The Insights tab — your weekly read

Seven cards you read once a week: who you can actually reach, which single document is jamming registration, where consent is short, how families answer, which languages they need, what they keep asking, and where your process is uneven.

The Insights tab is the slow one. Nothing on it needs looking at before Monday — daily triage stays on your dashboard, and this week's call list stays on the Enrollment tab.

Can we actually reach them?

One fraction: households with at least one channel that would work right now, out of every household on your roster.

Two rules make that number worth reading, and both make it *lower* than the figure most systems print:

• Nobody leaves the bottom of the fraction. A family with no phone number, no email address and no consent stays in the denominator and counts as unreachable. A contactability figure that goes *up* when your directory gets worse is not measuring anything.
• A channel only counts if it would work. A recorded consent that was withdrawn, a do-not-contact request, a channel you have not switched on yet — none of them count, however filled-in the field looks.

Siblings are one household. 164 applicant rows can be 138 families, and a guardian with three children is one ringing phone, not three points of a percentage.

Underneath, every household below the bar is listed by surname with the one action that fixes it. The fraction is the summary; the list is the work.

One thing this card cannot see yet. Nothing in the product records a bounced email or a disconnected number against the contact it came from, so an address that has been failing for months still counts as reachable here. The card says so on its own face. That is our gap, not your roster's, and it is on our list.

What is jamming registration

The headline is the last missing document: among families whose file is complete except for one item, which item is it. That is the queue where one phone call finishes a registration today.

It is deliberately not the most-missing document. A family who has submitted nothing contributes five missing items, so a "most missing" ranking is dominated by families for whom the missing item is not the problem. Both rankings are shown, labelled, and never mixed up.

Broken down by grade and by language, so "fourteen of our twenty-two are Spanish-preferred families" is visible — that is a translation job, not twenty-two phone calls.

We hold whether each checklist item is in or missing, and nothing else. No document is stored, opened or read. See Family outreach.

Consent coverage, and what it held

Two numbers on one card, because separately they mislead: how many households you have permission to text, email and call, and how many messages the consent gate refused to send this month.

Coverage alone reads as paperwork. Held messages alone read as a fault in the product. Together they are one sentence you can act on — *we may not text twelve of our families, and that is why thirty-four messages did not go out*.

Every held message is also your evidence: your agreement promises this exception log on request, and this is it. See Consent & opt-ins.

Households who asked us to stop are counted separately and are never presented as a gap to close.

How families answer

Which channel your families actually use, counted per household.

A text reply and an email reply are recorded the same way in your roster, so replies that carried no channel of their own are shown together in one honestly-named row rather than split by guesswork. Calls, WhatsApp, website questions and STOP messages do name their channel and are counted separately.

This is a staffing signal. It does not say a channel *causes* enrolment, and Elira will never present it that way.

Language demand

Counted twice, because text and voice are different capabilities that fail differently: Elira writes and speaks English, Spanish and Arabic. A family who prefers another language is still contacted — in English — and the card counts how many, separately for texts and for calls.

The English row is a ceiling rather than a count: your roster records English when no preference was ever captured, so it includes families nobody asked.

What families keep asking

Questions Elira could not answer from your own material, grouped by topic. Nineteen bus questions in a month is one paragraph for your website, after which those nineteen questions stop arriving.

Questions that reached a human only because our model was unavailable are counted, labelled as ours, and kept out of the ranking. Our outage is not a gap in your knowledge base, and you cannot fix it.

The questions themselves stay on the AI Knowledge tab — this card is counts only.

Where your process is uneven

The last card compares one thing — how often a registration gets finished — across two groups your roster already records: students with an IEP flag, and families by preferred language. For example, *9 of 22 IEP-flagged students completed registration, against 25 of 37 with no flag*.

Three things about that card, and all three are deliberate:

• It is about your process, not about your families. It contains no names, no phone numbers and no student list, so it cannot be turned into a call list. Who to call today lives on the Enrollment tab, and that list does not know which families have an IEP flag or which language they prefer.
• Any group under ten is withheld outright. Not blurred, not rounded — withheld, with the numbers removed rather than hidden. In a school this size a small group is an identifiable child.
• It shows WHERE, never WHY. Two fractions side by side, and no single "gap" figure. What is behind a difference is yours to look into; nothing here claims to know it, and nothing here should be read as blaming a family for it.

Only a yes/no IEP flag is held. No disability category, no service plan and no document content is stored or read anywhere in this product.