When an algorithmic channel makes sense
With steady, large traffic and a catalogue with plenty to choose from. The model learns from data, so a small shop with little traffic will not give it enough material and the result will be random. This is not a channel for the start, but for the stage when traffic acquisition is already working.
Why we ask about incrementality, not ROAS
Because the model optimises for purchase probability, so by definition it targets the people closest to a decision. The dashboard can then show a great return that the business does not see in its profit and loss. The only answer is a test with the channel switched off for part of the traffic.
What we can really influence
The product feed, exclusions against the other channels, the budget and how it is measured. That sounds less impressive than bid optimisation, but in an algorithmic channel these four things account for most of the difference in results. JustIdea controls these four things; the vendor's model controls everything else.
Deep learning, AI and first-party data
RTB House is a DSP platform that uses deep learning, meaning deep neural networks, to buy ad space and choose products, instead of manually set rules. This kind of artificial intelligence learns from first-party data from your shop: products viewed, baskets and orders. For the advertiser it means few settings and a heavy dependence on data quality. AI on the platform side does not remove the need to work on the feed, the measurement and the incrementality test. RTB House product ads show on publishers' sites across the open internet.
