Marketing budgets for cars and mobility, tied to the deals that close.

The sections below cover long lead times, new customers versus repeat bookings, how far down the map a model can go, what to do when cars run out, and the seasons.

A car purchase starts with research and ends with a test drive, a valuation, a negotiation. The marketing that started it ran weeks before the sale. Measured on same-week sales, upper-funnel channels look weak and search looks strong.

The model is set up with the delay between the marketing and the outcome, so cause and effect line up. When a lead closes much later, for example after a follow-up programme, the sale is credited to whatever brought the lead in once the model is retrained. If search brought the lead, search gets the sale.

Rental and car-sharing businesses live on repeat use. Most bookings in a given week come from people who have used the service before and would book anyway. Paid marketing mostly wins new customers, and that is where its effect is easiest to see.

So the outcome is often new paying customers rather than total bookings, and channels are compared on the marginal cost of the next new customer. That cost, set against what a customer brings over time, is what decides how much to spend and where. A channel with a low cost per booking can look good while mostly reaching people who were coming back anyway.

Mobility services are local. A car-sharing service in two cities runs two businesses with different demand, competition and media, so each city gets its own model with the same setup, and the numbers can be compared side by side. What the first model learns about each channel becomes the starting point for the second.

There is a limit to how far down this goes. Within a small country, splitting into regions leaves too little data per model, so we don't build sub-national models for markets the size of Norway. In larger countries, with several states or regions each the size of a small country, regional models can make sense. Where you want to test a channel locally, a regional experiment run outside the model can still teach it something, and the result goes in as a prior.

A rental company with every car booked, or a dealer short of the models people want, can't sell more however good the marketing is. Weeks like that look like weak media.

It is tempting to add stock or fleet as a variable in the model. It would explain the past better, but future stock is unknown, so it would make scenarios and forecasts less precise. Instead, periods of unusually low or high supply are recorded on the timeline in Odins, so anyone reading the results sees the context, and the model keeps its focus on what marketing does.

Mobility companies often sell through partners paid per booking or per sale, from affiliate networks to comparison and booking sites. Because the cost follows the sales, a model that treats these partners as ordinary channels credits them with sales they did not cause, and can hand them a large share of the budget.

We keep pay-per-sale partners in the model at their historical spend, so the other channels are estimated correctly, and keep them locked in the recommendations. The reallocation is then driven by the channels you actually choose how much to spend on. A partner placement bought at a fixed cost can be modeled like any other channel.

Demand for cars and rentals moves through the year: summer travel, holidays, the spring car market. The model separates that seasonal demand from the effect of the marketing, so a strong month is not read as a strong channel, and the budget can follow the months where the next part of it works hardest.

Beyond the season, we keep controls few: large price changes and launches that are known ahead. Interest rates matter for car sales; central-bank decisions are known ahead and can go in. Things the business can't know for next quarter stay out of the forecasting model.

Odins is a managed marketing mix modeling platform. We connect the ad platforms and your sales or booking data, build and maintain one model per city or market, and review every recommendation before it reaches you. Data flows in continuously and the model is retrained once a month, while scenarios and forecasts are yours to run whenever you need them, in the platform or from your own AI assistant.

Moving budget toward channels with headroom typically gives 5 to 15% more effect from the budget you already have. Companies like Nettbil and Hyre run their marketing budgets on Odins. See the managed service.