Guides
Articles in guides from the ROASfinity blog.
Shopify
Google Ads
Slack
Meta Ads
Stock photos and example labels, not customers.
In this section
What AI ad assistants can and cannot do safely
Assistants are good at reading numbers and drafting changes. The risk starts when a draft turns into a live change nobody reviewed.
6 min read→ Guides
Read-back explained with an example
Reading an object back after a change sounds like extra work until the day it catches a change that was only half applied.
5 min read→ Guides
MER vs ROAS on one page
If a client only reads one paragraph of your report, it should explain why two numbers that sound alike disagree.
4 min read→ GuidesHow to audit your agency tool stack
Most agencies did not design their stack. It grew one tool at a time, and nobody has written down which tool is the source for which number.
6 min read→ GuidesBreak-even ROAS: a worked example with made-up numbers
The formula fits on one line. The work is deciding which costs belong in the margin.
5 min read→ GuidesHow to run a simple holdout test for an ecommerce store
When you must know whether a channel works, attribution is not enough. A holdout test is the simplest honest answer.
6 min read→ Guides
Three permission levels for AI assistants in ad accounts
Giving an assistant access is not one decision. It is three, and each one needs its own conditions.
5 min read→ Guides
What to do when MER drops
A falling MER is a symptom. The fastest way to make it worse is to cut budget before you know the cause.
6 min read→ Guides
A creative test budget, worked through with made-up numbers
The arithmetic is simple. The discipline is deciding the rules before the first euro is spent.
4 min read→ Guides
How to set a MER target with a client
A target that nobody can derive will be argued about every month. A target built from the margin can be checked.
4 min read→ Guides
How to read a cohort table for an ecommerce store
A cohort table shows whether customers come back. Read it row by row, and compare groups at the same age.
5 min read→ GuidesHow to report a promotion without flattering the numbers
A promotion lifts every number on the dashboard. The report has to say how much of that lift the ads can actually take credit for.
5 min read→ GuidesSetting ad targets when shipping costs change
A shipping change moves the break-even line. If the target stays where it was, the ads look better or worse than they really are.
4 min read→ GuidesPlanning ad budgets around seasonality
A seasonal peak can make a weak month look strong and a normal month look weak. Plan against the calendar, not against the last four weeks.
5 min read→ GuidesReading MER in a subscription store
In a subscription store the first order is only the start. A single day of MER does not show whether the customer ever pays again.
5 min read→ Guides
How to build a hook library for ad creative tests
Most teams rewrite their openings from scratch every month. A hook library turns each test into something the next brief can use.
5 min read→ Guides
How to brief a video editor for ad variants
A good editor can turn out many versions in a day. Without a brief, those versions test many things at once and teach nothing.
5 min read→ GuidesReviewing AI-generated ads in ten minutes
AI drafts ads quickly. The review is what stops a quick draft from becoming an expensive mistake.
4 min read→ GuidesWriting prompt standards for creative teams
When every writer keeps a private prompt, the output changes with the person. A standard makes the work repeatable and easy to review.
5 min read→ Guides
A brand safety checklist for AI-made ads
Brand safety is mostly about what is never said. A short checklist, run on every batch, catches most of the problems before they reach a customer.
4 min read→ Guides
Ten numbers for a client dashboard, in reading order
A dashboard with thirty numbers gets skimmed. A dashboard with ten defined numbers gets read, questioned and acted on.
4 min read→ Guides
How to read a CPM change in an ad account
CPM is the price of a thousand impressions. When it moves, the cause is usually the audience, the mix of places or the competition for the same people.
4 min read→ GuidesHow to read a CTR drop in an ad account
CTR is the share of impressions that get a click. When it falls, the ad is the first suspect and often the last one to check.
4 min read→ Guides
How to read a conversion rate change for an online store
Conversion rate is orders divided by sessions. A change in it is easy to blame on the ads. The order of checks decides whether you find the real cause.
5 min read→ Guides
A five-minute daily check for an ecommerce ad account
Most ad problems grow for days before anyone looks. A short check each morning catches them while they are still small.
4 min read→ GuidesHow to read a new customer rate change in an online store
The new customer rate is the share of orders from people who have not bought before. It moves when the ads bring new buyers, and also when the repeat base changes size.
4 min read→ Guides
How to read an average order value change in an online store
Average order value is revenue divided by orders. A higher number feels like progress, but it can come from one large order, a discount rule or a shift in what people buy.
4 min read→ GuidesHow to read a return rate change in an ecommerce store
A return rate that jumps can mean a product problem, a sizing problem or returns that simply arrive late. Each one needs a different action, so find out which one it is first.
4 min read→ GuidesHow to read the gap between platform ROAS and blended ROAS
Platform ROAS and blended ROAS rarely match. The gap is not an error to fix. It is a set of differences to explain before any budget decision is made.
