The most common email an account manager receives is not a complaint. It is a question that the report already answers: “why is CPM higher this week?”, “are we on budget?”, “which creative is winning?”. Each one takes five minutes to answer and interrupts something else.
Screebly is an assistant that lives on a shared report and answers those questions from the report itself. This is what it does, what it will not do, and why it only appears when the agency decides.
How it appears
An agency writes a short campaign brief on the campaign: objective, audience, flight dates, budget, anything the client should be reminded of. If the brief is empty, the assistant does not exist on that campaign’s reports. If it is filled, a small chat bubble appears in the corner of every shared report for that campaign.
That is the whole switch. There is no separate toggle to forget, and there is no assistant on a report the agency has not given a context for. An assistant without a brief can only restate the charts, which is not worth a client’s time.
What it answers well
It reads the report the client is looking at: every KPI, chart series, table row and text block, plus the brief and the campaign’s dates and budget.
- “Are we on track?” It knows the flight window and the budget, works out how far through the flight you are and how much is spent, and says whether pacing looks even. Date arithmetic is done by the system, not guessed by the model, so the answer is consistent.
- “Why did clicks drop in week three?” It looks for what else changed in the same period on the page: spend, impressions, a creative’s share, and explains the relationship.
- “Which creative should we scale?” It ranks what the report shows and says on which metric.
- “What does VTR mean?” It explains the metric in the client’s language.
- “Who are you?” It says it is Screebl’s report assistant, not a person, and not a general chatbot.
Answers lead with numbers from the page and keep adjectives to a minimum. Clients told us they trust an assistant more when it sounds like an analyst than when it sounds like a brochure.
What it refuses, on purpose
- Other reports and other campaigns. It sees one page. “How does this compare to last quarter?” gets an honest “I can only see this report; your account manager can pull the comparison.”
- Anything outside the report. Market news, competitor spend, general marketing advice. It will say it cannot help with that and point to the agency.
- Actions. It cannot change a budget, pause a campaign or send an email. It says so if asked.
- Its own instructions. Attempts to extract how it is configured are declined.
- Reports it has no right to see. The assistant has the same gate as the report. If a client’s access key is required and not supplied, the assistant does not answer, because answering would be a way around the key.
Why it is opt-in per campaign
Two reasons. First, the assistant should only exist where someone has given it something to reason with. Second, agencies want to decide which clients get it, and when. Some clients want the bubble on day one; some would rather have the account manager as the only voice for the first months. Making it a property of the campaign brief keeps that decision with the agency.
What the client sees
A small purple bubble, bottom right of the report. Opening it shows a few starter questions based on the campaign. The conversation stays in the client’s browser on that report; it is not shared with other viewers of the link and not stored on the report. A “clear” control wipes it.
The bubble is not part of the report’s exports. PDFs come out clean.
What the agency sees
Nothing changes in the agency’s workflow except one field on the campaign. Write the brief the way you would brief a new colleague: what the campaign is trying to do, for whom, with what money, over what dates. Two paragraphs is plenty. Update it when the plan changes; the assistant reads the current version.
Where this goes
Today the assistant is a first-line analyst for one report. The questions it cannot answer, comparisons across periods and campaigns, are data questions, and they are the next thing on the roadmap. We would rather it says “I cannot see that” than guess, which is also the standard we would want from a junior analyst.
If you want to try it, write a brief on one campaign and open its shared report. The bubble is there, and the first question most people ask is “are we on budget?”, which is exactly the question it was built for.
