is normal: in DDI's Global Leadership Forecast 2025, only 20% of HR leaders reported having leaders ready now for their most critical roles, and internal candidates could immediately fill just 49% of critical positions. A first-draft plan showing half your critical roles with no Ready Now cover is not a failed plan. It is an accurate one, and accuracy is what makes the next step worth funding.
Close the gap with development that has a deadline
Every scored gap implies an action, and this is where plans separate into the two kinds. The failing kind writes "develop strategic thinking" next to a name and moves on. The working kind writes: "Owns the pricing-model overhaul in Q3. Presents the result at the January board meeting. Readiness review in December."
The pattern is exposure plus deadline. People close readiness gaps by doing pieces of the bigger job under real conditions: running the budget cycle, holding the room in an escalation, carrying an initiative that crosses functions. Courses and coaching support that work; they don't replace it.
Put a date on every development action and a name next to every date. Undated development intentions have roughly the shelf life of New Year's resolutions, and for the same reason.
Review on a cadence the calendar enforces
Plans decay. People get promoted, resign, burn out, and change their minds about what they want. A plan reviewed annually is wrong for most of the year, and everyone senses it, which is why nobody consults it when a real vacancy hits.
The cadence that works in practice:
- Quarterly, 30 minutes, per executive-team role: did anything change? Any flight risk? Development actions on track?
- Twice yearly, half a day, full plan: rescore where there's new evidence, recalibrate across leaders, update readiness windows, and retire stale actions.
- Immediately, on trigger events: a resignation, a reorg, or a candidate leaving forces a review of every plan that person touched. This is where a stress test earns its place: model the departure, watch which other plans go thin, and fix the cascade before it happens for real.
The reviews only happen if they are on a calendar with an owner. "We review continuously" means never.
What a finished plan looks like
Finished is the wrong word, but a plan that works has a recognizable shape:
| Component | Failing version | Working version | | --- | --- | --- | | Scope | Every manager, or only the CEO | 15 to 30 roles chosen by cost of vacancy | | Criteria | "Gut feel," refreshed annually | Weighted, observable dimensions per role | | Scores | Unattributed numbers | Evidence-backed, dated, calibrated | | Readiness | Everyone is "ready with support" | Honest windows someone would act on | | Development | Themes without dates | Exposure assignments with deadlines | | Maintenance | Rebuilt before each board meeting | Quarterly touch, twice-yearly deep review |
On tooling: a spreadsheet can genuinely hold all of this for a while. What a spreadsheet can't do is remember why a score changed, show you bench depth across thirty roles at a glance, or model what happens to four other plans when one person leaves. When the review meetings start spending more time reconstructing history than making decisions, that is the moment purpose-built succession planning software stops being a luxury.
But the order matters: criteria first, honest scores second, software third. A tool pointed at a popularity list just makes the popularity list load faster.