What your audit score means
The number out of 100 on Data Quality is an average of four things MarqIQ can measure on a person record: how complete it is, whether its email and phone are usable, how recently anything happened on it, and whether its values are written consistently. It is a score about records, not about your marketing. This article says what moves it, what a band means in practice and what it does not prove. For the report that grades a tool rather than its records, read Run an integration audit report.
Before you start
One number, one screen
The Integration Audit Report has no single score, and neither does the Data Audit tab. If somebody has quoted you a MarqIQ number out of 100, it came from Data Quality, under Intelligence Cloud, and it is the average of the Overall quality score across your people.
What it is measured on
The score is worked out per profile, and a profile is a person after records from every connected tool have been resolved onto one identity. The number on the ring is the mean of those profile scores, and the line beside it states the denominator: “average across 41,208 people”. Read that line with the number, every time.
- Up to 60,000 profiles are read. Records that carry an email fill that window first, because identity resolution needs one, and within each group the newest-created come first.
- Where your store is larger than 60,000, the score therefore describes your newest 60,000 records rather than an even spread of your history. That is the one figure in MarqIQ measured that way, and it is worth saying aloud before anyone repeats it.
- Records with no email at all cannot join a resolved profile, and are counted separately on the Intelligence Cloud overview rather than scored here.
- A record you have deleted in the tool is excluded.
Take care
Two workspaces with the same score are not comparable. The score rewards filled fields, so a workspace that collects a phone number on every form starts higher than one that deliberately asks for two fields. Compare your own score against your own score, one quarter to the next.
The four components, and exactly what moves each one
The What the score is made of card shows all four as their own averages, each with its share of the score: complete 35%, valid 30%, fresh 20% and consistent 15%.
| Component | What it counts | What moves it |
|---|---|---|
| Completeness | How many of seven fields are filled: email, first name, last name, company, job title, phone and country | Each missing field costs about 14 points on that record. A record with all seven scores 100. Company means a company held in one of your tools; a name MarqIQ reads off the email domain does not count as filled |
| Validity | Whether what is there can be used | Starts at 100. An unusable email costs 50. A phone number too short to dial costs 15. A personal address such as Gmail or Outlook is a valid address and costs nothing. It never goes below zero |
| Freshness | How long since anything happened on the record | Under 30 days scores 100, under 90 scores 80, under 180 scores 60, under a year scores 40, older scores 20 |
| Consistency | Whether values are written the way a report can group them | Starts at 100. A country written as a code of three characters or fewer that MarqIQ cannot resolve costs 20. A job title that does not classify costs 15. Every value that Normalization would change costs 10, up to 30 |
Which date counts as the last activity
MarqIQ takes the newest date your tool holds for something that happened with the person: a reply, a website visit, a form submission, an email open or click, or a call, meeting or note logged on the record. It reads every field whose name says so, including your custom fields, and never one fixed field. A record being edited, an email sent to the person, and a meeting booked for a future date are not counted as activity.
People with no activity date
Where none of a person’s records carries a last-activity date, freshness cannot be judged. MarqIQ leaves freshness out of that person’s overall and re-weights the other three. The Fresh average is taken over the people who do have a date, and the card says how many that is. The date MarqIQ last synced a record is never used in its place: it says when MarqIQ read the record, not when the person last did anything.
What each band means in practice
Each profile also gets a word, and How people grade shows how your people split across the four. The shape of that bar tells you more than the average does.
| Band | Score | What it looks like on a record |
|---|---|---|
| Excellent | 90 and above | Nearly every field filled, a usable work email, something happened in the last month or two |
| Good | 75 to 89 | One or two fields missing, or activity a few months old |
| Fair | 50 to 74 | Several fields missing, or an old record, or values that need standardising before they group |
| Poor | Under 50 | Usually an unusable email — that single check costs half of validity — or a record with little more than a name |
A high average with a long Poor tail is a different problem from an even Fair across the board. The first means one intake route is producing bad records; the second means your fields are thin everywhere.
What a poor score costs, in terms a board understands
Each component maps onto something that breaks downstream. Say the consequence, not the score.
- Low completeness means routing and segmentation guess. A lead with no company and no job title cannot be sent to the right rep or the right nurture, so it goes to the default — which is usually the slowest one.
