Media Insights

What this page is for

  • The Media Insights page is a weekly report on your paid campaigns, written for you by our AI based on your own data: how your advertising is going, what changed against the previous period, what performs really well or really badly, and what to do next. It reads the campaign figures of every advertising platform you run (Google Ads, Facebook Ads, TikTok, Pinterest …): spend, impressions, clicks, video views, conversions and the cost per result, for each campaign and each campaign objective.
  • Why you need it: the Digital Media pages show every number; this page answers the questions of a weekly status meeting without the meeting. Every week it writes a verdict for the period, the two or three moves that matter, the best and worst performers, and the next steps — each backed by the exact figures, so you can check them.
  • What you gain concretely by using this page every week: (1) a one-line verdict per period — on track or not — with the numbers behind it; (2) the significant moves against the previous period, with their direction and their cause; (3) the campaigns that deserve more budget and the ones that waste it, compared only with campaigns bought for the same purpose; (4) a short list of actions in priority order, each naming the campaigns to act on; (5) a warning when a platform's data stopped arriving, so you never mistake a missing feed for a collapse.
  • One table row = one record of the report. The figures cover the last 7 or the last 30 complete days (the period control), each compared with the 7 or 30 days before. The page shows the latest report only; a new one replaces it every week, on your review day.

The controls at the top

The two controls choose which report you read; they do not filter rows away. To return to the default view, clear them with the Reset button.

  • page — whose story you read: executive (the default) is the account-wide story — how the money splits between platforms, how the whole account paces, which platform does better for the same objective, which feeds are missing; each platform name (Google Ads, Facebook Ads …) is that platform's own story, campaign by campaign. The executive page never repeats a platform's findings, so read it first, then each platform in turn.
  • period — the window the records are about: 30d (the default) is the last 30 complete days compared with the 30 days before; 7d is the last 7 complete days compared with the 7 days before. Every record belongs to one period, and the report writes a complete set for each, so switch to 7d to see what is happening right now and whether it confirms or reverses the month.
  • The date (the box on the right) — the date of the report you are reading: the day of the latest weekly run. You don't change it; the page always shows the latest report.

The four tables, in screen order

Each table is one section of the report, and the four together are one week's story in a fixed order: the verdict, what moved, what stands out, what to do. The tables are short on purpose: at most one row in Overview, four in Comparison and Facts, three in Actions, and a quiet week has fewer — the report is told not to pad. In every table the first column is the row's verdict, and the row's colour follows it: green for good, red for bad or urgent, orange for a warning, grey for neutral.

  • Overview — How is your paid media performing? — one row per period, the verdict. status — on_track (the period's objectives are being met), mixed (some good, some bad), off_track (the main figures move the wrong way), no_activity (no spend in the period), data_gap (a platform's data stopped arriving and dominates the period). issue — the verdict in one line, with at most two rounded figures. detail — two to four sentences without numbers: the state, the biggest move, the direction of the next steps. evidence — the two to four numbers that define the period, with the previous value and the change where there is one. You might read that spend rose by a quarter while conversions fell, and that the cost per conversion is the problem of the month.
  • Comparison — What changed versus the previous period? — the significant moves against the previous period, most important first. direction — the move and whether it is good news: up_good (a result grew — more conversions, more clicks), up_bad (a cost grew — a higher cost per conversion, more spend for the same result), down_good (a cost fell), down_bad (a result fell), neutral (a deliberate move of money, a shift between platforms or objectives, a pause). issue — the move in one line; evidence — the figures: value, previous value, change in percent. A stable figure is not a move; this table lists only what should change a decision. You might find that one platform now carries all the spend, or that the cost per conversion of your sales campaigns improved while their volume held.
  • Facts — Your best and worst performers, anomalies and gaps — the current state, not the change. kind — highlight (performs really well now), lowlight (performs really badly now), neutral, warning (something to watch — for example a large share of spend in campaigns without a clear objective), anomaly (a figure that does not fit the rest), data_gap (a platform's data ends before the period does). issue and evidence as above. One finding lives in one table only: if the point is the change, it is a Comparison; if the point is the level, it is a Fact — the same number is never cited twice. You might find the campaign with the lowest cost per sale and the most sales, or a platform whose last data is three weeks old.
  • Actions — Recommended next steps, in priority order — what to do, each as "do this, on these campaigns, because". priority — high (act this week), medium, low. verb — the one thing to do: increase, decrease (a budget or a bid), pause (a campaign), reallocate (money between platforms or objectives), maintain (keep as is), investigate (look at a campaign's setup), fix_data (restore a platform's data feed). issue — the action in one sentence. detail — the reason, without numbers. campaigns — the campaigns to act on, by their exact name as in the platform, separated by semicolons; null when the action is not about particular campaigns (a feed, a platform, a budget split). Actions name only campaigns that are running and spent money in the period; a paused campaign may explain a move but is never an action.

