Before Handing an Agency a $200k/Month Ad Budget: What to Verify About Their Tech Stack

Key Takeaways

• At a large budget, the tech stack matters more than the creative reel — because a broken measurement setup wastes the entire spend efficiently while the dashboard glows green.
• Verify measurement first: server-side conversion tracking (CAPI/sGTM), first-party collection, and values on events — client-side-only tracking at this budget is a standing, expensive leak.
• Insist on ownership: you own the ad accounts, tracking config, models and dashboards under your credentials and keep them if you leave — an agency that owns your stack owns you.
• Verify signal protection against cookie loss, consent and iOS restrictions, because at scale every point of lost signal degrades every optimisation decision the budget funds.
• Demand incrementality, not last-click: how will they prove the growth was caused (holdouts, geo tests, MMM), not just claim credit via platform ROAS?
• Check reconciliation: their numbers must tie to your booked revenue and your CFO's ledger, not only to platform dashboards, or you are scaling on a distorted picture.

At This Budget, the Stack Matters More Than the Reel

When you are about to hand an agency a large ad budget — call it $200k a month, though the logic holds anywhere near that scale — the instinct is to evaluate them on the things that are easy to evaluate: the creative reel, the client logos, the case studies, the chemistry of the pitch. Those are not irrelevant, but they are the wrong first priority, and prioritising them is how sophisticated companies end up wasting enormous budgets with impressive-looking agencies. The thing that most determines whether a large budget produces profit or is quietly incinerated is the agency's tech stack — how it measures, who owns the data, how it protects the signal that drives every optimisation, and whether its numbers reflect reality — because at this scale, a broken measurement setup does not merely underperform. It wastes the entire budget efficiently, optimising hard toward a distorted picture, while the dashboard glows green and everyone believes it is working.

This is the crucial and counter-intuitive point: at a large budget, the failure mode is not obvious underperformance that you can see and fix; it is confident optimisation against broken measurement, which looks like success right up until you reconcile the numbers to your actual profit and discover the gap. A great creative team pointing a large budget at a broken measurement stack will produce beautiful ads, a healthy-looking ROAS, and a business that is not making the money the dashboard implies — because the stack that was supposed to tell them what is working was lying to them, and they optimised toward the lie with real money and real diligence. The bigger the budget, the more this matters, because the bigger the budget, the more money flows through the broken measurement before anyone notices.

So the diligence that protects a large budget is stack diligence, and it is the diligence most companies skip because it is technical, unglamorous, and not what the pitch is designed to showcase. The agency will want to talk about creative, strategy and results; you need to make them talk about measurement, ownership, signal and reconciliation, and you need to verify their answers rather than accept them. The rest of this guide is exactly what to verify, in five areas, with the questions to ask and the red flags that should stop the deal — because it is far cheaper to discover a broken stack in diligence than to discover it six months and a million dollars into an engagement.

Area One: Measurement and Tracking

The first and most important area is how the agency measures conversions, because measurement is the foundation on which every optimisation decision the budget funds is built. In a world of cookie loss, browser tracking prevention and iOS privacy restrictions, client-side (browser) tracking alone misses a large and growing share of conversions — which means an agency relying on client-side pixels is optimising a large budget against incomplete, distorted data, and does not fully know which spend is working. The standard you should verify is server-side measurement: a conversions API (CAPI) and server-side tag management that collect conversion data server-to-server, which is more complete, more resilient to privacy changes, and harder to lose. Ask the agency directly whether they implement server-side tracking, ask them to describe their setup in technical detail, and — this is the part most companies skip — ask to see it, or to have your own technical person verify it, because the difference between an agency that says it does server-side tracking and one that actually does it well is enormous and invisible from the pitch.

