How do I know if my paid media is actually driving growth?
Your ads are working if sales would have been lower without them. The only way to know that is to measure incrementality: compare results where ads ran against a comparable group where they did not. Platform-reported ROAS shows what your ads touched, not what they caused.
Start with three checks. First, compare blended results (total revenue or new customers divided by total marketing spend) over time against what each platform reports. If the platforms together claim more conversions than you actually gained, attribution is double counting. Second, watch what happens when spend changes. If you cut a channel by 30% and total sales barely move, that channel was likely taking credit for demand you already had. Third, run a controlled test on your biggest channel, such as a geo-holdout or a platform conversion lift study.
A clean answer on the channel taking the most budget is worth more than a dashboard of attributed numbers across all of them.
What is incrementality, and how is it different from attribution?
Attribution assigns credit for a conversion to the ads a customer touched. Incrementality measures how many conversions happened only because of the ads. Attribution answers who gets credit. Incrementality answers what the spend caused, and only that answer tells you whether to spend more.
Attribution models (last click, data-driven, multi-touch) are useful for day-to-day optimization inside a channel, but they tend to credit ads for purchases that would have happened anyway, especially retargeting and branded search. Incrementality uses a control group that did not see the ads, so the difference in outcomes is the true lift.
A practical setup uses both: attribution for daily decisions, and periodic incrementality tests to calibrate it. If a test shows a channel drives half the conversions its platform reports, apply that factor to its reported numbers until the next test.
What is a geo-holdout test, and how do you run one?
A geo-holdout test turns advertising off in a set of matched geographic markets while keeping it on in others, then compares sales between the two groups. The gap, adjusted for normal differences between markets, is the incremental lift from the ads.
The core steps: (1) pick one clear question and one primary metric, such as new customers or revenue, (2) choose test and control markets that have historically moved together, (3) size the test so the expected lift is detectable, which means enough markets, enough spend and enough weeks, often three to six, (4) hold other major changes steady in those markets during the test, and (5) read the results with a method chosen in advance, such as synthetic control or difference-in-differences, and report a confidence range rather than a single number.
Most geo tests that fail, fail at design, with too few markets or too little spend to see a signal. Open-source tools such as Meta's GeoLift can help with market selection and power analysis.
When does a startup need marketing mix modeling (MMM)?
Later than most think. Marketing mix modeling becomes useful once you spend meaningfully across several channels, including hard-to-track ones like Connected TV, and have roughly two years of consistent weekly data with real variation in spend.
MMM is a statistical model that estimates how much each channel contributes to sales over time, while accounting for seasonality, pricing, promotions and outside events. It depends on inputs many early teams have not organized yet: clean weekly spend and impressions by channel, sales, price changes and a calendar of promotions.
Before investing in a model, get those inputs right and run a few incrementality tests. Test results are the best way to calibrate an MMM and check that it is telling the truth. Open-source options such as Google's Meridian and Meta's Robyn have lowered the cost of building one, but the hard part is still the data and the judgment, not the code.
How should I split my budget across marketing channels?
Allocate by marginal return, not average return. Put the next dollar where it produces the most incremental customers, and stop increasing a channel once its cost per incremental customer rises above what a customer is worth to you.
Every channel has diminishing returns. The first dollars reach the most likely buyers, and later dollars cost more for each new customer. That is why a channel with a strong average ROAS can still be overfunded.
A practical approach: rank channels by incremental cost per acquisition (using test results where you have them), set a target based on your payback goal, and shift budget in steps of 10 to 20 percent so you can see the effect of each move. Keep a test-and-learn reserve for new channels and creative so you are not only optimizing what you already know. Revisit the split monthly, and after any big change in pricing, creative or season.
How do I calculate CAC, LTV and payback period correctly?
CAC is total acquisition cost divided by new customers acquired. LTV is the gross profit a customer generates over their lifetime, not their revenue. Payback period is the number of months of gross profit it takes to earn back CAC. Use gross margin and fully loaded costs, or the numbers will flatter you.
The most common mistakes are calculating LTV on revenue instead of gross profit, leaving agency fees, tools, creative and team costs out of CAC, counting platform-attributed conversions instead of actual new customers, and assuming that early cohort retention will hold as you scale.
Build it by monthly cohort. For each month of new customers, track cumulative gross profit per customer and mark the month it passes CAC. That curve tells you more than a single LTV to CAC ratio. Targets vary by business model, but many subscription and DTC businesses aim to pay back CAC within 12 months, and investors will look closely at how the curve changes as spend grows.
Should I hire a fractional growth lead, a full-time head of growth or an agency?
Hire fractional when you need senior judgment before you can justify a full-time salary, an agency when you need hands to execute, and a full-time head of growth once growth is your main lever and the work fills a week. Many startups need the first before the other two.
Agencies are built for execution: running campaigns, producing creative and managing bids. They are rarely set up to question whether the budget itself is right. A fractional growth lead, typically one to three days a week for a set term, sets strategy, builds measurement, holds agencies accountable and often helps hire the eventual full-time leader. A full-time head of growth makes sense once you have product-market fit, meaningful monthly spend, and enough channels and people to manage.
A useful test: if no one on your team can explain what your last marketing dollar returned, you need senior judgment before you need more execution.
Is Connected TV (CTV) advertising worth it for a startup or DTC brand?
Connected TV can work for brands with a broad audience, a clear visual story and enough budget to test properly, but it is one of the hardest channels to measure with click-based attribution. Plan the measurement before you buy the media.
Connected TV is streaming television delivered to internet-connected TVs and devices, bought programmatically or directly from publishers. It reaches audiences who are hard to find on social and search, and it can be targeted more precisely than traditional TV. The catch is that viewers rarely click, so last-click attribution undervalues it.
Before launch, decide how you will judge it: a geo-holdout test, a matched-market comparison, or lift in branded search and direct traffic in exposed areas. Then confirm the basics that drive results: where ads actually run, frequency caps, completion rates and whether delivery matches what you bought. A focused test with a clear read beats a broad buy you cannot evaluate.
How do I know if my marketing agency is doing a good job?
Judge an agency on the business outcomes it moves and how honestly it reports them, not on platform metrics it controls. A good agency can show incremental results, tells you what is not working, and puts your budget where the next dollar returns the most, even when that means spending less.
Five questions worth asking: (1) Do they report blended, business-level results alongside platform numbers? (2) Have they proposed or run an incrementality test in the last six months? (3) Can they explain the account structure, targeting and creative testing plan in plain language? (4) Do their fees reward spending more rather than spending better? (5) Do you own your ad accounts, data and creative?
An independent audit of the accounts and the reporting, usually a few weeks of work, answers most of these and gives you a fair basis for the next conversation with your agency.
What does working with a growth and measurement advisor look like?
At Andrashko Advisory, most engagements start with a fixed-fee diagnostic: two to three weeks reviewing your spend, measurement and growth model, ending with a written readout of what is working, what is not and what to do next. From there, clients choose a defined project, ongoing advisory or a fractional growth lead.
The work spans six practices: paid media and advertising, measurement and incrementality, financial modeling and analysis, growth and operating advisory, go-to-market and new business, and AI workflow enablement. The best fit is a venture-backed consumer, DTC, subscription, fintech or B2B SaaS brand spending roughly $100K or more a month on paid media without a senior growth or measurement lead.
Matt Andrashko brings 12+ years, the last six agency-side, including a $75M+ annual paid media portfolio across 30+ brands and $20M+ a year allocated on incrementality and geo-holdout tests. The easiest first step is a 30-minute intro call.