Are You Using the Right Measurement Method for Your Spend Level?

There are many different ways to market your business, and correspondingly there are lots of ways to measure your marketing’s effectiveness.

This is a rough roadmap for companies who are looking to understand what marketing measurement tactics they should be considering, whether they are about to spend their first dollar on media, or they regularly spend $50M or more.

A note before you read: this is a linear sequence, where each stage builds on the last. It’s ordered based on when methods become statistically viable, and are cost-justified based on spend volume. But viability and cost aren’t the only dimensions that matter.

The question you need to ask at every stage: what is the most important question I cannot currently answer about my marketing, and what is the cheapest and fastest way to get a reliable answer to it?

That question and this sequence will produce the same answer for many companies, but not all. There are real exceptions, particularly for B2B companies, funded startups, and many category-specific businesses. Always keep in mind your business model, channel mix, organizational maturity, competitive context, and sales cycle.

Stage 1: The Data Foundation

Annual Media Spend: $0 – $10K

At this level of spend, you aren’t going to generate any meaningful statistically significant signal. Measurement here is almost entirely about not making mistakes. Start with a clean data infrastructure, so that when you grow, you’ll have something to work with.

Don’t invest in any paid measurement tool at this point. You don’t have enough spend to generate the data volume that any legitimate model needs. Put your money into creative or additional spend.

The most important thing to do now is get your basic analytics configured and set up correctly. Fixing it retroactively can be painful, expensive, or even impossible. Basic tracking, events on things people do that matter, good naming conventions. You want a defined data strategy for your implementation, but don’t waste money on expensive consultants at this stage.

Adopt at Stage 1

GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Platform dashboards (Meta Ads Manager, Google Ads)
Basic conversion tracking (purchases, leads, sign-ups)

Stage 2: Post-Purchase Surveys

Annual Media Spend: $10K – $50K

Hopefully at this level you have some real customers from that spend, but you still don’t have enough variability for statistical modelling. Do not buy an MTA or MMM tool. You almost certainly don’t have the conversion volume yet to make the models effective.

The best thing you can add here is post-purchase surveys. You can capture information you otherwise would never get: dark social, word of mouth, podcast discovery, offline touchpoints. They are cheap to run, give you immediate feedback, and right now are the most honest signal you are going to have access to.

If you have enough traffic, start running simple A/B tests on your landing pages and ad creatives.

Last click will still be your go-to here, but treat it as highly directional, not truth. Your survey data will probably contradict it heavily, and when it does, trust your survey data more.

Adopt at Stage 2

Post-purchase surveys
Simple A/B tests

Keep Using

GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Platform dashboards (Meta Ads Manager, Google Ads)
Basic conversion tracking (purchases, leads, sign-ups)

Stage 3: Your First Attribution Layer

Annual Media Spend: $50K – $200K

At this point you likely have enough conversion volume to run some rules-based MTA and get some value out of it, and to start seeing some cross-channel patterns.

It’s also the point where the ROAS your platforms are reporting is unreliable. The platforms will tell you their ROAS is great regardless of reality. At best it’s now a directional signal to compare to your new MTA and your survey data.

Platform lift tests are probably viable for your biggest channels. They are free, but with a big caveat: they are biased toward their own platforms. Run them to directionally understand what’s working, not to validate your overall spend. If you run a lift test on Meta and it says your campaigns are driving 3x incremental revenue, read that with heavy skepticism. These tests are designed to increase your budgets to the ad platforms.

If you haven’t implemented server-side tracking, you should have that done by this stage, to make sure you’re recovering any conversions that slip through because of consent.

Adopt at Stage 3

Rules-based MTA tool
Platform lift tests
Server-side tracking to recover consent-gap conversions

Keep Using

Post-purchase surveys
Simple A/B tests
GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Platform dashboards (Meta Ads Manager, Google Ads)
Basic conversion tracking (purchases, leads, sign-ups)

Stop Using

Platform ROAS metrics as a reliable signal

Stage 4: ML-Driven MTA and Incrementality Tests

Annual Media Spend: $200K – $500K

Your rules-based MTA is already hitting its ceiling. Linear, time-decay, and position-based fixed models can no longer be trusted. Fixed models assign credit via a formula rather than by what the data actually shows. Upgrade to an MTA that uses ML to drive its results.

Also start running your first real Geo Holdout Test, specifically on your top spending channel. At this stage you’re going to begin to get an actual picture of what’s causing revenue versus what’s only correlated with it. The first test will likely show that your true incremental ROAS is much lower than your attributed ROAS. That gap is the misallocation that better measurement will help you recover. That’s normal and good to know, so don’t panic. One geo test is not a full measurement program, but it could be your most important single data point.