Go deeper than 'do you do server-side'. Ask whether they collect first-party data properly, whether their conversion events carry values (so the platforms can optimise toward high-value conversions, not just conversion counts), whether events are deduplicated correctly across browser and server, and how they handle the specific signal-loss challenges of your platforms and your market. An agency that answers these fluently and specifically has a real measurement practice; one that gives vague reassurances or treats measurement as something the platforms handle automatically is one whose measurement you cannot trust with a large budget. The tell is depth: a competent measurement practice can go arbitrarily deep on the technical detail because it lives there every day, while a superficial one runs out of substance quickly and retreats to reassurance.

The reason this matters so much at a large budget is compounding: every optimisation decision — which audiences to scale, which creative to fund, which products to push, where to move budget — is made on the basis of the measurement, so if the measurement is degraded by a third or more (which client-side-only tracking easily loses to modern privacy restrictions), then every one of those decisions is made on a distorted picture, and at $200k a month those distorted decisions move enormous sums in the wrong direction. Verifying the measurement stack is not a technical nicety; it is the single highest-leverage piece of diligence you can do before handing over a large budget, because it determines whether the agency can see clearly enough to spend your money well. This is the foundation of any credible performance marketing engagement, and its absence is disqualifying at scale.

Area Two: Data and Account Ownership

The second area is ownership, and it is the one companies most often get wrong because it is not urgent until it is catastrophic. Verify, before you sign, that you will own everything the agency builds and touches: the ad accounts (under your business manager and your credentials, not the agency's), the tracking and measurement configuration, any models and dashboards, and the data itself. This matters for two reasons. The first is leverage and continuity: if the agency owns your ad accounts, your tracking setup and your data, then leaving the agency means losing your accounts, your measurement history, your optimisation learnings and your data — which makes you captive, because the cost of switching is not just finding a new agency but rebuilding your entire growth infrastructure from scratch. An agency that owns your stack owns you, and it knows it.

The second reason is that ownership is a proxy for the agency's whole posture toward you. An agency that insists on building everything in its own accounts and tools, that resists giving you full access, that treats the measurement setup as its proprietary asset rather than yours, is an agency optimising for your dependency rather than your capability — and that posture will show up in a hundred other ways over the engagement. An agency that builds everything under your credentials, gives you full access, and treats the infrastructure as yours to keep is one whose incentives are aligned with your long-term capability, and that posture too will show up everywhere. So the ownership question is both a concrete protection (you keep your infrastructure if you leave) and a diagnostic (it reveals whether the agency is building your capability or its own indispensability).

The specific things to verify and get in writing: the ad accounts are owned by your business entity, with the agency granted access rather than ownership; the server-side tracking and measurement infrastructure is built under your domain and your accounts; any dashboards, models and playbooks are yours and documented; and on termination, you retain everything and the agency's access is simply revoked, with no loss of your accounts, data or measurement. An agency that agrees to all of this readily is one you can trust with a large budget without becoming captive; an agency that resists any of it is telling you that it is planning for your dependency, which is exactly the wrong foundation for handing over $200k a month. Verify ownership before the budget flows, because it is nearly impossible to claw back afterward.

Area Three: Signal Protection and Privacy

The third area is how the agency protects signal against the ongoing erosion of tracking — cookie deprecation, browser tracking prevention, consent requirements, and platform privacy restrictions — because at a large budget, every point of lost signal degrades every optimisation decision, and signal loss is not static but worsening. Ask the agency how it handles these challenges concretely: how it maintains conversion signal despite cookie loss (which is where server-side tracking and first-party data come in), how it handles consent and privacy compliance in your markets (which is both a legal requirement and a signal question, because non-consented data cannot and should not be used), and how it copes with platform-specific restrictions like iOS's privacy changes that have degraded mobile app and web measurement.

