Advertising platforms
Google and Meta operate their own auctions, delivery systems and automated campaign features. Lonta configures and manages the client’s accounts; the platforms control their underlying systems.
Back
Lonta AI technology
From campaign data to a decision you can inspect.
Lonta AI combines advertising and CRM data, machine learning and specialist review to run marketing for home service companies. Software helps find patterns and prepare work. A named account lead owns the decisions, execution and explanation.
What a change record contains
- Source
- The account, call or CRM record
- Reason
- The issue and the proposed change
- Owner
- The person responsible for the work
- Follow-up
- The outcome and review window
How data becomes reviewed marketing work
The service workflow connects analysis to execution and back to business outcomes. The client’s tracking setup determines which records can be connected.
Connect the records
Advertising, calls and CRM outcomes provide different parts of the picture. The available connections depend on the client’s accounts, access and tracking setup.
Check what the data means
A call, a booked visit and a sold job are different events. Missing outcomes, duplicate inquiries and changes in tracking need to stay visible.
Analyze and estimate
Software processes repetitive account work. Internal models help surface patterns and predictions; those estimates need a defined question and an explanation of their limits.
Review with a specialist
The account lead checks the recommendation against service area, answered hours, job economics and technician capacity. A model cannot supply those business decisions by itself.
Execute within the plan
The team implements approved campaign, creative, page and tracking work inside agreed budget guardrails. Changes have a reason, a timestamp and an owner.
Follow the outcome
Calls and CRM outcomes remain connected to the marketing record. Weekly briefs and monthly reviews explain what changed and what happened afterward.
Three parts of the system, with different responsibilities
Lonta’s analysis and models
Our internal predictive models and account analysis support the team’s decisions. They work alongside the platforms and the available client records. An estimate needs to be checked against the outcome it claims to predict.
People responsible for the work
Account managers keep the client context clear. Marketing specialists execute campaigns. Product and ML expertise help interpret the data. A named lead explains the recommendation and owns the follow-through.
Google Ads and Google are trademarks of Google LLC. Meta, Facebook and Instagram are trademarks of Meta Platforms, Inc. Microsoft, Microsoft Advertising and Bing are trademarks of the Microsoft group of companies. Lonta AI is not affiliated with, endorsed by, sponsored by, authorized by, or otherwise approved by any of these companies.
Why the CRM outcome matters
A call tells you that someone made contact. A booked visit tells you the dispatcher found a slot. Sold work and collected revenue answer later business questions. Mixing those events can make an account look efficient while the shop stays short of profitable work.
Lonta connects calls and CRM outcomes to the marketing record so the review can follow that difference. A record that cannot be matched needs to remain unresolved. Available history and tracking quality affect how complete the picture can be.
Google’s offline conversion documentation describes importing outcomes that happen after an ad interaction. That capability does not establish that an ad caused a sale. Read our guide to branded searches and new-customer jobs for that distinction.
How to judge a predictive model
Start with the business question and the moment the estimate is made. Predicting a booked visit from a new inquiry is different from reviewing a completed call or forecasting seasonal demand. Each task needs its own inputs, outcome and follow-up window.
Ask which records were held out of training, whether repeat customers crossed the test boundary, how often the model misses a useful opportunity and whether it beats a simple rule. If a score is described as a probability, compare it with the observed frequency of the outcome.
These are evaluation questions you can apply to any vendor’s claim. This page publishes no model accuracy, client lift or proprietary training-data figures. The AI lead-scoring audit includes a worked example and an open review worksheet.
The person responsible stays visible
Co-founder Alex Lapin brings ML and advertising-algorithm experience to Lonta’s internal models. Alex Kadyrov connects product analytics with measurement. Dan Mash shapes marketing strategy, while Art Sedin connects strategy, delivery and operating processes.
Michael S. leads the marketing team. Account and project managers keep the client’s priorities, approvals and next steps organized. A recommendation still has to fit the service area, the phones the shop can answer and the work the crew can deliver.
The service includes a weekly written brief and a monthly strategy review. See the account review standard for what gets checked and the people behind delivery for their roles.
Your accounts and your business context
Advertising accounts stay under the client’s email, billing and payment method. The website stays on the client’s domain and hosting. Tracking, audiences, keyword lists, recordings, CRM data and creative source files remain in the client’s systems.
The service terms describe ownership, confidentiality, AI use and approval of changes. The privacy policy identifies processing and retention. The free strategy starts with the account’s current records and constraints, before recommending a plan.
Questions about Lonta AI’s technology
Is Lonta AI software that the client has to operate?
Lonta AI is a managed marketing service. The team runs the campaigns, pages, creative, tracking and reporting. The client provides access and business context, while a named account lead owns the plan and delivery.
Does Lonta develop its own machine learning models?
Yes. Lonta develops internal predictive models to support account analysis, with co-founder Alex Lapin focused on machine learning and advertising algorithms. A prediction is an estimate for a defined task; its usefulness depends on the data, evaluation and decision it supports.
Who approves work prepared with AI?
A named person approves changes before they go live, as described in Lonta’s service terms. Specialists consider the client’s budget, service area, capacity and job economics alongside the software output.
Who owns the advertising accounts and client data?
The client owns the accounts and work. Advertising billing stays with the client, and tracking, audiences, recordings, CRM data and creative source files remain in the client’s systems. Access and data handling are described in the service terms and privacy policy.
Can a lead score guarantee a booked or sold job?
No. A score estimates an outcome under stated conditions. Booking, attendance, job completion and collected revenue are separate events. A useful evaluation names the outcome, follow-up window and errors before the score influences a business decision.
Read the evidence behind the questions
Platform and evaluation references checked October 9, 2026. They explain the underlying concepts; they do not certify Lonta’s models.
- Google Ads: offline conversion imports.
- scikit-learn: data leakage and preprocessing.
- scikit-learn: evaluating models on held-out data.
- scikit-learn: probability calibration.
Start with your account’s constraints
Get the free written strategy for your market, service area and measurement gaps. It arrives in three business days, and you keep the findings whether you hire Lonta AI or not.