OllieLabs Closed beta Opening to a small number of advertisers at a time. Access is by request. Request access

Software that runs your paid media on its own.

Not a dashboard and not a report. It joins every door your leads arrive through into one funnel, takes the stages that actually closed, and moves the money against them — inside limits you set, with nobody logged in.

Built by one operator: a chemical engineer, nine years running paid acquisition, US$100M+ in media managed.

Request beta access

Budget reallocating itself until it matches the revenue that actually closed Two stacked bars for the same advertising account. The upper bar is what the account spends, split across four campaigns; the lower bar is the revenue that actually closed from each of them. The two do not match: the campaign taking forty two per cent of the budget produced six per cent of the revenue. The upper bar then moves on its own until its shape matches the lower one, and repeats. Nobody logs in to do it. Your budget, moving itself reallocated against closed revenue, unattended EVERY 6 HOURS ILLUSTRATIVE ACCOUNT WHAT IT SPENDS · R$ 12,000/day WHAT ACTUALLY CLOSED · R$ 214,900 Retarget 7d Lookalike 2% Search · brand Search · non‑brand The top bar keeps walking into the bottom one. Nobody logs in to make that happen.
Four campaigns in one account: the one taking 42% of the budget closed 6% of the revenue.

The account runs itself. On a number that is actually true.

Leads arrive through every door you have. We join them into one funnel, and you tick which stages of it go back — the whole funnel is available, not just the sale. Those feed the thing that decides what to buy next. Returning the stages is the input. Operating the media is the product.

The whole loop: every door, one funnel, the stages you choose sent back, and software that runs the media against them Leads arrive through eight unrelated doors at once — WhatsApp, a site form, a Meta lead ad, a phone call, a Google form, a LinkedIn form, someone walking in, a spreadsheet — drawn as a tangle of crossing wires. OllieLabs captures them and resolves them to one person, which fills an ordered sales funnel: four thousand one hundred and twenty prospects narrowing to seventy four purchases. The advertiser ticks which stages go back — here Qualified and Purchase — and those feed the operator, which runs the media on its own: it moves budget between campaigns, sets bids, targets and audiences, turns creative on and off, and returns the chosen stages. Those decisions land on Google Ads, Meta, TikTok, LinkedIn, Pinterest and analytics, where the ads run and fill the doors again. EVERY DOOR YOU ACTUALLY HAVE ONE ORDERED FUNNEL WhatsApp Site form Lead ad Phone Google form LinkedIn Walk‑in Spreadsheet Ollie captures and joins every door, resolved to one person Prospect 4,120 Contact 1,840 Scheduled 610 Qualified 212 Purchase 74 · R$112k the whole funnel is available — you tick what goes back the stages you chose moves budget, bids and creative The agent runs the media on its own WHERE OLLIE RUNS THEM the ads run again Without the bottom half you have a report. With it, the account moves on its own. Each door writes somewhere else, and writes something different. A lead ad returns an ad id but no creative, a WhatsApp thread returns a click id, a phone call returns nothing at all. Making them one funnel is the work, not a setting. The same person shows up as a phone number here, an email there and a click id somewhere else. We keep every touch and resolve them to one identity, so the journey survives a CRM that overwrites the origin on each visit. Yours, not ours. You declare the stages your business actually has and every contact is placed on them. The volumes here are one account’s month; what matters is that each stage is a real, countable thing rather than a label on a chart. The whole funnel is available, and you tick what gets returned. Sending only purchases starves a young account of signal; sending everything teaches the auction to buy cheap leads that never close. Most accounts land on one mid-funnel stage plus the sale — and you can change it without touching a tag. It reads the stages you chose and acts in the account, on its own: moves budget between campaigns, sets bids, targets and audiences, turns creative on and off, and returns those stages to the platforms. Every move argues from revenue rather than from what the platform counted for itself. It stays inside limits you declare, logs every change with the number behind it, and reverts in one click — or you keep it in shadow mode, where it proposes and waits. The fold below takes this apart step by step. The same places the decisions have to land. Google Ads and Meta are wired and proved end to end today; TikTok, LinkedIn, Pinterest and analytics destinations are the same call in a different envelope.
Sending data back is table stakes now — three categories of tool already do it. What almost nobody does is decide the next spend against that number: the budget, the bids, the creative, without a person in the loop. That is the half we build; the rest is the input it needs to be worth anything.

Request beta accessSee the price list

Four things have to be true. Only the fourth one is the product.

The first three are plumbing, and every tool in this market sells one or another of them. The fourth is the one that changes what your money buys — and it is worth nothing if the first three are not true.