The figures in an evidence cell are written the same way everywhere: the platform or campaign, the metric, the value, and in brackets the change against the previous period and the previous value — for example "spend 1,200 EUR (+24 % vs 968 EUR)". In the issue and detail columns a campaign is called by a generic handle ("the winter campaign"); its exact name, as in the platform, is in the evidence and in the campaigns column, so the two can be matched. Every amount is in your reporting currency. The metrics are the platform's own: impressions (how many times the ads were shown), clicks, video views and completed views (watched to the end), link clicks, conversions (what you set as the result in each platform — an order, a lead, a call), spend, and the spend share (a campaign's share of its platform's spend in the period, or a platform's share of the whole account's); and the costs per result: CPM (per thousand impressions), CPC (per click), CPV (per video view), CPCV (per completed view), CPLC (per link click), CPA (per conversion). Their definitions are at https://airdan.ai/blogs/kpi-definitions/cpm-cost-per-mille, https://airdan.ai/blogs/kpi-definitions/cpc-cost-per-click, https://airdan.ai/blogs/kpi-definitions/cost-per-fully-completed-view-cpcv and https://airdan.ai/blogs/kpi-definitions/cpa-cost-per-action.

Three rules the report follows, which explain what you will and will not see in it:

  • Efficiency means the objective's own cost per result. A campaign bought for Awareness is judged on CPM, for Traffic or Engagement on CPC, for Video on CPV, for Conversions or Sales on CPA — never on another objective's measure, so an Awareness campaign with an expensive CPC is not a finding. Campaigns are compared only with campaigns of the same objective, on one platform or across platforms, and always with the volume of results next to the cost per result: "the most sales at the lowest CPA", never a league table of costs alone. Campaigns with no objective (unclassified) are never ranked; when they carry a large share of the spend, a warning asks for the objectives to be mapped — the Data › Campaign Objectives page shows the mapping.
  • Small numbers never make a headline. A change on a tiny base (less than 50 in your currency or less than 1,000 impressions in the previous period, or less than 50 spent now) is flagged as low volume and never carries a finding on its own; if it is mentioned at all, the evidence shows the base.
  • A missing feed is a data gap, never a performance change. When a platform's data ends before the period does, the report says so once — as a data_gap Fact, or as the Overview status when it dominates — and judges that platform only on the days its data covers. A platform whose last day of data is three or more days before the report counts as stale.

The weekly routine: what to look for and what to do

Go through the steps below once a week, on the day the new report arrives, in order. Each step tells you where to look and what to do with what you find. Write the actions you take on in a document or export the Actions table (see "Notes"); at the end you should have a short list of changes to make in the platforms, each with its campaigns, the change, the date you made it and the person who makes it.

  1. Read the verdict and check that the data is complete. How to find: page = executive, period = 30d; read the Overview row, then the Facts table for a data_gap row.

    • If status = data_gap, or a Fact of kind data_gap names a platform, the platform's data stopped arriving on the date in the evidence: tell us, we restore the feed — the platform's account access has usually expired or changed. Until it is restored, read that platform's rows for the period its data covers only, and do not act on its "declines".
    • If status = no_activity, nothing was spent in the period: if that is intended, there is nothing to read this week; if not, check in the platform whether the campaigns are paused or the budget exhausted.
    • If status = on_track, mixed or off_track, read the detail sentences for the biggest move and the direction of the next steps, then continue.

  2. Understand what moved and what stands out before you act. How to find: on the same page and period, read the Comparison table top to bottom, then the Facts table. For every row the evidence gives the figures; to see them day by day, open Digital Media › All Campaigns or the platform's own page (Digital Media › Google, › Facebook …) for the same 30 days.

    • If direction = up_bad or down_bad, the move is against you: the Actions table usually carries the response; read the two together.
    • If direction = neutral, money was moved on purpose (a shift between platforms, a pause): confirm it matches what you decided; if it does not, someone changed a budget you did not plan.
    • If kind = highlight, the campaign or platform deserves more: it will appear under increase or reallocate in the Actions table when the report is confident; if it does not, it is a candidate for your own test.
    • If kind = warning about unclassified campaigns, part of your spend sits in campaigns without an objective and is not judged at all.

  3. Carry out the actions, in priority order. How to find: the Actions table, sorted by priority; the campaigns column names the exact campaigns. Start with the high rows.

