Client reports now open with a written summary

An agent now writes the narrative that opens a client report, and you can edit every word of it before the client reads it. Brands became a place where portal and reporting settings live, and the client activity feed now says what actually happened.

Reporting carried the day. The report gained a written opening, the activity feed was rewritten so a client can read it without asking what a row means, and brands became a real settings layer.

For our partners

1. Client reports can now open with a written summary of the period, drafted by an agent and editable by you before anyone outside sees it.
2. The activity feed in the client portal says what happened: quiet checks roll up by week, flagged checks name what they saw, and ranking movement shows up both ways.
3. A brand now has its own Portal tab and Reporting tab, so portal pages and report sections can be set once for a group of clients.
4. The report’s keyword trend bars count only the keywords a campaign tracks today. A partner asked why those bars fell to zero after a tracking list was trimmed; that is resolved, and the bars now read against one consistent set of terms.
5. The AI Visibility section plots each check as a point on a real timeline instead of squeezing two checks into a band, and a new card lists the buyer questions the numbers come from.

1. The report summary

The summary sits at the top of the report: a title, a read of the period in plain language, and what comes next. It names its comparisons correctly, so a quarterly report compares against the quarter before it and a six month report against the six months before it, never against last month.

Beside it, on your reporting page only, the agent leaves notes for the agency: what to watch, what is underway, what it needs from the client, the one takeaway, and where the account should be in sixty to ninety days. Those notes never reach the client’s report or the PDF.

An Edit control opens the title, the narrative and the looking ahead list. Saving updates the portal and rebuilds the stored PDF for that month, and the agent will not overwrite a period you have edited. Whether the client sees the summary at all is a report setting, off until you turn it on. We are enabling the writer partner by partner.

2. The activity feed

Franchise clients were reading rows like “flagged 2 items for the team to follow up on” and then asking what the items were. Quiet monitoring checks now collapse into one weekly row with the checks listed. A flagged check gets its own row, names what it saw, and later says it was checked again and nothing new came up.

Tasks show their title, their state and a link to the page they touched. Every row expands into a plain explanation of what that kind of work is, so a client new to the portal is not guessing. And rankings are in the feed now: each Monday a row records the keywords that moved into the top three, onto page one, off page one or back, plus the first time one reaches number one.

3. The brand layer

Brands appear in the left nav once a workspace has one. A brand’s Portal tab sets which pages its clients see in their portal; its Reporting tab sets which sections their reports carry. Settings resolve workspace first, then brand, then client, so deciding what twenty franchisees see is one decision rather than twenty, and a single client can still pin its own answer over the brand.

A client’s own settings sheet now tells you which layer an unset switch is following, with a link to it. A client that belongs to no brand behaves exactly as before.

Behind the scenes

The crawl score was grading sites too kindly. A site could show thirty errors and over a hundred warnings on the card and still read 100, because each problem was divided across every page before it cost anything. We rebuilt the scoring: every page is graded on its own, a problem that comes from one template costs less than the same count of page specific errors, and pages closer to the home page weigh more. Scores will read lower from the next crawl and the spread finally matches what the issue list says.

One of our agents asked a partner whether a photo gallery showed on their page. The photos were fine. The gallery animates in on scroll and our page renderer does not scroll, so the agent took a blank picture as truth and handed its own blind spot to a human who could not see it. We added a line to the instructions telling an agent that when what it sees and what the page says disagree, that is ours to look at, not the partner’s to confirm. In the lab, the model on that job believed the blank picture some of the time without the line and none of the time with it.

Site edits got a guard. Every change our agents make to an existing page is now read back and compared before the edit is reported as done. If anything came back missing, the page is restored to exactly what it was and the edit reports as not done, so a page is never quietly damaged and never falsely successful. We also taught the editor to read and write one more page builder, which had been returning blank pages to agents that needed to work on them.