Can you trust the field you report channels from?

Which field it reads

MarqIQ picks the field your reports read as “where a person came from” by what the field holds, not by a fixed name. A custom field is read the same way as a native one. The method at the foot of the analysis names the field it chose and why.

Where no field clearly records it, the analysis says so and lists what you hold that comes close, each with its fill rate. If one of those is your source field, open the Field Intelligence tab and choose it under Fields we rely on. This analysis reads it from then on.

What it checks

Six things, listed most serious first under What is distorting your channel numbers. Each carries the evidence, what it costs you in the budget conversation, and how teams normally fix it.

People with no source at all

How many people have nothing in the field, and whether it is getting better or worse. The trend compares people created in the last 90 days against the whole population, and it is only stated with at least 30 recent people. A quarter of the database or a worsening trend makes it a fault rather than a risk.

These people are left out of every source report, which makes each channel look smaller than it is. When a fifth of your records are missing, a channel’s share can be wrong by that much in either direction.

Buckets that name no channel

Values such as “Other”, “Unknown”, “Web” and “Offline Sources” are counted separately from real channels and tagged names no channel in the table. They are a fault when they hold three in ten sourced people, or a quarter of your won revenue.

One channel spelled several ways

“LinkedIn”, “linkedin” and “LinkedIns” are three values to your tool and one channel to you, so a real channel looks like three small ones. Two values are called the same source only when they differ in spelling — capitals, punctuation, plurals. Meaning is never guessed, so two real channels are never merged.

A source field that is really a notes field

Raised when the field holds 40 or more values and at least half of them are used by fewer than five people. The field is free text in all but name. The analysis sorts your own values into the handful of channels they already are, using your counts, rather than handing you a stock list of channels.

A source that brings people but not customers

The largest source that brings at least a tenth of your sourced people and whose people become customers at less than your own average, stated with both counts and the account average beside it. It needs at least 30 customers in the whole account before it is raised at all, and the gap has to be too large for chance.

Two limits are worth knowing. A source below a tenth of your sourced people is never raised here, however poorly it converts. Nor is a catch-all value such as “Other”, because the check above already covers those.

The finding tells you not to cut the channel on this alone. Check what it costs per customer rather than per lead, and whether its people are earlier in a long cycle, which the Funnel analysis shows.

Where your tool and MarqIQ disagree

This check needs the tracking script live. People are matched by email to a form fill MarqIQ captured on your own website, and the two answers are compared by channel family — paid search, email, events — not by exact words.

• People your tool calls direct, blank or a catch-all while MarqIQ recorded a specific first touch. Raised once at least 20 people are matched and a tenth of them, and at least five, disagree. A quarter of them makes it a fault rather than a risk.
• Where those people arrived from paid advertising, the finding says so, because money was spent to get them and the channel is getting no credit.
• A straight clash — one channel in your tool, a different one in MarqIQ — is raised once 15% of matched people clash, and at least five of them do. Usually one side records the first touch and the other the converting touch, and nobody has said which the company reports on.

Every source, side by side

Under the findings there is one row per source value, up to 25 of them.

What it will not do

• It changes nothing in your tool. The method says so in as many words, and every fix is a description of work, not an action button.
• It never merges two values because they mean something similar. Spelling only.
• It does not tell you to rewrite a field your tool will not let you rewrite. Where the source field cannot be meaningfully edited, the fix names a second field instead.
• It quotes no benchmark and no industry average.

The four numbers at the top

The header carries People, With a source, Different values and Customers, and a headline sentence naming the field and how many things are distorting it now against how many are worth fixing before the next budget review. When nothing is wrong, the headline says the field is in good shape and gives the fill rate and the leading value.