Rank last week's influencer campaigns and post them to Slack

Every Monday at 9am, build a ranked digest of last week's influencer campaign performance from Later and post it to our #influencer-marketing channel in Slack. Run this on a cron schedule. Later sends no outgoing webhooks, so a schedule is the only way to start this.

First, use Later's Get Campaign Performance to fetch per-campaign metrics for the trailing seven days, sorted by engagement descending. Later requires an explicit metrics list on every request and returns only the metrics you ask for, so request only the ones the digest actually displays: engagement, impressions, engagement rate, and post count. Supported metric names differ per endpoint, so check the list on Get Campaign Performance rather than assuming a shared set.

Then make a second Get Campaign Performance call covering the seven days immediately before that window, holding the same metrics, sort, and filters constant. Match campaigns across the two calls by campaign id so you can compute the week over week change in engagement for each one. Campaigns that only appear in the current week are new, not recovered, so treat them separately instead of reporting an infinite gain.

Then call Get Instance Performance Over Time bucketed by week, covering the past three weeks, so you can compare the past week against the prior two and say where the program as a whole is heading rather than judging it off a single week.

Compose the digest and send it with Slack Bot's Send a Message to #influencer-marketing. Lead with a one line headline on overall program direction, drawn from the three week trend. Then list the top five campaigns by engagement, each with its headline metrics on one line. Then flag any campaign whose engagement fell more than 25 percent week over week, naming the campaign, the drop, and the current week's numbers, so the team knows where to intervene before spend keeps flowing. If nothing fell past the threshold, say that in one line rather than dropping the section, so the message shape stays the same every week.

Details to respect when calling Later. All dates are YYYY-MM-DD in UTC, and the window between the start and end date cannot exceed two years. The dateBasis parameter defaults to post_date, which filters by publication date, while performance_date filters by when the metric was recorded. Pick one and use it consistently across both weekly windows, otherwise the comparison is not like for like. The campaignIds parameter accepts at most 50 ids per request, so batch if we track more than that. Responses are wrapped as data plus nextCursor, so follow the cursor until it comes back null, holding every other parameter constant while paging. The rate limit is 120 requests per minute per IP, so back off and retry on a 429 rather than failing the run.

Format the Slack message in Slack mrkdwn, which uses single asterisks for bold rather than double. Keep it scannable: a headline line, then the ranked five, then the flags. No preamble and no closing pleasantries. Later is read only, so this workflow reports and never writes anything back to Later.