Run a mock round

ROI is the incremental profit attributable to a campaign divided by its cost. The hard part is the word incremental.

I start by agreeing on the metric that represents value, revenue, margin or pipeline, and the attribution window. Then I try to isolate the campaign's true effect. Where possible I use a holdout or geo test, because correlation between spend and sales is weak evidence.

I calculate fully loaded cost, including media, creative, tools and people, not just ad spend. For brand campaigns I use proxy measures such as branded search lift and assisted conversions, and I am honest that long term brand effects are measured over quarters, not weeks.

I report ROI with the assumptions and confidence level, and I compare against alternative uses of the budget. A campaign with a three times return is only good or bad relative to what else the money could do.

Click through rate is clicks divided by impressions. Benchmarks vary widely by channel and industry, so what matters most is the trend and the comparison against your own previous ads and your competitors.

To improve it I work through the main levers. Audience: is the ad shown to people with real intent? I tighten targeting and exclude irrelevant placements. Creative: the headline and image must match the searcher's exact problem, so I mirror their language and lead with the benefit. Offer: a clearer, lower friction promise, such as a free tool rather than a generic demo. Landing alignment: the ad and the page must promise the same thing.

I test one variable at a time, usually the headline first, then the creative, and I let each test run long enough to be confident. I watch cost per click too, because a high click rate with a high bounce is not real success.

CAC is the cost of acquiring one customer. LTV is the profit that customer generates over their lifetime. Their ratio tells you whether growth is sustainable.

I calculate CAC as total sales and marketing cost divided by new customers in a period. I calculate LTV using average revenue per account, gross margin and churn, so LTV equals average revenue per account times gross margin divided by churn rate. Using margin matters because revenue is not profit.

A healthy benchmark is LTV at least three times CAC, with payback under twelve to eighteen months. Below that, growth burns cash faster than it creates value.

The ratio also shows where to act. High CAC may mean the channel is saturated or the offer is weak. Low LTV may mean poor retention or underpricing. I track both alongside churn, because rising CAC and rising churn together are an early warning.

I would not choose on gut feel. I would decide using payback, intent and time horizon.

Paid search is fast and controllable but stops the moment spend stops. SEO compounds and can lower long run acquisition cost, but it takes months and is harder to attribute.

I would model both. For keywords with high commercial intent and proven conversion, paid often makes sense immediately, especially if payback is under a target such as twelve months. For informational, high volume topics where we have topical authority, SEO is usually the better long term investment.

I would run a small paid test to measure true conversion rate on a keyword cluster, then use that data to decide whether to invest in content for the same terms.

The answer is usually a portfolio: paid for demand capture now, SEO for demand generation later, reviewed quarterly.

Organic reach is the number of people who see your content without paid promotion. It comes from followers, shares and the platform's ranking algorithm, and it is essentially free but unpredictable and shrinking on most networks.

Paid reach is the number of people who see content because you paid to distribute it, through boosted posts, search ads or display. It is predictable, targetable and scalable, but it stops when the budget stops.

The two work together. Organic builds trust and authority and validates which messages resonate. Paid amplifies the winners and reaches audiences beyond your existing followers.

In practice I use organic content as a testing ground, then put budget behind the posts and keywords that already show strong engagement, rather than paying to distribute a message nobody has validated.

Audience targeting decides who sees an ad, and it matters because showing the right message to the wrong people wastes money and damages relevance scores.

It works on a few dimensions. Demographics cover age, location, income and job title. Interests and behaviours group people by what they read, buy or engage with. Intent targeting reaches people actively searching for a solution. Retargeting reaches people who already interacted with you, and lookalike audiences find new people similar to your best customers.

Good targeting narrows enough to be relevant but not so much that costs rise or the audience is exhausted. I usually start broad with strong creative, then let data show which segments convert, and I layer exclusions so I am not paying to reach existing customers.

Targeting and creative must match, or even a well targeted ad feels irrelevant.

A SaaS funnel maps the stages a user passes through, and each stage has its own metrics.

Awareness: impressions, reach and cost per thousand. Are we in front of the right audience?

Acquisition: visitors, click through rate, cost per click and cost per acquisition. Are we buying attention efficiently?

Activation: sign ups and the percentage that reach the first value moment, such as creating a first project. Activation is often where the biggest leaks are.

Revenue: conversion to paid, average contract value and monthly recurring revenue. Retention: churn, net revenue retention and expansion.

I focus on the stage with the largest drop-off relative to benchmark, because a small improvement near the top multiplies through the whole funnel. I report one primary metric per stage so the team knows what it is optimising.

I start with a hypothesis tied to a metric: for example, moving social proof above the form will lift sign up conversion by ten percent because trust is the main blocker.

I define one primary metric and the minimum sample size before starting, so I cannot stop early when I like the result. I split traffic randomly and keep everything else identical, changing one meaningful element per test.

I record the baseline conversion rate and run until the test reaches significance. I check segment results too, such as mobile versus desktop, because an average win can hide a loss on mobile.

If the variant wins I implement it and start the next test. If it loses, the learning still has value. I maintain a testing backlog and document results so we compound knowledge instead of repeating experiments.