Stop using GA4 attribution reports. They are completely unhelpful at this point.

Adopt at Stage 4

ML-driven MTA
A geo holdout experiment on your largest channel
Creative-level attribution (if you have heavy paid social with variants)

Keep Using

Post-purchase surveys
Simple A/B tests
Platform lift tests
Server-side tracking to recover consent-gap conversions
GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Basic conversion tracking (purchases, leads, sign-ups)

Stop Using

Rules-based MTA tool
Platform ROAS metrics as a reliable signal
GA4 attribution reports

Stage 5: Agile MMM (Managed)

Annual Media Spend: $500K – $1M

At this stage you probably have enough variability in your spend, across enough channels, that a Bayesian MMM can start producing a directional signal that provides value. The key point for an Agile MMM is that it requires a heavy human and managed service angle. The outputs can be meaningful, but the confidence intervals can be wide, so the recommendations are still directional, not precise. How well it works depends on your tactics. The more channels you are using, with more spend and more variability, the better the signal the models will produce, versus always-on evergreen campaigns running on two channels.

Agile MMMs work even better when you actually move budgets around regularly, month to month or week to week. Run campaigns in flights rather than having anything always on. Test new channels periodically. Brands with rigid always-on spend across minimal channels will probably not get enough signal from these models.

At this level of spend, stop using any single platform’s attribution numbers entirely. The walled garden self-reporting maze is now going to actively mislead your budget decisions. Your Agile MMM outputs and incrementality tests are going to be more reliable than either Meta or Google’s self-reporting.

Depending on your data, this is also when you want to start hardening your first-party data collection, either with a clean CRM or a basic early CDP.

Adopt at Stage 5

Agile MMM managed service
First-party data hardening: a clean CRM or early CDP

Keep Using

Post-purchase surveys
Simple A/B tests
ML-driven MTA
Geo holdout experiments (quarterly cadence at minimum)
Creative-level attribution (if you have heavy paid social with variants)
Platform lift tests
Server-side tracking to recover consent-gap conversions
GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Basic conversion tracking (purchases, leads, sign-ups)

Stop Using

Platform dashboards (Meta Ads Manager, Google Ads) as a primary signal
Rules-based MTA tool
Platform ROAS metrics as a reliable signal
GA4 attribution reports

Stage 6: Full Triangulation

Annual Media Spend: $1M – $5M

At this point the three-method stack of attribution, incrementality, and measurement is fully operational, with each method watching the others. Your MMM should drive your strategic budget allocation, your geo holdouts help validate and calibrate your MMM, and the ML-driven MTA handles your day-to-day creative and campaign optimization.

Using last-click attribution at this point is actively counterproductive. It will point you in the wrong direction and undermine confidence in better data. Platform native lift tests should also be retired as your primary incrementality source. You should now have much better geo holdout data.

Keep in mind that triangulation only works if the systems are actively talking to each other. You need to use your incrementality tests to calibrate your MMM’s assumptions. If they are independent and nobody is reconciling their outputs, you are going to have three confusing numbers. Having a good human in the loop, whether that’s an internal analyst or an outside partner, is what makes this work.

Other tactics worth exploring at this stage: brand lift studies, monitoring your AEO/GEO, and a first-party data strategy leading you toward a CDP or consolidated data warehouse.

Adopt at Stage 6

Full triangulation (MMM, incrementality, and MTA working together)
Always-on geo holdout program (not just one-off tests)
Brand lift studies
AEO/GEO monitoring
First-party data strategy (CDP or data warehouse consolidation)

Keep Using

Post-purchase surveys
Simple A/B tests
Agile MMM managed service
ML-driven MTA
Creative-level attribution (if you have heavy paid social with variants)
Server-side tracking to recover consent-gap conversions
GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Basic conversion tracking (purchases, leads, sign-ups)

Stop Using

Platform dashboards (Meta Ads Manager, Google Ads) as a primary signal
Rules-based MTA tool
Platform ROAS metrics as a reliable signal
GA4 attribution reports
Platform lift tests as primary incrementality source
Last-click attribution of any kind

Stage 7: Build vs. Buy

Annual Media Spend: $5M – $10M

Once you get to this point, even a 3 to 5% efficiency gain from better measurement is worth hundreds of thousands of dollars annually. The ROI on measurement is now obvious to any CFO. The strategic question shifts: is a managed service still the right model, or is it time to start building your own internal measurement capability?

