AI SDR
In short
Why do my contacts have no ICP fit score in an AI SDR sequence?
Scoring needs a LinkedIn profile address, and contacts without one are skipped with nothing shown; that explains most blank scores. The other three causes are the sequence not being an AI SDR sequence, the contact having joined before scoring could run, and several contacts sharing one profile address. A lead source threshold is measured separately from the score you see afterwards.
Quick fix
- Open one of the unscored contacts and look for a LinkedIn profile address. If the field is empty, that is your answer.
- Confirm the sequence is an AI SDR sequence. Regular sequences do not score anyone.
- Check when the contact joined. Scoring happens on the way in, and contacts already in the sequence are not filled in afterwards.
- Remove and re-add a contact whose profile address you have since filled in; that puts it through scoring again.
- If several contacts carry the same profile address, correct them; one evaluation is being written across all of them.
Didn't work? Contact Reply support; the team can check your account directly.
Symptom
Inside an AI SDR sequence, some or all contacts show no fit score and no LinkedIn activity score. Hovering produces no written explanation because there is nothing to explain. Other contacts in the same sequence are scored normally.
What the two scores are
Inside an AI SDR sequence each contact can be given two separate readings:
- a fit score, for how closely the person matches the offer attached to that sequence
- a LinkedIn activity score, for how active they are on LinkedIn
Hovering over either one shows the reasoning behind the number in plain words. Both are worked out per sequence, not per person, and each sequence scores from scratch: which is why the same contact can read differently in two sequences.
The four reasons a score does not appear
| Cause | Tell-tale sign |
|---|---|
| No LinkedIn profile address on the contact | Most of the unscored contacts have the field empty |
| Not an AI SDR sequence | Nobody in the sequence is scored, and no columns exist on the People page |
| The contact joined before scoring could run | Older contacts blank, newer ones scored |
| Several contacts share one profile address | Identical scores on people who are clearly different |
1. The contact has no LinkedIn profile address
This is by far the most common answer. The evaluation reads the person's LinkedIn profile, so without an address there is nothing to read. Contacts in that state are skipped without a message, a marker or a placeholder, and the column is left blank.
How to recognize it: open two or three unscored contacts. If the profile address is missing on all of them, stop here.
What to do: fill in the profile addresses and put the contacts through the sequence again. Imports and list uploads are the usual source of the gap, so it is worth checking the column mapping on whatever brought them in.
2. Scoring only runs in AI SDR sequences
Neither score exists outside an AI SDR sequence. A regular sequence never evaluates anyone, and the columns do not appear on the main People page at all, so there is no account-wide view of who fits and who does not.
How to recognize it: nothing in the sequence is scored, not one contact.
What to do: if you want people scored, they need to be in an AI SDR sequence with an offer attached. There is no way to score a list in place.
3. The contact was already in the sequence
Scores are worked out at the moment a contact joins a sequence. Contacts that were already there when scoring became available are not evaluated retrospectively, and nothing goes back to fill them in.
How to recognize it: the blanks line up with a date. Everyone added before a certain point is unscored; everyone since is fine.
What to do: remove the contacts and add them again if the score matters. Bear in mind that re-adding restarts them in the sequence, so weigh that against the value of the number.
4. Several contacts share one LinkedIn profile address
When the same profile address appears on more than one contact, a single evaluation is carried out and the result is written to all of them. The score is accurate for one of those people and misleading for the rest.
How to recognize it: two or more contacts with clearly different names, companies or job titles carry the same score and the same reasoning text.
What to do: find the duplicated address and correct it. Copy-paste during manual entry and a mis-mapped column on import are the two common sources.
A lead source threshold and a displayed score are different numbers
This is the most common confusion on this page, and it is not a fault.
When you set a lead source to accept only people at or above a given percentage fit, that filtering happens at the source, before the contact ever reaches your sequence. The score you see afterwards is a separate calculation, made against the offer attached to that sequence.
Two measurements, two moments, two reference points. A contact can therefore be admitted at one number and displayed with another, including displayed below the threshold that let it in. Neither number is wrong; they are answering different questions.
If you want the displayed scores to sit higher, the lever is the offer on the sequence, not the threshold on the source.
One caution before you reach for that lever. Editing the ICP on a sequence can replace the contact-search filters that sequence is using, so a change made to improve the scores can quietly change who is being searched for as well. Note your current filters before you edit, and check them afterwards.
FAQ
Why is the same person scored differently in two sequences?
Two reasons, and the second is the one people do not expect.
The first is the obvious one: the score measures the person against that sequence's offer, not against a fixed profile. Two sequences selling different things will reasonably disagree about the same contact.
The second is that the score is worked out afresh for every sequence and is not perfectly repeatable. It is produced by a language model rather than a formula, so two runs on the same person against the same offer can land a little apart. Nothing is carried over between sequences, so there is no shared answer to be consistent with.
Treat the score as a strong signal for sorting and filtering, not as a precise measurement. A contact at 58 and a contact at 62 are not meaningfully different; a contact at 20 and a contact at 80 are.
Why do I see contacts below my threshold?
Because the threshold was applied at the lead source and the visible score was calculated later, against the sequence's offer. See the section above; the two numbers are not measuring the same thing.
Where are the score columns on the People page?
They are not there. Both scores live inside AI SDR sequences only, so there is no way to sort or filter your whole contact list by fit.
Still stuck? Contact Reply support
If these steps didn't solve the problem, the Reply support team can look at your account directly. Open the Reply Help Center and send the team a message.
To get an answer faster, include:
- the name of the AI SDR sequence and roughly when the contacts were added
- whether the unscored contacts have a LinkedIn profile address
- how the contacts arrived – import, lead source, or added by hand
- any fit threshold you set on the lead source
Don't send passwords, API keys, or access tokens.
Related
- Why did a lead source add no contacts? – where the threshold is applied
- Why did Autopilot stop adding contacts?
- Why did a search return too few contacts?
Build with Reply
- REST API: Contacts – read contacts and their sequence membership. The scores themselves are not on the API
- The scores are not available to an assistant either. They are shown in the sequence only