Reporting and optimisation decisions

0:00 Reporting and optimisation

Ads Manager has hundreds of possible columns. TikTok, LinkedIn and Snapchat have hundreds more. It's easy to spend an hour staring at numbers and walk away with nothing but a headache. Great paid social managers don't look at more numbers. They look at the right numbers in the right order, and they decide in advance what they'll do when those numbers move. In this lecture you'll learn how to build a standard column set, how to diagnose problems top-down, how to write decision rules, how to report attribution honestly, and how to build a weekly report sheet, with a few automated alerts and AI help for the commentary.

0:47 Standard column set

A good report answers three questions. What happened? Why? What will we do next? Everything else is supporting detail. Start with a standard column set, saved as a preset, so you read the same metrics every time. Delivery: spend, impressions, reach, frequency and CPM. Creative: hook rate, hold rate, link click-through rate and link cost per click. Conversion: results, cost per result, conversion value and ROAS. And funnel: landing page views, add to carts and checkouts. Use link clicks, not all clicks, because all clicks include likes, profile taps and people expanding the text. Link clicks show real intent to visit your site.

1:32 Diagnose top-down

When results change, diagnose from the top down. Think of it like checking a car that won't start: fuel first, then battery, then engine. One: did spend or delivery change, maybe a budget edit, a learning reset or a policy issue? Two: did CPM change, maybe seasonal competition or audience saturation? Three: did click-through change, maybe creative fatigue or a weak new ad? Four: did the landing page view rate change, maybe page speed or a tracking break? Five: did the conversion rate change, maybe the offer, stock, price or a website bug? Here's the key idea. Many ads stopped working problems are really website, stock or pricing problems.

2:19 Example 1: London bakery (illustrative)

Let's do a simple worked example with illustrative numbers. A London bakery's click-and-collect campaign sees cost per order jump from six pounds to eleven in a week. Top-down check. Spend: stable. CPM: flat. Click-through: stable. Landing page view rate: stable. Conversion rate: halved. So the problem is after the click. The team places an order themselves and finds the online ordering system has started forcing people to create an account. The fix is on the website, not in Ads Manager. They remove forced sign-up, and cost per order heads back towards six pounds. No ad was changed.

3:01 Decision rules

Now decision rules. Write them down before you need them, so emotions don't make the call. Pause an ad when it spends two to three times your target CPA without a conversion, or its CPA stays far above target after enough spend. Scale an ad set when CPA stays below target for several days, and increase budget in steps. Refresh creative when frequency rises and click-through falls for a sustained period. And investigate when platform conversions and store or CRM data diverge sharply. You can automate some of these as notifications in Ads Manager or TikTok's automated rules. Start with alerts, not automatic actions, and keep the rules few and simple.

3:49 Honest attribution

A word on attribution in reports. Each platform reports conversions using its own attribution setting, for example seven-day click and one-day view. Always note the setting in your report, keep it consistent when comparing periods, and compare against your source of truth, your store or CRM. Add blended metrics like MER, total revenue divided by total marketing spend, and new-customer CAC from store data. They're blunt but hard to fool. If a platform's ROAS climbs while MER and new customers stay flat, the platform is probably taking credit for sales that would have happened anyway.

4:30 Hands-on: weekly report sheet

Now build the weekly report sheet. Export a weekly breakdown by ad into a tab called Ads. In a Report tab, SUMIFS formulas total spend, purchases and impressions for the chosen week, and calculate CPM, link click-through, landing page view rate, hook rate and platform CPA. Type in new customers from your store, and calculate blended new-customer CAC. Then add a ranked creative table using SORT and FILTER, with conditional formatting: green when CPA is under target, red above one and a half times target. All the formulas are in the lesson text. Ten minutes to set up, and then every Monday takes five minutes instead of an hour.

5:17 Example 2: Karachi fashion (illustrative)

Now a realistic Monday review, with illustrative numbers. A Karachi fashion brand's platform CPA rose from nineteen hundred rupees to twenty-seven hundred. Top-down. Spend and delivery: unchanged. CPM: up eight percent, mild seasonal pressure. Link click-through: down from one point four percent to zero point nine on the two biggest ads, and their frequency is above four. Landing page views and conversion rate: stable. Diagnosis: creative fatigue on the top two ads, not a website problem. Action: launch three new hooks on the winning concept, promote one new concept from testing, and leave budgets alone for a week. The next week, CPA heads back towards target.

6:03 AI and the client report

Can AI help with reporting? Yes. Paste the weekly report table, aggregated and with no personal data, into an assistant and ask for a draft in the what, why, next format, using only reasons supported by the numbers and your notes. AI is good at spotting which metric moved. You supply the context it can't know, like a stock-out, a site change or a public holiday. Then verify every figure. And make the report readable for a busy owner: three summary bullets, KPIs against target, top and bottom creatives with thumbnails and reasons, tests and learnings, and next steps, in plain language like: people click but don't buy because delivery costs appear late.

6:52 Mistakes and measures

Common mistakes. Reporting all-clicks click-through rate. Changing attribution settings between reports. Reacting to one bad day. And reports full of numbers with no insight and no next steps. How do you measure success? Every weekly report ends with dated actions and owners, and last week's actions are marked done or not done. Decisions follow the written rules rather than the mood of the meeting. And platform and store numbers are reconciled monthly, with the ratios explained. When that's in place, reporting stops being a chore and becomes the engine of improvement.

7:32 Recap and try this now

Let's recap. Good reports answer what happened, why, and what we'll do next. Use a standard column set with link metrics. Diagnose top-down: delivery, CPM, click-through, landing page view rate, conversion rate. Write decision rules in advance, automate alerts before actions, note attribution settings and cross-check with MER and store data. Here's your try this now. Build the weekly report sheet from the lesson text with last week's data, run a top-down diagnosis on your biggest campaign, and write three decision rules for next week, each with an owner.