5 min read→ GuidesWhat a payback window tells you about ad spend
A payback window is the time it takes a group of new customers to return their acquisition cost in margin. It shows how long money is tied up, which matters as much as the ratio.
4 min read→ GuidesHow agencies set spend guardrails for client ad accounts
A guardrail is a limit set before anyone needs it. It stops one bad day, or one bad change, from turning into a bad month.
5 min read→ GuidesA simple weekly meeting rhythm for an ad agency
Most agencies meet too often and decide too little. A short, fixed rhythm gives each meeting one job and one written output.
4 min read→ GuidesHow to run an ad account audit in one hour
A full audit can take days. A focused hour answers the questions that cost the most money when they go wrong.
6 min read→ GuidesWhat to log when an AI assistant changes an ad
An AI assistant can change an ad in seconds. The log has to make that change as traceable as one made by a person.
5 min read→ Guides
How to split work between people and AI in an agency
The question is not whether to use AI. It is which steps an assistant can draft, which steps a person must own, and where the line sits in writing.
5 min read→ GuidesHow to triage a sudden spend spike in a client ad account
A spend spike is an alarm, not a diagnosis. Confirm the number first, then find out whether the money was well spent before anything changes.
5 min read→ GuidesHow to explain a flat week to an ecommerce client
A flat week is not a failure, but it still needs a clear answer. Say what happened, what did not change, and what you will check next.
4 min read→ GuidesReading blended and platform numbers together without double counting
Blended numbers and platform numbers answer different questions. Put them side by side with their definitions, or they will be argued about forever.
5 min read→ GuidesHow to build a promo calendar for an ecommerce client
A promo calendar shows every offer before it runs, so spend, margin and reporting can be planned together instead of argued after the fact.
5 min read→ GuidesHow to write a metric tree for an ecommerce ad account
A metric tree starts from one number the client cares about and breaks it into the parts that move it. It shows where to look when that number changes.
5 min read→ GuidesA review loop for AI-written ad copy before it goes live
An AI draft is fast, and a fast draft can carry a wrong claim into a live ad in minutes. A fixed review loop with a named reviewer keeps the speed and the control.
4 min read→ Guides
How to write an account handover note that the next person can act on
A handover fails when the next person has to guess. A short written note, in a fixed order, lets the work start on the first day.
4 min read→ GuidesNaming ads when many people launch in the same week
Two people with the same naming habit still create a mess in the reports within a week. A shared rule with a registry prevents most clashes.
4 min read→ GuidesWhen to roll back an ad change, and when to wait
Rolling back a change feels like a fix, but it is also a change. It needs the same discipline as the change it reverses.
4 min read→ GuidesA plain service level for client work, on one page
A service level tells the client what to expect, in plain words. Written once, it prevents most arguments about what was promised.
4 min read→ GuidesHow to write ad copy variants that test one thing at a time
A variant that changes the headline, the offer and the picture at once cannot tell you which change worked. Good variants change one thing and say what it is.
4 min read→ GuidesHow to decide what to test next in ad creative
Most teams have more test ideas than budget. A short scoring habit turns a long list into one test that is worth running this month.
4 min read→ GuidesWhat to do with a losing ad creative
A losing creative is not automatically a bad creative. Check whether the test was fair before you pause it, and write the reason down either way.
4 min read→ GuidesHow many ad creatives to launch at once
Launching more creatives gives more ideas at once, but it also splits the budget. The right number comes from the budget, not from the length of the idea list.
4 min read→ GuidesHow to keep a creative learnings log that people use
Creative tests teach a team a lot, then the lessons fade when people change. A log with a fixed format keeps each lesson where the next person will look.
4 min read→ Guides
How to run a client dashboard review, step by step
A dashboard review that ends with 'looks fine' wastes the client's time. A fixed order turns the same screen into decisions.
5 min read→ GuidesHow to handle urgent client requests without breaking the week
Every urgent request feels like the most important one. A fixed triage rule keeps real emergencies first and stops the rest from taking over the week.
4 min read→ GuidesSetting AI usage rules for your agency team
Many AI mistakes in agencies start with unclear rules, not with the tools. One page with clear limits gives the team a shared default.
5 min read→ GuidesHow to train a new account manager in the first 30 days
A new account manager learns from the client's numbers and the team's habits. A fixed plan puts both in front of them, in a set order.
5 min read→ Guides
How to keep client data clean before an AI assistant reads it
An AI assistant answers from the data it can see. Messy names, mixed definitions and old exports produce confident wrong answers, so cleaning comes first.
5 min read→
Built for shop owners, creators and agencies. Stock photos, not customers.Run the next client review from one screen.
We open one agency at a time, with your own accounts connected read-only first.
Request access