How do you know if paid ads are working?

Paid ads are working when they return more value than they cost — measured in real revenue or durable customer behavior, not vanity metrics like impressions or clicks. The trap most early-stage founders fall into is optimizing for signals that feel like progress but don't connect to the unit economics that determine whether the business survives. Before you can answer whether ads are working, you need three numbers: your cost per acquisition (CPA), your customer lifetime value (LTV), and your payback period.

The only question that matters: does LTV exceed CPA?

Every paid channel is ultimately a machine that converts dollars into customers. The machine is working if you get back more than you put in, on a timeline your cash position can survive. This means you need to know your LTV before you can meaningfully evaluate any ad spend. If your LTV is $200 and your CPA is $180, ads are technically 'working' — but you have almost no margin for error, creative fatigue, or platform volatility. If your LTV is $1,200 and your CPA is $180, you have a real growth lever.

The mistake early founders make is launching ads without a confirmed LTV baseline. If you haven't run the product long enough to know 6-month or 12-month retention, you're guessing at the denominator. In that case, use a more conservative leading indicator: first-order gross margin per customer, or subscription revenue at 90 days. Anchor ads to a number you can actually observe, not one you're projecting.

Payback period matters as much as the ratio. A 3:1 LTV:CPA ratio is meaningless if it takes 24 months to recoup acquisition cost and you have 8 months of runway. Model your payback period explicitly. For most early-stage consumer products, a 6-month payback is healthy. For B2B SaaS with annual contracts, 12 months is often acceptable. Know your number before you scale.

Which metrics to track at each stage of the funnel

The funnel from ad impression to retained customer has four stages: reach, click, conversion, and retention. Each stage has its own failure modes, and diagnosing where the funnel breaks is how you fix it.

At the top, CPM (cost per thousand impressions) and CTR (click-through rate) tell you whether your creative is resonating with the audience you're targeting. A CTR below 1% on Facebook or Instagram usually means your creative isn't stopping the scroll — the message doesn't match what the audience cares about. A high CTR but low conversion rate means the ad over-promised, or the landing page fails to continue the story the ad started.

At the bottom, track conversion rate on the landing page and CPA by campaign, ad set, and creative. Don't aggregate too early — a single high-performing creative can mask three losing ones, and you'll burn budget on the losers until you separate them. Once someone converts, track their behavior at 7, 30, and 90 days. Are they activating? Returning? Referring? Customers acquired through paid often have different retention profiles than organic customers, and that difference matters enormously for your LTV assumptions.

How to run a clean test before scaling

Scaling a channel that isn't proven yet is one of the most reliable ways to destroy startup capital. Before you increase budget, run a structured test: pick one audience, one offer, and three to five creative variants. Run for long enough to get statistical significance at the conversion level you care about — usually a minimum of 50 conversions per variant. Don't call it early because one variant 'looks good' after 10 conversions.

Control as many variables as possible during testing. If you change your landing page and your ad creative at the same time, you won't know which change moved the needle. Change one thing at a time. This sounds obvious but almost no one does it when they're moving fast and excited about new ideas.

Garry Tan's engineering principle of tying technical choices to real user outcomes applies directly here: every ad test should have a pre-defined success metric and a pre-defined timeline before you look at results. Deciding what 'good' looks like after you've already seen the data is how confirmation bias gets expensive. Write it down first: 'This test succeeds if CPA is under $X with at least 50 conversions in 14 days.' Then don't touch it.

Warning signs that ads are not working — even when the dashboard looks fine

Platform dashboards are optimistic by design. They attribute conversions using the most favorable window, smooth out volatility, and surface the metrics that keep you spending. Here are the signs that your ads aren't actually working, even when the numbers look decent on screen.

First: your organic baseline isn't growing. If you turn off paid and conversion volume collapses entirely, you've bought transactions, not a business. Healthy paid acquisition should reinforce a brand that's also building organic demand over time. Pure paid dependency means you're renting customers, not earning them.

Second: cohort LTV is declining by acquisition channel. Pull LTV by cohort and break it out by paid versus organic, or by campaign if you have the data. If paid cohorts churn faster or spend less over 90 days, you're acquiring lower-quality customers — often because targeting has drifted toward people who respond to discounts or aggressive creative but don't actually have the problem your product solves.

Third: you're dependent on one creative or one audience. If 80% of your paid results come from a single ad or a single audience segment, you have fragility, not a working channel. Platforms change, audiences saturate, and creative fatigue is real. A working paid channel has redundancy across multiple offers and multiple audiences.

Fourth: you've never turned it off. The only clean way to measure paid incrementality is to run a holdout test — a period where you pause spend for a segment and measure what happens. If you've never done this, you don't actually know what your ads are contributing versus what would have happened organically anyway.

When to keep spending versus when to stop

The decision to scale, hold, or cut paid spend should follow a simple framework: if CPA is below your acceptable threshold and payback period fits your runway, scale cautiously — increase budget by 20-30% at a time, not 2x, because algorithms need time to recalibrate and costs often spike with aggressive scaling. If CPA is at or above threshold, pause and fix the funnel before adding money. If CPA has been above threshold for more than two testing cycles and you haven't found a combination that works, consider whether the channel is right for your product at this stage.

Not every product category responds to paid at early stages. Paul Graham's observation that truly promising work often looks uncertain from the outside applies to channel strategy too — the fact that a competitor is running aggressive paid campaigns doesn't mean paid is the right move for you right now. Some products grow through word of mouth, integrations, or content before they're ready to scale paid. Forcing paid onto a product that hasn't found its natural growth motion is expensive and demoralizing.

The clearest signal to keep spending is this: when you increase budget, CPA stays roughly flat and the downstream cohort quality stays the same or improves. That's a real signal. When CPA rises sharply with budget increases, you've saturated your best audience and you're reaching lower-quality prospects to maintain volume. That's the moment to invest in expanding your creative library and audience strategy before scaling further.

“It's more promising if something you're interested in is difficult, especially if it's more difficult for other people than it is for you.”

— Paul Graham, source

The one thing to do

Before you spend another dollar on ads, write down your acceptable CPA, your confirmed LTV, and your payback period — then evaluate every campaign against those three numbers, not platform-reported ROAS.

Frequently asked questions

What's a good LTV:CPA ratio for early-stage startups?

A ratio of 3:1 is a commonly cited floor, but what actually matters is whether your payback period fits your runway. A 2:1 ratio with a 3-month payback can be healthier than a 5:1 ratio with an 18-month payback if you're capital-constrained.

How long should I run a paid ad test before deciding it works?

Run until you have at least 50 conversions per variant at the stage of the funnel you're optimizing for. Time minimums matter too — at least 7 days to account for day-of-week variation, and 14 days is safer for most campaigns.

Should I trust the ROAS number my ad platform reports?

Use it as a directional signal, not a ground truth. Platform-reported ROAS uses attribution windows that typically over-credit ads for conversions that would have happened anyway. Always cross-reference with your own backend data and consider running holdout tests to measure true incrementality.

When is it too early to run paid ads?

If you don't have a confirmed conversion rate on your landing page from organic or direct traffic, paid will just amplify an unproven funnel at high cost. Get to a conversion rate you can benchmark first, even if that means sending 200 people manually before running a single paid dollar.

Sources

More playbook answers · Growth Prophet home