Run RTB House from Your Dealership CRM

Deep learning bidding aimed at considered purchases, sold behind a contract, and its real standing with the ad gateway.

RTB House buys retargeting inventory across the open exchanges, the same broad territory as every other platform in this family. What it sells as its difference is the bidding layer. Rather than rule based bidding tuned by an account manager, it runs deep learning models that decide who to bid on, how much, and which creative to render, and it retrains those models continuously against outcomes.

The claim attached to that architecture is that it performs better on considered purchases than standard dynamic retargeting does. The reasoning is worth understanding because it is not marketing noise. A rule based system optimises toward the click, which suits a low value item somebody buys on impulse. A model trained on longer conversion paths can learn that a particular visitor pattern predicts a purchase five weeks out and bid accordingly instead of chasing whoever is closest to clicking today.

A vehicle is the archetypal considered purchase. Weeks of research, multiple people involved, several site visits, a large sum of money. If the argument for model driven bidding holds anywhere outside high end retail, it holds in automotive. The wider channel case sits on programmatic display advertising for dealers and is not repeated here.

The case is the mismatch between how most retargeting optimises and how a vehicle is actually bought. Standard dynamic retargeting spends most of its budget on the people most likely to click soon, which in automotive means bidding hardest on browsers and shoppers who were coming back anyway. A system optimising for the eventual purchase rather than the next click distributes money differently, and on a long consideration cycle that difference compounds.

The honest objection is that you cannot inspect it. A learned model is a black box by construction, and neither your store nor we can audit why it bid what it bid. When performance is good that is a comfortable trade. When performance drops there is no lever to pull and no explanation to give a general manager, only a request to give it more time and more data. A dealership that needs to understand its media should think hard before buying a system that structurally cannot explain itself.

The second objection is commercial. This is not a platform a single rooftop opens on a Tuesday afternoon. Access sits behind a contract, and the commitment is designed for advertisers spending at a level most independent stores do not reach in a quarter. If the model needs volume to learn, a small budget is also the worst possible way to test it, which means the cheap trial that would tell you whether the claim holds is not really available. Set expectations against what to expect from car sales leads before anybody signs anything.

Available on request, which on this directory is a status with a published test rather than a polite deferral. The test is that no adapter file exists in the LeadLocate ad gateway for the platform. That is the case here. There is no partial implementation, no stub returning encouraging numbers, and no code anywhere in the CRM that has ever spoken to their API.

Access class is CERTIFIED, which describes the vendor's gate rather than ours and is the strictest of the three on this directory. A contract, a certification or a minimum spend stands between any party and API access. That is a commercial negotiation, not a form, and it does not move because a dealership is impatient.

Keeping the two apart matters, because they fail independently. If the adapter existed tomorrow, the contract would still stand between your store and a live connection. A signed contract, equally, is worth nothing without an adapter to use it. Two gates, two owners. On a CERTIFIED platform the vendor's gate is usually the slower and the more expensive of the pair, and any vendor presenting this kind of platform as something they can simply switch on for you is skipping the half that costs money.

Stated as a design rather than a description, because none of it runs and none of it is scoped. The same four row shape appears on every page in this directory, so a platform we have built and one we have not can be compared line for line.

On a platform whose entire proposition is a model learning from behaviour, the bottom row deserves a second read. What a model can learn from is bounded by what reaches it, and the section below is specific about what would not.

If this platform were built, it would not be a special case. Every network answers the same contract.

A single library inside LeadLocate defines what any advertising platform must be able to do: hand us an authorization URL, exchange the returned code for a token, list the accounts you own, deploy a campaign, pause or resume it, return spend and results, and prove on demand that the connection still works. Seven operations, one adapter file per network. That shape is why adding a platform is scoped work rather than an open ended project, and why turning one on does not disturb the ones already running.

Credentials are encrypted before storage using AES-256-GCM. Every action taken against a network is written to an append only event log carrying the company, the network, the user and the detail, stamped in UTC on purpose, because these tables are shared across dealerships and a timezone relative timestamp is one nobody can reconcile afterwards. Campaign objects record which credentials created them and re-resolve those same credentials later, so a credential withdrawn at the end of a contract does not orphan the work built under it.

Twenty networks are registered against that contract and one is implemented. RTB House is not among the twenty at all, which is what available on request means here rather than a softer way of saying soon. Every entry in this directory carries a status with a stated test beside it, including the ones that say no.

There are two ways to advertise through LeadLocate and you pick.

The first is the webXintel Ads Network. It is included, our team builds and operates the campaigns from the selections you make in Leads Manager, and it arrives as one line rather than a stack of platform invoices. Nothing to set up, no contract to negotiate, and no dependency at all on the state of this page.

