Paid ads judged on what they sold

Measurement first, because the rest is unfalsifiable without it

AI ads management is paid media run against enquiries and revenue instead of clicks. In practice that means fixing the measurement first, producing enough creative to keep testing, and reporting in a form that survives a hard question from the person paying. For businesses in Pakistan it also means accounting for the enquiries that arrive on WhatsApp and never touch a conversion tag.

If the platform reports forty conversions and the CRM shows nine enquiries, one of those numbers is fiction and both are being used to make decisions. Until they can be reconciled, every optimisation is being made against a signal nobody has verified.

That reconciliation is the first piece of work. Conversion events firing on delivery rather than on click, a real enquiry traced end to end, and a shared definition of what counts as a lead. It is unexciting and it changes the outcome more than any bidding strategy.

• Platform conversions reconciled against records in the CRM
• Events that fire when an enquiry is received, not when a button is pressed
• Offline conversions fed back, so the platform optimises toward buyers rather than form-fillers
• One agreed definition of a lead, used by sales and by the ad account

Creative volume is the lever most accounts never pull

Modern ad platforms decide most of the targeting themselves. What is left under an advertiser's control is the offer, the audience signal and the creative, and of those the creative is the one that can be varied fastest.

This is where AI earns its place. Producing thirty genuine variations in the time a team previously produced three means the account can keep testing instead of settling. The judgement about which ideas are worth testing stays human; the production does not have to be.

The account structure argument, settled practically

There is a long-running argument about consolidated versus granular campaign structures. In practice it resolves by volume. Below a certain number of conversions a week, splitting the account starves every segment of the data it needs, and the algorithm never learns anything.

Most small and mid-sized advertisers are below that line and are running structures designed for accounts twenty times their size. Consolidating usually improves results immediately, and it is free.

Reporting that survives a hard question

A report should state what was spent, what came back, and what is uncertain. The uncertainty matters: attribution is imperfect, view-through is contested, and anyone presenting a single confident number is presenting a choice of model as though it were a measurement.

The version that earns trust says what the platform claims, what the business's own records show, and where the two disagree. Being open about the gap is what makes the rest of the report credible.

How the fee is structured, and why that matters here

Paid media is recurring work permanently, so it sits outside the four ownership models the rest of the systems use. There are two ways to buy it. Either your team is trained to run it, with a one time build fee and a small retainer afterwards, or Wobble runs it and you rent the output as an agreed monthly list of campaigns and creative.

The first is the wrong choice if nobody in the business will ever pick it up. The second is the wrong choice if you want to own something at the end. Both are set out against each other in the four ways to own this.

Whichever of the two you take, there is a monthly figure, and its structure is worth arguing about before you sign. A percentage of spend pays the provider more when you spend more, which is a poor incentive on the days when the correct advice is to spend less. A flat monthly fee removes that, and turns the conversation into one about results rather than budget.

What moves it: how many platforms are being run, since each is its own account structure and its own reporting; how much creative is produced each month, which is the lever most accounts underuse and the largest variable input; and whether conversion tracking already works, because an account that cannot attribute a sale cannot be optimised and fixing that comes first.

Before quoting, anyone should look at your account history and your current tracking setup. Then ask what their fee does when your spend halves. If the answer is that it halves too, you have learned something useful about whose interest the advice will serve.

• How many platforms and accounts are in scope
• How much new creative is produced each month
• Whether conversion tracking and offline sales are already recorded correctly
• Whether the fee is flat or a share of spend, and what happens to it when spend drops

What paid media cannot fix

If the page the traffic lands on does not convert, paid media buys the same failure at a faster rate. Fixing the destination first is almost always cheaper than raising the budget.

If margin cannot support the cost of acquisition in the category, no amount of optimisation closes that gap. That is arithmetic rather than execution, and it is better established before a budget is committed than after three months of it.

What the machine is not allowed to change, and who is on the account

Budget moves, bid strategy changes and anything going live in your name are prepared automatically and released only after human approval. That is not caution for its own sake. An automated rule reacting to a two day dip by cutting spend on the campaign that was working is the most common way an account gets quietly wrecked, and the damage stays invisible for a fortnight because the reported cost per click improves while the revenue does not.

So the split is written down before anything runs. Reporting, anomaly detection, creative variant generation and audience refreshes run unattended. Spend changes, new offers and any public reply to a customer comment wait until somebody signs them off. When something unusual happens, the alert names the campaign and the change it wants to make rather than making it.

Ali owns marketing and sales at Wobble and is the person on the account. The published examples sit on the work page: a Google Search and call-only engine built for Zavcom, a United States internet provider, at an average cost per click of $1.54 across 89 conversions in ninety days, and intent-segmented search for Amari Sayulita, a boutique hotel in Mexico. Wobble is answerable for what those accounts do, month to month, with no commitment beyond the month you are in.

Running it in-house is the right answer more often than an agency page will admit. If one person can own the account, has the taste to judge creative, and looks at it three times a week, keep it inside. The case for outside help is creative volume and the measurement layer underneath, which is where most internal accounts are actually stuck. Either way the ad accounts, the pixel and the historical data stay in your own accounts and never sit inside an agency container.

Common questions

Why do my ad platform conversions not match my CRM?

Usually because the platform counts an event that fires on click or on form submission rather than on an enquiry actually being received, and because attribution windows differ between the platform and the business's own records. Reconciling the two before optimising anything is what stops the account being tuned against a number nobody has verified.

How is AI used in paid media management?

Mostly for creative production volume and for reading results. Producing thirty genuine ad variations in the time a team previously produced three lets an account keep testing rather than settling. It is also useful for clustering which messages worked and for drafting the analysis. Judgement about what to test stays human.

Should campaigns be consolidated or split into many ad sets?

It resolves by volume rather than by preference. If any ad set produces conversions in single digits per week, it is not generating enough signal to learn from, and merging it into a larger one usually outperforms optimising it. Many smaller advertisers run structures designed for far larger accounts.

What is a realistic cost per lead?

A realistic cost per lead varies so widely by category, market and offer that any figure quoted before looking at your account is a sales device. The number that matters is not cost per lead anyway; it is cost per lead that sales accepted, which is often two or three times higher and is the only version connected to revenue.

Can you run ads for both Pakistan and international markets?

Yes, and they behave differently enough that treating them as one programme is a common and expensive mistake. Cost per result, competitive density, payment behaviour and the channel people expect to be contacted on all differ, so budgets, creative and follow-up are planned per market.

How long before paid media results are readable?

Enough conversions have to accumulate for the difference between variants to be distinguishable from noise, which is a volume question rather than a calendar one. A high-volume account can read a test in days; a low-volume one may take a month, and reading it sooner produces confident conclusions that do not repeat.