One buyer crossing four channels over twenty one days before a single sale A single buyer, drawn as a person on the left, moves along a timeline through four channels: a Meta ad on day one, a WhatsApp conversation on day two, a phone call on day nine, and a visit to the store on day twenty one, ending in a sale of one thousand five hundred reais. Four separate systems each recorded one of those moments and none of them recorded the other three. ONE BUYER · FOUR CHANNELS · TWENTY‑ONE DAYS Meta adday 1 WhatsAppday 2 Phone callday 9 In storeday 21 R$1,500 Four systems each saw one of these moments. None of them saw the other three, and the CRM overwrote where she came from on every visit. We keep all four and put them back together.

Request beta accessSee the price list

Who else is in this, and what they actually do.

There is no vacuum here. Three groups already sell parts of this and they are good at the part they sell. The four rows below are the same four from the fold above.

  The pipe Measurement suites Auto‑operators OllieLabs
Join every door to one persona lead ad, a chat, a call, a walk‑in on the sitearound a checkoutnoany door
Let you choose which stage goes backnot just the purchase nothe salenoany stage
Return it to the ad platformswith the real value attached yesyesnoyes
Operate the account against itbudget, bids, creative nonoon platform numbersyes

The pipe is Tracklution, TagServer and server‑side GTM — and it is our own entry plan, not our product. The measurement suites are Northbeam, Triple Whale and Hyros: the returning half stopped being ours alone, but all of them are built around a checkout, so a sale that closes in a chat or across a counter was never written down. The auto‑operators do move budget and creative — on what the platform reports about itself.

So the order we build in is deliberate.

  1. Entry: the connection. Table stakes, priced like table stakes.
  2. Core: the account operated on its own, against a number you can defend. The measurement is what makes the operation trustworthy; the operation is what you pay for.
  3. After: journeys, messaging, creative. Worth building, and worthless before the two above are true.

Request beta accessSee the price list

An agency has to profit on a portfolio.

That one constraint caps what you are allowed to buy. There are exactly three shapes, and you have met all of them.

A dedicated pod

You pay
Full cost of the pod, plus margin
You get
Real attention
The catch

You are paying an outside company to manage a team that is effectively yours.

One pod, N accounts

You pay
A fee diluted across a portfolio
You get
A price you can afford
The catch

Split across media, design and supervision, nobody has the hours left to learn your business.

Your own team, in house

You pay
Salaries, tools and the hiring
You get
People who work only for you
The catch

They are judged on sales while optimising toward leads, because closed revenue never reaches the platform.

400h 200h 80h 13h 1 6 10 20 30 Software: divides nothing The pod's hours, divided Team hours each account gets, per month Accounts handled by one pod
Drag the handle. A pod is two and a half full‑time people: one media manager, one designer, and a supervisor shared with other accounts. About 400 hours a month, and every account you add divides the same 400 again. The dashed line never moves, because software is not splitting anyone's hours.

The constraint

Three shapes, one limit: headcount. In none of them does anyone have the hours left to learn your business.

The software runs as a copilot beside whoever does the work — your agency, your team, or us.

Request beta accessSee the price list

In software, one more account costs about nothing.

Which does not soften the trade‑off, it deletes it: the attention of a dedicated team at the price of a portfolio seat.

Then why hasn't software already won?

Everything built before sold tools for a human to operate: bid managers, dashboards, audit suites. The human stayed the bottleneck, so the headcount math held. Three things changed that, at once.

The signal broke

App tracking prompts and the end of third‑party cookies cut what the platforms can see. They optimise toward a form view because a form view is all that still reaches them.

The pipe back opened

Sending real revenue back stopped being a partnership and became an API call. Google moved offline conversion import out of the Ads API on 15 June 2026, keeping only an allowlist of integrations that were already sending; we were on the wrong side of it and ship on the Data Manager API instead. Meta takes it through the Conversions API. Anyone can write to both.

The judgment moved

Smart Bidding and Advantage+ genuinely work now, which moves the bottleneck off the operator and onto the signal you feed them. That is the part nobody fixed, and the part a model can now hold.

Request beta accessSee the price list

You decide how much autonomy to hand over.

Nothing touches your account without permission. Most advertisers start at Data, see the gap, then move up.

$20k – $50k
TIER 1

DataWe connect your ad accounts and your sales data, join the identities, report the funnel, and send the real conversions back to the platforms as offline events. What it does not do is change your campaigns: no recommendations, no edits.

Attribution, funnel reporting, and the real conversions sent back as offline events.

