Which part of sending a client report actually needs a human? Not the refresh. Not the reminder. Probably not the addressing either, assuming your contact list is in order. Most of what agencies call automated report delivery is clerical work that got promoted to account management because it happens to travel by email.
What does need you is small and easy to lose in the plumbing: the two or three sentences that tell the client what they are looking at and what you did about it. That is what makes someone open the message on a Monday instead of filing it. Automate the wrong step and you end up with a very efficient pipeline that pushes a voiceless report at 09:00 every week, which clients quietly stop opening around week five.
So the question worth answering is not whether you can schedule the send. You can. It is where the handoff from data to inbox actually runs, and which link in that chain carries your judgment.
What should fire on its own, and what shouldn’t
Delivery is four steps: refresh the data, trigger the notification, route it to the right people, and say something about it. The first three are rules. Given the same inputs they produce the same output every time, and a person doing them by hand is just a slower rule with a worse error rate.
The fourth is not a rule. “Video completion rate dropped from 71% to 58% because we rotated in the new 30-second cut mid-flight” is not derivable from the file. It comes from knowing what you changed and why. That distinction is the whole design brief for report delivery automation: mechanise the three steps that have no opinion in them, and make the one that does impossible to skip.
The failure mode agencies reach for is the opposite. They automate the sentence too, with a template that says “Please find your weekly performance report attached. Let us know if you have any questions.” Nobody has questions about that message, because it contains no claim to have questions about.
The four-step handoff, from export to inbox
Step one: the data has to be current before anything else fires
Everything downstream is worthless if this step is stale. In practice that means the report is reading from the latest export, and the client is reading the report rather than a copy of it.
This is where PDFs make scheduling structurally unsafe. A scheduled PDF is a promise you made at build time about numbers that keep moving. A live link is at least honest: if you update the underlying data on Tuesday, the Monday link shows Tuesday’s numbers. Screebl works this way by design, with one campaign export per report and a reusable template doing the layout, so refreshing a report is an upload rather than a rebuild. That removes the step where most manual delivery goes wrong, which is not the sending but the re-production.
Step two: the trigger
Pick one condition and be literal about it. “Every Monday at 10:00” is a trigger. “When the report is ready” is not, unless something specific marks it ready. The version I would trust: the account manager marks the report as published, and the notification fires on a schedule only for reports that are published. A report that nobody touched does not get announced.
Step three: routing
Routing is the step everyone assumes is trivial and it is the one that produces the embarrassing incidents. Each client should receive their own report and nothing else, and that has to hold when a recipient forwards the message to three colleagues you have never met. Per-client access keys and a portal where a client only ever sees their own reports do more for this than any care taken over the To: field, because they survive forwarding. If you want the longer version of that conversation, including what a client’s IT team will ask you about it, secure report sharing covers it properly.
Step four: the commentary
Short, written by a person, written before the send and not after. Three or four sentences is plenty. More on what belongs in them below.
Where automated report delivery quietly breaks
The breakages are not dramatic. That is the problem. A manual send fails loudly, because you notice you have not done it. Automation fails silently and keeps running.
Stale or partial data. The export lands late, the schedule does not care, and the client receives a month that ends on the 27th presented as a full month. Programmatic conversions get restated after the fact in DV360 and Campaign Manager 360, which means a Monday morning send can show a conversion count that is genuinely lower than the one the client screenshotted on Friday. Either build a check that the data covers the period you claim, or set the send late enough that it does.
Wrong routing after a change nobody told you about. The client-side marketing manager moves to a different brand. Procurement gets added to the distribution list and now receives spend figures in a context you did not write for. A colleague reuses last quarter’s recipient list for a client who changed agencies of record. None of these throw an error.
Dead accounts and unread bounces. A scheduled send from an unmonitored address bounces into an unmonitored mailbox. Two months later the client says they have not seen a report since June and they are right. If you automate the send, someone has to own the bounce list, and it should be a person on the account rather than whoever set the automation up.
Reports that arrive for campaigns that ended. Flight closed on the 14th, the schedule keeps firing, and the client gets three more notifications about a flat line. Ending a cadence deliberately is part of running one.
Setting the send time so the report doesn’t arrive before the numbers do
Work backwards from when your data is trustworthy, not forwards from when the client’s week starts. For a weekly report closing Sunday, that usually means Tuesday rather than Monday at 07:00, because someone has to look at the file first, and because platform figures for the last day of a period are the least settled figures in the report.
Monthly is stricter. If the report is going to be read next to an invoice, it has to reconcile with the media plan before it goes out, and that reconciliation is human work. Send after it, never in parallel with it. The wider question of how often to report at all is a separate decision, and worth making explicitly rather than inheriting: weekly or monthly goes through the trade-offs.
One more scheduling habit that pays for itself: put the send an hour or two before a window when you are actually at your desk. Client report notifications produce replies, and replies that sit for a day cost you more goodwill than the automation saved.
The three lines that keep a scheduled send from reading like a utility bill
Whatever field you write into, whether that is the notification text or a text block at the top of the report itself, these are the lines worth writing:
- What moved. One number, with direction and size. “Spend is at 94% of plan with four days left” tells them where they stand before they click.
- Why, or that you do not know yet. “CPM is up 18% because we added connected TV inventory on the 9th” is an explanation. “We are still checking the CPM increase with the trading team” is also fine, and more credible than silence.
- What happens next, and what you need from them. A decision to make, an asset to approve, or nothing at all. Say which.
Pre-empting the obvious question is the highest-value thing in that block. If VTR fell and you know they will ask, answer it before they do.
None of this takes long once it has a home in the workflow. The reason it gets skipped is that manual sending eats the time it would have taken, which is the actual argument for scheduled client reports: the ten minutes you spend writing ten “your report is ready” emails is ten minutes not spent on the only part of the message the client reads twice.
What happens when someone replies to a message a machine sent
They will reply. Assume it, and make sure the reply lands with the person on the account rather than in a no-reply void or a shared inbox nobody owns. A scheduled notification that cannot be answered is worse than no notification.
Treat each reply as a finding about the report rather than an interruption. “Is this including the search spend?” means your report does not make channel scope obvious. “Why is reach lower than last month?” means the frequency context is missing from the page. Over a quarter those replies are a decent backlog for your template. Client report questions goes further into reading them that way.
For the narrow, factual questions, Screebl can carry some of that load with an optional assistant on the shared report, which answers from that one report only, respects the client’s access key, refuses anything outside it, and only appears when you have written a campaign brief for the campaign. It handles “what period does this cover” without waking anyone up. It does not handle “should we move budget to CTV next month”, and it should not.
That split is the whole point. Schedule the refresh, the trigger and the routing, and keep the sentence that explains the quarter in human hands. If you want to see how the live link, access keys and the client portal fit your current reporting day, the plans are the place to start.
