How do you tier accounts so research time goes where it pays?
Score every account 0 to 3 on fit, signal, deal size and access, remove anything with weak fit, and sort the rest into tiers A, B and C by total score. Tier A gets one-to-one outreach and about 75 minutes per account per 4-week cycle, covering every contact and follow-up; tier B gets a shared message with one custom line and about 15; tier C gets a segment message and about 3. Those budgets are our guidance, not a study.
Updated
Why tier accounts at all?
Because your hours are fixed and your accounts are not equal. In our worked example below, equal effort gives every account about 14 minutes: the 10 strongest get a thin first line, and the 60 weakest soak up time they will never repay.
It answers a different question from how much time to research a prospect, which sets minutes per person. This page decides which accounts earn those minutes in the first place.
It also assumes you already know who belongs on the list. If you don't, build that first: how to build an ideal customer profile.
How do you score an account?
Four criteria, each scored 0 to 3. Our scoring, so adjust the anchors to your market.
| Criterion | What you're asking | 0 | 3 |
|---|---|---|---|
| Fit | How closely does it match your ICP, with no disqualifiers? | Misses the profile | Matches every box and your best wins look like it |
| Signal | Is something changing now? | Nothing visible | A trigger you can name, dated in the last 30 days (our cutoff) |
| Deal size | What could it be worth to you? | Below your minimum | Top of your range, or room to expand |
| Access | Can you reach the people who matter? | No verified contact | Two or three verified contacts, one a warm path |
Three rules keep the scoring honest.
- Fit is a gate, not a score. Fit of 0 or 1 means the account is off the list, whatever else it scores. A hot signal at a bad-fit company is a distraction.
- Signal means an external change you can point to. A new executive, a funding round, a hiring push. Our page on sales trigger events covers six and how to spot them free. If you can't name the event and the date, score it 0.
- Access of 0 caps the account at tier B until you find a verified contact.
Now add the four scores. The maximum is 12.
| Total score | Tier |
|---|---|
| 9 to 12 | A |
| 6 to 8 | B |
| 2 to 5 (fit of 2 or more) | C |
Here is the scoring on four made-up accounts.
| Account (invented) | Fit | Signal | Size | Access | Total | Result |
|---|---|---|---|---|---|---|
| Kestrel Freight | 3 | 3 | 3 | 2 | 11 | Tier A |
| Quarrybank Dental Group | 3 | 0 | 2 | 1 | 6 | Tier B |
| Tidewell Print | 2 | 1 | 1 | 1 | 5 | Tier C |
| Fernhollow Labs | 1 | 3 | 3 | 3 | 10 | Off the list (fit gate) |
Fernhollow scores 10, higher than Quarrybank. It still doesn't make the list, because the fit is wrong. That is the rule doing its job.
Can AI fill in the scores, and how do you check its work?
AI can do the reading, which is the slow part. It cannot do the judging for you. Our guidance: let it score Fit and Signal from pages you give it, and keep a human check on every score that moves an account up.
Give it the account's own pages (home page, careers page, newsroom) pasted in, plus your ICP and disqualifiers. Here is a prompt of ours:
Score [COMPANY] 0 to 3 on Fit and Signal, using only the pasted text.
Today's date: [DATE]
ICP: [INDUSTRY, SIZE RANGE, DISQUALIFIERS]
Fit: 0 = misses the ICP or hits a disqualifier; 3 = matches every part.
Signal: 0 = no named, dated change; 3 = a named change dated in the last 30 days.
For each score give: the score, one exact quote that supports it, and its page URL. For Signal, also give the event and its date.
If the text doesn't say, write UNKNOWN. Never guess.
Pages: [PASTED TEXT AND URLS]
Then run the check:
- Open the quote. For every account that lands in tier A, and every account its Fit score removes (0 or 1), find the quote on the page. No quote, no score.
- Confirm the date. Any Signal above 0 needs an event and a date you can see yourself. A model can invent one, so a missing or fuzzy date drops the score to 0.
- Spot-check the rest. Pull about 1 in 5 of the tier B and C accounts at random and open their quotes too. If more than one in five is wrong, tighten the ICP in the prompt before you trust any of it.
- Handle UNKNOWN. An UNKNOWN Signal is 0. An UNKNOWN Fit means you score Fit yourself, because a 0 there would remove the account on missing text alone.
Keep Deal size and Access out of the prompt. Deal size comes from your own range, and Access needs verified contacts, which a model reading a web page cannot confirm. Score those two yourself.
How much time should each tier get?