This area overlaps with measurement but deserves separate attention because it is forward-looking: an agency's stack might measure adequately today and be about to fall off a cliff as the next round of privacy changes lands, and you are handing over a budget for the future, not just the present. So verify not just that the agency's measurement works now but that it is built for the direction of travel — server-side by default, first-party data as the foundation, consent handled properly, and an active practice of adapting to platform changes rather than being surprised by them. An agency that treats signal protection as an ongoing discipline it invests in is one whose measurement will still be good in a year; one that treats it as a solved problem or the platforms' responsibility is one whose measurement is quietly decaying and will degrade the returns on your budget as it does.

There is also a compliance dimension that is genuinely important at scale and that a good agency takes seriously: handling personal data and consent correctly is a legal obligation, not an optional refinement, and an agency that is cavalier about consent or that proposes tracking approaches that skirt privacy rules is exposing you to real regulatory risk on top of the signal risk. A competent modern agency treats privacy-compliant, consent-respecting, first-party, server-side measurement as the standard — because it is both the compliant approach and, not coincidentally, the most durable and complete one. Verify that the agency's signal-protection approach is compliant as well as effective, because at a large budget you are exposing a large surface area, and an agency that protects signal by cutting corners on privacy is trading a regulatory problem for a measurement one.

Area Four: Attribution and Incrementality

The fourth area is how the agency proves that its work is actually causing growth, rather than taking credit for growth that would have happened anyway — and at a large budget, this is the difference between a genuine return and an expensive illusion. Most agencies report platform ROAS and last-click attribution, which systematically over-credit the agency, because the platforms grade their own homework and last-click assigns all the credit to the final touch. An agency spending a large budget can post a healthy platform-reported ROAS while contributing far less incremental value than the number implies, because much of what it is 'driving' is demand that already existed and would have converted anyway. The only way to know the difference is incrementality measurement, and you should verify that the agency does it.

Ask concretely how the agency measures incrementality: does it run holdout tests (withholding spend from a representative control group and measuring the difference), geo experiments (varying spend across matched geographies), or media-mix modelling (statistically estimating each channel's incremental contribution)? An agency that has a real incrementality practice can describe its methods, has run these tests, and thinks natively in terms of caused versus correlated effects; an agency that has never run an incrementality test and cannot explain how it would is optimising your large budget toward a number that overstates its own contribution, and neither you nor it actually knows how much of the growth is real. At a small budget this is a tolerable imprecision; at $200k a month it is a large sum of money being allocated on a systematically flattering metric.

This also connects to how the agency will be held accountable, which is why it matters for the whole relationship, not just the reporting. An agency that measures incrementality has chosen to be held to caused results, which means it has to actually deliver them; an agency that reports only platform ROAS has chosen a metric that flatters it regardless of its true contribution, which is exactly what an agency that was not confident in its incremental value would do. So the incrementality question is both a measurement check (can they tell what is real?) and an accountability check (are they willing to be measured on what is real?). Before handing over a large budget, verify that the agency will prove incrementality rather than claim credit, because the alternative is paying a large sum for a number that cannot distinguish your agency's contribution from your own baseline demand.

Area Five: Reporting and Reconciliation

The fifth area ties the others together: do the agency's numbers reconcile to your actual business results — the revenue you booked, the contribution your finance function recognises — or do they live only in platform dashboards that never touch your ledger? This is the check that catches the gap between a great-looking dashboard and a disappointing bank balance, and it is the check that most companies never make until the gap becomes undeniable. An agency reporting a healthy platform ROAS on a large budget may be reporting a number that, when you reconcile the platform-claimed conversions against what your finance function actually booked, does not add up — because the platforms each claim credit for overlapping conversions, modelled conversions inflate the totals, and none of it has been tied back to your real revenue. At a large budget, an unreconciled reporting practice can hide a very large discrepancy between claimed and actual results.