The second is your own account, connected to the CRM so campaigns sit on your billing, in your name, with your spend history intact. That is the path this page documents, and for RTB House it does not exist. If your store or your agency already holds a contract with them, it stays exactly where it is and nothing inside the CRM speaks to it.

One limit stated plainly, because it holds across the whole gateway even where a network is fully built: pushing, pausing and reading numbers from a network is done by our team today, not from your dashboard. That is a deliberate constraint while the connection layer is young, and opening it up is queued work rather than a promise.

Three things in order, and only one of them is engineering.

First the commercial gate. A contract, a certification or a minimum spend has to be agreed before API access is issued, and on a CERTIFIED platform that conversation sets the floor for everything after it. Second the adapter: the seven operations written against their endpoints, a credential block added to the gateway configuration, and a verification call that proves a connection rather than assuming one. Third the setup on your side, meaning an advertiser entity, a billing relationship and whatever site audience the campaigns are meant to work from.

On measurement, this platform needs a longer and more honest window than most. Delivery data reads quickly because impressions accumulate fast. The thing you are actually buying, a model that learns your buyers, cannot be judged on that timescale at all. A learning system given two weeks and a small budget will look mediocre whether or not it is good, and the same system given a fair test is a meaningful commitment of money before you know the answer. That asymmetry is the real cost here, and it deserves to be in the plan rather than discovered halfway through a contract. None of this is a promise about volume or about what any of it closes at, and nobody can promise those.

Every gap below is true whether or not an adapter is ever written, and on this platform one of them undercuts the core pitch.

We do not emit a vehicle catalog or an inventory feed to any advertising platform. Deep learning creative optimisation renders individual products, and it does that from a feed. There is no code anywhere in our stack that generates one. Without a feed, the model would be choosing among generic creative rather than among your units, which is not the product anybody is buying when they choose this vendor. Inventory Link ingests your feed and merchandises your units on your own properties, a different job entirely, and that gap would remain after any adapter work.

We do not install or manage pixels. A retargeting model has nothing to learn from without site behaviour reaching it, and the tag that carries that behaviour lives on your website. Placing it, maintaining it through site changes and keeping consent handling correct are not services we provide.

We do not run a server side conversions API and we do not upload offline conversions. A model optimising toward purchases needs purchase outcomes fed back to it, and feeding them back is precisely what we have no code to do. On this platform that is the sharpest gap on the list.

We do not export a hashed audience. Your customer records stay inside the CRM.

No campaign has ever been deployed to RTB House from LeadLocate, because no code exists that could deploy one, and no credential of theirs has ever been held by us.

Say so, and say it with your store attached. Requests are recorded against the platform they name, and the tally is the input to build order, because a dealership ready to spend on a channel is a better signal than any internal roadmap meeting.

What a request buys you here is an honest reading of whether the platform fits your store at all. On a CERTIFIED platform the first question is not engineering, it is whether your monthly media budget clears the commitment and whether your site traffic is deep enough for a model to learn anything. If the answer to either is no, hearing that in a week is worth considerably more than a position in a queue.

If the honest answer turns out to be that a lower barrier retargeting platform, or the one network already connected, gets you most of the way for a fraction of the commitment, you will hear that too, because selling a build you do not need is a bad trade for both of us. Meanwhile the managed path is live and waits on none of this, and pricing sits on the pricing page without changing based on which networks you turn on.

Does LeadLocate integrate with RTB House today?

No. The published test for available on request is that no adapter file exists in the ad gateway for the platform, and that is the case here. Nothing in the CRM has ever called their API and no campaign has ever been deployed to them from our side.

Can I use my own RTB House ad account?

Not through the CRM. Bringing your own advertising account is the half of the product this directory documents platform by platform, and for RTB House that half is empty. Any contract you hold with them operates entirely outside LeadLocate.

What happens if I request RTB House?

It is recorded against RTB House specifically, not against retargeting in general, and counted. That tally is the input to build order, and on a platform gated by a contract you will also get a straight reading of whether your store clears the commercial floor.

Is deep learning bidding actually better for a dealership?

The argument is plausible for long consideration purchases, and a vehicle is the clearest example there is. What nobody can tell you is why the model bid the way it did, so the trade is potential performance in exchange for an explanation you will not get.

Why does the CERTIFIED label matter if the adapter does not exist anyway?

Because the two gates fail independently and the vendor's is usually the expensive one. A contract, a certification or a minimum spend is required before API access, and clearing it alone still leaves you without an adapter, just as an adapter alone leaves you without access.

Could the model optimise against my actual closed deals?

Not through us. That would need outcomes fed back through a conversions interface, and we have no code that uploads offline conversions to any platform. Finishing an adapter would not close that gap, which matters more here than on most entries.

Requests are counted against the platform they name. Tell us your monthly media budget and how much website traffic your store sees, and you will get a straight answer about fit before anything is scheduled.