Our guidance, per account, per 4-week cycle. A cycle is one full pass through the account: research, first touches and follow-ups.
| Tier | Share of accounts | Research and writing per account | Who you contact |
|---|---|---|---|
| A | Roughly the top 10% | About 75 minutes | Two or three people, mapped before you write |
| B | Roughly 30% | About 15 minutes | The most likely owner of the problem |
| C | The rest | About 3 minutes of your time (the segment message is written once, for the whole group) | One role per account |
These are not the per-prospect research caps from the research-time page linked above. Those cap up-front research on one person; these cover a whole account for a 4-week cycle. Tier A's 75 minutes is about 15 minutes of research on each of two or three contacts, plus writing and every follow-up. Tier B's 15 is a 3-minute look plus writing. Tier C's 3 is a one-minute check plus a share of the segment message.
What matters is the gap: in this split, one tier A account gets 25 times the minutes of one tier C account. Keep a steep gap even if you change every number. If your best accounts are bigger, give A more. If your deals are small and fast, shrink A and B.
For mapping who to contact in tier A, see how many people are in a B2B buying committee and how to find the decision maker. Once a tier A account takes the meeting, how to secure a concrete next step before the call ends covers the end of that call.
What changes in the outreach for each tier?
The time is only worth spending if the message changes with it.
| Tier A: one-to-one | Tier B: one-to-few | Tier C: one-to-many | |
|---|---|---|---|
| Message | Written for this account, about its own situation | A shared message for 5 to 15 accounts with the same trigger or problem, plus one custom line | One message per segment, with variables for name, company and role |
| Research | The event, the people, what they've said publicly, what they'd lose by waiting | One fact per account, found in a few minutes | None per account. Verify the contact and stop |
| Channels | Email, phone and LinkedIn, in an order you chose | Email plus one other channel | Email, usually |
| Follow-up | Each touch adds something new for this account | Follow-ups share a theme across the group | Standard sequence |
Low effort per account still means a message written for a specific group. That takes less time per account, not less thought.
What should you do with the accounts that are off the list?
Remove them and write down why. Every account you remove frees its whole budget for the rest of the list. Record each removal in a sheet: account, date, reason (for example, "no outbound team" or "below our minimum deal size"). Check that sheet before you import any new list, so the same accounts don't come back.
Worked example: how do you spend a weekly budget on 100 accounts?
Our example, invented numbers: one rep has 6 hours a week for account research and personalization, and a list of 100 qualified accounts. Over a 4-week cycle, that is 6 × 4 = 24 hours, or 1,440 minutes.
Step 1: sort the accounts. After scoring, the list splits into 10 in tier A, 30 in tier B and 60 in tier C.
Step 2: price each tier.
| Tier | Accounts | Minutes each | Minutes total |
|---|---|---|---|
| A | 10 | 75 | 750 |
| B | 30 | 15 | 450 |
| C | 60 | 3 | 180 |
| Review and re-scoring | n/a | n/a | 60 |
| Total | 100 | 1,440 |
Tier C detail: 60 contact checks at 1 minute (60) plus two segment messages written once (120) = 180, or 3 minutes per account.
The arithmetic: 750 + 450 + 180 + 60 = 1,440. That's 13.8 minutes per account on average across the 100 (1,380 ÷ 100), before the review block.
Step 3: compare it with equal effort. Spread the same 1,380 account minutes evenly and every account gets 13.8. Your 10 tier A accounts get 13.8 minutes each instead of 75, and the 60 tier C accounts get almost 11 minutes more each than they need. Same hours, worse use.
Step 4: set the weekly rhythm. Divide the cycle by 4. Tier A is about 188 minutes a week, roughly 3 hours, or two or three accounts. Tier B is about 113 minutes. The rest goes to tier C and the review. Block the A time first, early in the week, while you have the energy.
When should an account move between tiers?
Our rules, written down in advance:
- Up: a B or C account shows a new trigger you can name and date. Re-score it within one working day. If it reaches tier A, start the tier A work this week, while the event is still news.
- Down: an A or B account finishes two full cycles (8 weeks) with no reply, no meeting and no other engagement. Move it down one tier.
- Remove: a disqualifier removes the account at once. Tier A is not protected. Log it as described above.
- Review: once a month, re-score the list. Accounts with a stale signal lose those points.
Every quarter, compare outcomes by tier: meetings and deals from A, B and C. Divide the meetings and deals from each tier by the minutes that tier got. If tier A doesn't return more per account than tier C, the score isn't separating good accounts from average ones, so re-weight it. If tier A holds more than about 10% of the list, raise its threshold from 9 to 10. To plan the activity those tiers need, start from quota math, and test a custom first line against a segment line on your own list before you trust either.
Sources
Sources accessed October 5, 2026.
None: this page uses no outside statistics. The scoring, tier shares, time budgets and every worked example are our own guidance and math, and the example numbers are made up.