There’s no wrong answer. Hiring a senior marketing data scientist you can trust is going to cost $150–$200K or more, sometimes 2x to 3x that depending on your market. Hiring internally wins if measurement is core to your competitive differentiation, you have enough data complexity to keep them occupied, or you want to build proprietary models. Managed services win if you want to avoid building your own infrastructure, you have a lean team, and you’d rather spend the extra cost on ad spend. It just depends on what you are building.

At this point there are also a wide variety of additions to consider, depending on your business: creative analytics at scale, a full AEO/GEO program, audience holdout testing on email, an enterprise MMM engagement if you’re doing significant offline media like TV or retail.

Adopt at Stage 7

Creative analytics at scale
Full AEO/GEO program
Audience holdout testing for email and CRM channels

Evaluate Seriously

Hire an internal senior data scientist vs. continuing with managed services
Enterprise MMM engagement for offline channel inclusion (if TV, OOH, or retail media are significant)

Keep Using

Post-purchase surveys
Simple A/B tests
Full triangulation (MMM, incrementality, MTA)
Always-on geo holdout program
Brand lift studies
AEO/GEO monitoring
First-party data strategy (CDP or data warehouse consolidation)
Agile MMM managed service
ML-driven MTA
Creative-level attribution (if you have heavy paid social with variants)
Server-side tracking to recover consent-gap conversions
GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Basic conversion tracking (purchases, leads, sign-ups)

Stop Using

Platform dashboards (Meta Ads Manager, Google Ads) as a primary signal
Rules-based MTA tool
Platform ROAS metrics as a reliable signal
GA4 attribution reports
Platform lift tests as primary incrementality source
Last-click attribution of any kind

Stage 8: The Enterprise Measurement Stack

Annual Media Spend: $10M – $50M+

If you weren’t there before, you likely now have to account for offline channels: TV, OOH, retail media, direct mail, sponsorships. Something there is probably important enough to require inclusion in your models. You should be investing in a true quarterly econometrics MMM from a major analytics consultancy, validated by continuous geo holdouts.

If you have a heavy amount of offline channel data, look to retire any managed service that cannot handle it. Agile digital-only MMM tools are inadequate at this scale for planning, though they can still be useful for monthly tactical digital optimization.

You should also no longer have a single “authoritative” measurement source. At this scale you need multiple independent measurement sources that cross-check each other.

The biggest failure at enterprise scale is organizational, not technical. It doesn’t matter if you have the best data, the best models, the best tech. You can still make bad decisions if your measurement team isn’t talking to your media buyers, if the CFO doesn’t trust your methodology, or if it takes multiple months to act on monthly data. Your measurement infrastructure has to be built in parallel with your data strategy and decision-making infrastructure.

Adopt at Stage 8

Enterprise MMM consultancy engagement
Unified online and offline measurement
Dedicated internal measurement team with a head of marketing science
Custom causal inference models for your highest-spend channels
Full AEO/GEO program with dedicated tooling
Incrementality testing as the standard methodology for every new channel

Keep Using

Post-purchase surveys
Simple A/B tests
Full triangulation (MMM, incrementality, MTA)
Always-on geo holdout program
Brand lift studies
AEO/GEO monitoring
First-party data strategy (CDP or data warehouse consolidation)
Agile MMM managed service (for tactical digital optimization)
ML-driven MTA
Creative analytics at scale
Full AEO/GEO program with dedicated tooling
Audience holdout testing for email and CRM channels
Creative-level attribution (if you have heavy paid social with variants)
Server-side tracking to recover consent-gap conversions
GA4 including proper event tracking
UTM parameters (consistent ones) on every link
Basic conversion tracking (purchases, leads, sign-ups)

Stop Using

Platform dashboards (Meta Ads Manager, Google Ads) as a primary signal
Rules-based MTA tool
Platform ROAS metrics as a reliable signal
GA4 attribution reports
Platform lift tests as primary incrementality source
Last-click attribution of any kind

Additional Notes

Post-purchase surveys never stop adding value

This is one method that basically never stops being useful at any level, because surveys will capture things that no model ever will: word of mouth, Reddit research, AI-assisted discovery. More and more discovery is happening in the dark, and self-reported attribution gets more valuable over time as a result, not less.

Getting off last-click is the hardest political step

Someone, whether it’s a founder or the CFO, has been using this number for years to make decisions, and the numbers feel real and familiar. The biggest challenge after Stage 4 is simply getting completely off last-click. The data will be better. The internal sell is the hard part.

AEO/GEO can be adopted earlier than the roadmap suggests

It’s a brand measurement discipline, not a performance one. It becomes strategically important once your brand has enough volume that being included in AI answers meaningfully affects how people discover you. For most companies under $1M in spend, the best thing you can do is create valuable content and build authority signals through digital PR.