$0/month

TIER 2 Recommended

AssistantEverything in Data, plus a daily list of changes we would make, each with the estimated impact in revenue. You approve them one by one. We never touch the account on our own.

Daily recommendations with estimated impact, applied in one click.

$0/month

TIER 3

AutomaticWe apply changes directly, inside limits you set: maximum budget shift per day, campaigns that are off-limits, and a full change log. Any change reverts in one click.

The software operates, with guardrails, a full change log and one-click revert.

$0/month

Every plan at this band

  • History12 months, refreshed daily
  • Always unlimitedAd accounts, campaigns and seats

Guided setup, $390 — optional, at any band. Every plan connects self‑serve and nothing is locked behind it. Buy it when you want a person doing the wiring and checking the join against sales you already know, before anyone reads a number off it. While the beta is closed, every band comes in through the same form as everyone else.

One metric, in coarse steps — and nothing here needs a call to explain it.

What you pay for
A flat monthly price, per account
Stepped by how much media you run. Coarse bands, so growing your budget by 10% does not grow your bill.
A guided setup, only if you want one
Optional at every band, because a product that cannot connect itself is a bug, not a fee. You are buying an afternoon of ours, not a key to the door.
Nothing else
No overage, no fee per campaign, per channel or per asset created. Adding WhatsApp or a fourth platform does not change the invoice.
What we will never charge
A share of your ad spend
The day it tells you to cut budget it would be cutting its own revenue. That is how agencies are paid.
Per seat
The gap we close is knowledge. Per seat would tax the exact thing we want: your whole team on the same numbers.
A success fee
We would be grading our own homework. The company measuring the revenue cannot also be paid on it.

That is the whole price list. No call to unlock it, no quote built around what we think you can pay. Accounts that join the beta keep the band they join on for twelve months.

Request beta accessSee the price list

When we are the wrong answer.

We would rather lose the call than lose your quarter.

Nobody can say what closed

No CRM, no spreadsheet, nobody marking the deal anywhere. There is no truth to join, and we would be guessing with better graphics.

You run very little media

Below a few hundred conversions a month there is not enough signal to model, and you can still hold the whole account in your head.

You want the rest of an agency

Brand strategy, PR, production beyond ads. We run acquisition and we do not pretend to be the other thing.

You want someone to blame

We hand you the truth and the controls, which is more accountability, not less. Some teams would rather an agency absorbed it.

Two things that look like exclusions and are not.

Selling on your own site. That removes the identity problem, not the ads problem. The platform still optimises toward whatever event you feed it, and a checkout is not the same as a profitable customer.

Having an in-house team. Your team still needs closed revenue going back into the auction. Without it they are optimising toward leads while being judged on sales, which is the same trap with a friendlier org chart.

Request beta accessSee the price list

Who is behind it.

Tayro Toledo Chemical engineer (USP) · Founder

An engineer who ended up in marketing, and spent nine years running paid acquisition across education, fintech, e‑commerce and healthcare. The same play every time: come in, drop the agency, rebuild the tracking, hit the target.

The record

  • −70%Acquisition cost, by bringing media in-house and removing the agency
  • −85%CAC at a brokerage, from US$545 to US$45
  • 18k → 90kStudents at a university in two and a half years, team from 9 to 35+
  • US$100M+In digital media managed across nine years

It worked every time, and it never scaled, because it needed me in the room. Writing it as software is the part the engineering degree was for.

What is already running

  • 7Advertisers in the closed beta right now, on live ad accounts
  • Google + MetaBoth wired end to end. Returning a closed sale is proved against live accounts, not against a mock
  • 10 monthsOf real WhatsApp conversations measured still carrying the id of the ad that started them, in a live account
  • 295Automated tests behind the join, the return ledger and the reprocessing, all passing

Seven accounts in, and none of them is paying yet — the price list is what it costs when the beta closes. The record above is what I did for other people’s companies; this is what exists of my own, and that is the honest state of it.

Request beta accessConnect on LinkedIn

Request access to the closed beta.

Seven advertisers are in. We open to a few more at a time, because each account brings one more combination of CRM, channel and platform — we prove it once against real sales, and from then on it is code, not a booking in someone’s calendar.

  1. You send this. Two minutes, and the last field is the one that matters.
  2. I read it myself and check whether the join is even possible on your setup. Usually inside two working days.
  3. If it fits this round, we wire one account and prove it against sales you already know, before you are asked for anything.

If the fit is wrong for this round, I say so and you stay on the list. No sequence, no drip.

Please add your name.
That address doesn't look right.
Include https:// at the start.
Pick a range.
Optional, and it is the answer that tells us whether we can help.