    • If verb = increase or decrease and the detail names a budget, change the daily budget of the named campaigns in the platform by a step you can afford — a fifth to a third of the current budget — note the date, and read next week's report before the next step. If it names a bid, change the bid or the target cost per result by a similar step.
    • If verb = reallocate, move the budget from the platform or objective the detail names to the one it recommends, by the same kind of step; the expected effect in the record says what should improve — that is what you check next week.
    • If verb = pause, pause the named campaigns in the platform; the evidence tells you what they cost for what result.
    • If verb = maintain, change nothing on the named campaigns: the report says they work; the action exists so that you do not touch them while you move other budgets.
    • If verb = investigate, open the named campaign in the platform and look at its setup — its objective, its budget, its schedule, its status — and at its daily figures on the platform's Digital Media page; the evidence tells you which figure does not add up. Decide, then write the change down like any other action.
    • If verb = fix_data, see step 1.

  4. Read each platform's own story. How to find: switch page to each platform name, keep period = 30d, and repeat steps 1 to 3. The platform pages carry the campaign-level findings; the executive page only carries what is true across platforms.

    • If the executive page recommends moving money between platforms, the platform pages tell you which campaigns on each side are the ones to grow or contain.
    • If a platform page's Overview says no_activity while the platform is running, its data has not arrived: step 1.

  5. Check the last 7 days for a reversal. How to find: switch period to 7d, on the executive page and then on each platform.

    • If the 7d Comparison shows a reversal — a move opposite to the 30d story (the report says so explicitly, in one row) — the week overrides the month: act on the 7d action and watch it weekly.
    • If the 7d and 30d stories agree, the 30d actions stand; the 7d set only tells you the move is still going on.
    • If a 7d row is flagged as low volume in its evidence, do not act on it alone: wait for the 30d set to confirm it.

  6. Check next week what changed. How to find: on the next report, read the Overview and the Comparison table first and compare them with the list of changes you wrote down, with their dates.

    • If an action you carried out no longer appears and the Comparison shows the expected move (a down_good on the cost per result, an up_good on the results), keep the change.
    • If the move is the opposite of the expected effect, go back one step on that campaign and tell us.
    • If the same action appears again, check that the change is live in the platform (budgets are set per campaign, and a change made in the wrong campaign is the usual cause).

Notes

  • Export a table. Hover over a table, click the three dots in its top-right corner and choose Export → Google Sheets with "Keep value formatting" ticked, or CSV. The export contains the rows you see, for the selected page and period, so you can keep the Actions table as this week's task list and pass it to whoever runs your campaigns.
  • How the data updates. The report is written once a week, on your review day, in the morning, as the last of the weekly reports; the page shows the latest report only, and the next one replaces it. The campaign figures arrive from each platform every night, so the windows end the day before the report: 7d is the last 7 complete days and 30d the last 30, each compared with the 7 or 30 days immediately before; the current day is never included. A platform whose data stops earlier is reported as a data gap, not as a drop.
  • What the numbers don't say. Conversions are what each platform counts as a result, and the platforms count differently — a sale in one, a lead in another, with their own attribution — so a cross-platform comparison always shows the volumes next to the cost per result, and you weigh them. Conversions can carry decimals (12.5) because a platform splits a conversion between several clicks. Completed video views are 0 on platforms that do not report them; that means not reported, not no completions. The change percentages are computed from the figures; the report never does arithmetic of its own, and every number it cites can be found on the Digital Media pages. The text of a record is written by an AI from these figures: the numbers are copied from your data, never invented; the verdict, the grouping and the priority are its judgment — check the evidence when something looks surprising.
  • What the report leaves out on purpose. It sees campaigns, not what is inside them: nothing about creatives and creative fatigue, audiences and targeting, ad groups and individual ads, quality scores, auction dynamics, landing pages, attribution changes or competitors. Its actions are budget, bid, pacing, scaling or pausing a campaign, investigating a campaign's setup, restoring a feed, and mapping objectives — nothing else.
  • Related pages. The figures behind every record, day by day and campaign by campaign, are on Digital Media › All Campaigns and on each platform's page (Digital Media › Google, › Facebook, › TikTok, › Pinterest); how the platforms and objectives compare on cost and efficiency is on Digital Media › Benchmarks. Which objective each campaign is mapped to is on Data › Campaign Objectives, and what was bought on Data › Campaign Types. For the keywords behind your Google search campaigns, go to Paid Search › Keywords and › Search Terms. If your plan includes Opportunity Gap (Insights menu), it tells you which searches to add to Google Ads and which to stop paying for.

Remember!

  • Read the executive page first, then every platform. The executive page never repeats a platform's findings; the campaign-level actions are on the platform pages.
  • Compare like with like. A cost per result means something only against campaigns bought for the same objective, with the volume of results next to it.
  • One change at a time, dated, checked next week. The report compares period with period; a change you cannot date is a change you cannot judge.