What’s actually broken? Pick the one that sounds like your last quarter. You get the number from the engagement that matched it, and what it costs to start.
Someone sent you? Skip to the receipts.
Answer A
When a platform and a revenue record disagree, every optimization above them is aimed at a figure nobody reconciled. The fix is making the two numbers agree first. The media plan comes after.
Benefits nonprofit serving 100,000+ rideshare and for-hire drivers. Same budget, double the lead volume, once the signal could be trusted.
Roughly what is the monthly ad spend?
Noted. It is in the form below, and you can change it there.
Answer B
A duplicate rate nobody measured. A container republished without its tags. An event firing on a button click instead of a completed enrollment. None of it shows up in the platform report, and all of it is visible within days of admin access.
Every cost per acquisition reported to that board had been divided by a count roughly four times too high. A two-month analytics blackout surfaced in the same pass.
Roughly what is the monthly ad spend?
Noted. It is in the form below, and you can change it there.
Answer C
A budget increase against a signal nobody can read gets spent the way the old budget was spent, on whatever the platform report said was working. Repair the layer first and the same dollars buy more. That is what the figure below is: no new spend, twice the leads.
Benefits nonprofit, 11 weeks. Cost per lead fell 55% and cost per acquisition 34% in the same window, with the budget held flat.
Roughly what is the monthly ad spend?
Noted. It is in the form below, and you can change it there.
Answer D
If the work cannot be evidenced, effort and intent stop being the argument. What the data shows was delivered is the argument, and that is a much shorter conversation.
Billing against deliverables that could not be evidenced. Found within nine business days of admin access, alongside the duplicate rate and the blackout.
Roughly what is the monthly ad spend?
Noted. It is in the form below, and you can change it there.
Meta counts a conversion when its pixel sees one. Your CRM counts an enrollment when someone finishes one. Between those two numbers sit duplicate records, untagged traffic, events that fire twice, and a republished GTM container that quietly stopped sending anything. The gap does not announce itself. It shows up as a media plan optimized toward the wrong creative, a channel defunded because it looked expensive, and a monthly report nobody can reconcile against the bank.
Every decision above the measurement layer inherits whatever that layer already got wrong. So Lagrange builds or repairs the layer first, then the systems that run on it, before anyone spends another month optimizing against a figure that was never true.
The principal ran marketing at director level or above for nine years before this practice, in-house at NBA Entertainment, Freedom Mortgage, NewRez and Prescryptive Health. The largest account on this page serves more than 100,000 drivers.
Still not sure which one you are? Pick the one that sounds like your last quarter.
One engagement, described until permission to name it is on file. The work is named. The figures are published.
Every figure above ships in the page at its real value. No count-up is applied by markup. Two figures from an earlier engagement, $58.83 per booking on $396.70 of spend, were retired from this page in August 2026 because the sample was seven bookings. They are still in the case study, labeled.
Read the case studies →Each row is a system that has shipped for a client. Where a figure is published, the row says which one. Where it is not, the row says what the system does and stops there.
GA4 and GTM rebuilt against a written event spec. Meta CAPI. Server-side tagging where the browser cannot be trusted. Consent mode. A UTM and attribution spec with the reconciliation queries that prove it. Duplicate and blackout detection built in, because both were live defects on the first account we published.
CS-01 · figures publishedA weekly read of the account against a funnel decision tree, cross-checked against last week’s changes, ending in a ranked list of optimizations you approve before anything moves. Strict naming. Creative rotated on evidence. Meta figures are published; Google is run the same way and its numbers are the client’s.
CS-01 · Meta figures publishedA channel is proven on a small structured spend before a media budget is committed to it, and the result is read only through the repaired measurement layer, because a test read through a broken one proves nothing. Testing roadmaps are sequenced in-account.
Pattern published in CS-02 · figures retired from this pageAn ad-generation system that holds a brand’s guidelines while folding in what is currently working, built twice, for a social platform and for a benefits nonprofit. A creative review process from zero: structuring rules written down, review GPTs, and a team that can run it without us.
Two deployments · figures pending a named numberDashboards with the rule attached: what fires when the number moves, and who moves it. CEO reporting moved from anecdote to a structure the board can question. A dashboard with that rule attached is an operating change. Without it, it is a report.
CS-03 · confidential, described onlyTechnical audits, schema markup, monitoring for search and for answer engines, content calendars built from what the queries actually say. Run as a secondary workstream for a client and as the growth channel for a product the studio owns.
Client secondary · owned ventureCustom GPTs on a team’s own process. Custom Claude skills and agent systems built for a client’s workflow. The team trained on the parts they own, so the system does not leave when we do.
CS-01 and CS-03 · 2 GPTs, 11 rules codifiedAutomated regulatory monitoring, built for the California AI disclosure law, so a change in the rules reaches you the day it is published.
CS-02The plumbing between the ad account, the CRM and the landing page, built so the record that lands in the CRM is the one the pixel saw. Landing pages and funnels with the measurement designed in from the first wireframe.