- Low validity means you are paying to reach people you cannot reach. An unusable email bounces, and bounces damage the sending reputation of every other campaign that quarter.
- Low freshness means your reachable audience is smaller than your database. The Audience health analysis is the screen that puts a number on that.
- Low consistency means spend is judged against the wrong bucket. “LinkedIn”, “linkedin” and “LinkedIn Ads” are three rows in a report and one channel in a budget, so each looks a third as effective as it is.
What a good score does not prove
This is the part to say before anyone builds a target around it.
- It says nothing about attribution. A record can score 100 with its source field holding “Direct” because nothing ever wrote a real one. Completeness counts filled, not correct.
- It says nothing about whether the right field is being read. That is what Field Intelligence answers.
- It says nothing about whether a first-touch field is being overwritten, because it reads the record as it stands today and not its change history.
- It does not measure your deals. Every component above is about a person record.
- It cannot see a field it does not know to look for. The seven completeness fields are fixed; a custom field your business depends on is not in the score, and Field Intelligence is where you check that one.
Setting a target for next quarter
A target on the overall number is hard to act on, because four things move it at different costs. Target the component instead, and let the overall follow.
Read What the score is made of and take the lowest of the four.
Read Core fields filled, per person on the same screen.
It lists the eight fields it tracks with the number and share of people carrying each, so you can see which single field is dragging completeness down. What each tool holds, underneath, gives the same eight fields tool by tool, over each tool’s own records, with a link to that tool’s Data Audit for every other field.
Pick one field and fix where it is captured, not where it is stored.
Adding a field to your main form moves new records only. The score covers your newest 60,000, so a capture fix shows up over the following weeks rather than the following day.
Read the lowest-scoring people at the foot of the screen.
The fifty lowest-scoring people are listed, each with the four parts beside the overall, and Open to go to the record in its tool where MarqIQ can link to it.
Read ten of them.
Ten usually tell you which import or which form produced the tail.
Write the target as a component and a denominator.
“Validity from 71 to 80 across roughly 41,000 profiles by the end of the quarter” can be checked. “Get to 85” cannot be argued with or planned against.
Tip
Freshness falls on its own as records age, so a workspace that changes nothing drifts downwards. A flat score quarter on quarter is a small win, not a stalemate.
What to say alongside the number
Whenever this score leaves MarqIQ — in a client report, a board pack or a QBR — send these three sentences with it. They stop the single most common misreading, which is that the score grades the marketing rather than the records.
- What it is: the average record-quality score across N profiles, on completeness, validity, freshness and consistency.
- What it is not: a measure of campaign performance, attribution accuracy or revenue.
- What was measured: the number of profiles, and whether that was your whole store or your newest 60,000.
If it did not work
The score dropped and nothing changed in your tool
Why: Freshness falls as records age, and a sync that brought in a batch of thin records lowers completeness the moment it lands.
Fix: Compare the four components rather than the overall. A fall in freshness alone is time passing; a fall in completeness alone is new records arriving with fields empty.
The profile count is far lower than your record count
Why: One profile can hold several records — the same person in two tools resolves to one — and records with no email cannot join a profile at all.
Fix: Open the Intelligence Cloud overview, which counts the records with no email per tool, so you can see which source is producing them.
Validity looks low although the addresses look fine
Why: Validity falls for two things only: an address that is not well formed, and a phone number too short to dial. A personal address such as Gmail does not lower it.
Fix: Read Unusable email on the Intelligence Cloud overview, under What is missing, for the number of malformed addresses. If that is low, the phone field is the cause: look at how your forms collect it.
Fresh reads “No dates”, or a person shows a dash under Fresh
Why: Freshness needs a last-activity date on the record. Where no record of a person carries one, that part is left out of their score and the other three are re-weighted.
Fix: Check whether the tool keeps a last-activity date for people. A tool that does not expose one leaves this part blank by design, and the other three still score.
The score has not moved since you fixed a field
Why: The result is stored against a fingerprint of your data and recomputed when that data changes. A change made in the tool reaches the score after the next sync, not immediately.
Fix: Run a sync from Connected tools, then reload Data Quality.
Next steps
Related articles
Was this article helpful?