Verify, therefore, that the agency reconciles its reporting to your booked revenue and can produce numbers that tie to your finance function's ledger, not just to the platform dashboards. Ask how they handle the fact that platform-reported conversions across channels sum to more than your actual sales (because of cross-channel double-counting), how they treat modelled conversions, and how they would present results to your CFO in terms that reconcile to what finance sees. An agency that can do this — that reports in reconciled, finance-grade numbers — is one whose reporting you can trust with a large budget; an agency that reports only platform ROAS and has never reconciled to a client's ledger is one whose numbers may be systematically overstating results, and you will not know by how much until you do the reconciliation yourself, by which point a lot of budget has flowed against a distorted picture.

Pulling the five areas together, the diligence before handing over a large budget is fundamentally about whether the agency can see clearly and honestly — whether its measurement is complete and owned, its signal protected, its incrementality proven, and its reporting reconciled to reality. An agency that passes all five can be trusted to spend a large budget well, because it is optimising against an accurate picture and being held to real results; an agency that fails any of them is one whose large-budget spend is being optimised against a distorted picture, however impressive its creative and its logos. It is far cheaper to run this diligence before the budget flows than to discover a broken stack after it — so make the agency prove its stack, verify the answers rather than accept them, and let a failure in any of the five areas stop the deal, because a large budget on a broken stack is not a risk, it is a certainty of expensive waste.

How to Actually Verify, Not Just Ask

A crucial and often-skipped point is that asking the agency the five sets of questions is not the same as verifying its stack, because a competent-sounding answer is exactly what an agency good at pitching produces whether or not the substance is there. The whole value of stack diligence collapses if you accept the agency's word, because the failure mode you are guarding against — confident optimisation against broken measurement — is invisible precisely to the people running it, who believe their stack works, so the agency answering 'yes, we do server-side tracking and reconcile to revenue' may sincerely believe it while the implementation is broken. Verification means going beyond the answer to the evidence: asking to see the actual setup, having your own technical person or an independent expert inspect it, and requiring the agency to demonstrate rather than assert. An agency confident in its stack will welcome this; an agency that resists letting anyone look closely is telling you something about what a close look would find.

Concretely, verification looks like this. For measurement, ask to see the server-side tracking configuration and have someone technical confirm it is actually collecting and deduplicating events, not just present in name. For ownership, read the contract terms on account and data ownership and confirm the accounts are actually under your business manager, not the agency's, before any budget flows. For signal protection, ask the agency to walk through, in technical specifics, how it handles a concrete scenario like consented versus non-consented users or a specific platform's restrictions, because specifics are hard to fake and reassurance is easy. For incrementality, ask to see an actual example of a holdout or geo test the agency has run — its design, its results, what it changed — because an agency with a real incrementality practice has artefacts to show, while one that has never run a test has only the vocabulary. And for reconciliation, ask the agency to reconcile a sample of its reported numbers against your own booked revenue during the diligence itself, because an agency that reconciles as a matter of course can do it on the spot, and one that cannot will reveal the gap in the attempt.

The principle is that the pitch is the agency at its most controlled and the evidence is the agency at its least controlled, so verification shifts your decision from the controlled evidence to the uncontrolled evidence, which is where the truth is. This costs you time and possibly the fee of an independent technical reviewer, and at a $200k-a-month budget that cost is trivial against the money at stake — a single month of budget optimised against a broken stack dwarfs the cost of verifying the stack first. So do not let the seniority of the pitch, the pressure to move quickly, or the awkwardness of asking to inspect someone's work talk you out of verification, because the entire point of stack diligence is to catch the broken stack before the budget flows, and you cannot catch it by accepting the answers of the people who cannot see that it is broken. Verify, then decide.

Methodology & Fairness

A note on how to read this. This is an opinionated guide published by Fluxsy, not an independent ranking or audit. Where we name other companies, agencies or tools we describe them only by their genuine, public positioning; nothing is an endorsement or a paid placement, and any may be right for one company and wrong for another. We have deliberately avoided inventing statistics, results or 'best provider' claims. The durable value is the evaluation framework, which holds no matter which partner you ultimately choose — including if that partner is not us. Verify every specific claim, ours included, against primary sources and your own diligence.