Capability level · receipts to be namedStart with the health check if you want the diagnosis fast. Start with the full audit if you want the diagnosis and the written spec to fix it. Start with the retainer if you already know what is broken and need it held steady. Every audit fee credits toward any build that follows.
Everything that touches a number gets checked: tags, events, attribution, and what the reported figures are actually counting. You get the findings and a ranked list of what to fix first, with the cost of leaving each item alone.
The health check, then the document that fixes it: a written UTM and attribution specification your team and your vendors build against, and the cost of inaction quantified line by line so the remediation list can be argued in dollars.
Tracking rots without announcing that it has. This is the standing check that a broken tag, a republished container or a renamed campaign does not cost you a month of data before anyone notices. Each month ends with one page: what we checked, what we caught, what it would have cost.
Three months minimum because the first month is the baseline, and a check with no baseline catches nothing. After that you cancel any month.
Larger builds, including the AI systems in rows 03 and 07 above, are quoted to scope after the audit, and the audit fee comes off the quote.
The form asks the three questions we would have asked over email: what you measure now, what is in the stack, and the spend band. The reply answers them, or asks the one thing it still needs.
One page: what gets checked, which tier fits, what you receive, what it costs. If the audit is not the right thing, that page says so and names what it is. The reply includes a booking link if you would rather walk it through on a 20-minute call; the call is optional and the note stands on its own.
Admin on GA4 and GTM, read on the ad accounts. Asked for after you have read the scope note, because before that it is a demand and after it is logistics.
Findings, the remediation list in priority order, and the walkthrough with whoever owns the fixes. The full audit adds the written spec and the cost of inaction.
Client names and logos appear only with written permission on file. Anonymity reads as discretion, and it means work can be published the moment there is a result. Two of the three studies were delivered inside a partner engagement; the partner is named on the card and the measurement work is ours.
“You hire us for one thing. You get better at seven things.”
Lagrange stands up products of its own and runs them the way it tells clients to: measurement wired before distribution, a written event spec before the first ad, a price on the page. They keep their own identities here, because restyling a portfolio brand into the studio’s colors would make this a page of themes instead of a page of assets.
Financial infrastructure for creators. Live MVP at lootcreator.com. GA4 and Search Console wired on day one; the event map is the first document in the repo.
Live MVP · pre-tractionNew Jersey landlord legal forms without a lawyer. 11 document types, 7 tools. Grown on organic search, run as the studio’s own SEO and AEO test bed.
Live at njlandlordforms.comAI-guided book recommendations from a short interview: goals, time, taste, then a personalized list. The same pattern Lagrange builds for clients, running as a consumer product.
Live at whatshouldireadnext.aiThe health check finds what is broken and ranks it. The full audit adds the written UTM and attribution specification your team builds against, and puts a dollar figure on each item left unfixed. If you have an engineer who will do the fixes and a vendor who needs a spec, take the full audit. If you want the diagnosis first and the spec later, take the health check; the fee credits toward the audit.
Your agency runs the media. This is the layer the media reports against. Most agencies inherit whatever measurement was there when they arrived, and their reports inherit its errors. The audit is written so your agency can act on it; it is not a pitch to replace them.
Then the report says so, in writing, and you have a document that proves it the next time someone asks. If the health check finds nothing worth fixing, the fee credits toward the retainer.
Admin on GA4 and GTM, read access on the ad accounts, and a look at the CRM’s conversion record. It is asked for at step three, after you have read the scope note. Nothing is changed in any account during the audit. You receive the findings; your team or your vendor applies them.
Yes, on Meta and on Google, as the paid media operating system in capability 02, and only against a measurement layer that has been repaired first. Meta figures are published on this page. Google accounts are run the same way and their numbers belong to the client.
No. Conversational acquisition is built by our partner Namakina. Two case studies on this page came out of that partnership. If chat is the right channel for you, we refer you and stay on the measurement side.
Because the first month is spent establishing the baseline, and a check with no baseline catches nothing. Months two and three are when it starts catching things. After three months it runs month to month and you cancel any month.
Names and logos appear only with written permission on file. Until then the client is described, the work is named and the figures are published, which is the part that can be checked.
It is quoted to scope after the audit, because the audit is what makes the scope honest. Whatever the audit cost comes off the quote.
A media plan built on a number nobody can reconcile is a guess with a budget attached.
“I open with an audit rather than a media plan, because everything downstream inherits whatever the measurement layer already got wrong.”
The practice is small on purpose. Where an engagement needs a specialist, you get their name before they start.
You hire us for one thing. You get better at seven things.
Three questions, one business day to a reply, a scope note you can decline.
Received. A reply lands within one business day at your address. The scope note follows within two. Nothing is charged until you say yes to it.
Read the case studies while you wait →Prefer email? hello@lagrangestudios.co reaches the same inbox.
La Grange Studios, LLC · New Jersey