Why are Fospha attributed conversions for view-through sources often greater than the last click or assisted conversions in GA?
Fospha’s Ensemble Models identify sales and revenue which are being mis-attributed to channels which have been influenced by a prior event such, as an impression or view through direct, organic, and brand PPC.
The machine-learning models identify the percentage of incremental credit which was influenced by prior impression activity, above the baseline of expected sales across the training window. It then generates many hypotheses about this and then further tests sequentially until it reaches a solution with a high-level of certainty.
This gives Fospha an amount of sales or revenue which can be used as a pool of credit to re-distribute back towards higher-funnel channels which are driving value but will be ignored by Google and unable to be measured with its own attribution due to iOS 14.5.
The model then judges the importance of the role of each of those platforms in generating lower funnel sales or revenue based on a measurement of the strength of the signals between inputs - impressions and spend, and outputs - sales and revenue - during the period of analysis.
How is do the Fospha Ensemble Models calculate performance?
The Fospha Ensemble Models calculate using clicks, post-purchase, and impressions/view data to measure the contribution of the whole funnel. Initial revenue is allocated using clicks and customer-provided information. We use a multi-stage algorithm to re-distribute credit back from non-paid channels to channels (e.g. paid social & display) where impressions play a role in influencing sales but can't be fully measured via clicks.
How do we track impressions?
For any impressions-led channel (Paid Social, Youtube, Display, Discovery), we model the aggregated daily impressions served on each channel to measure the impact on direct, natural search and brand PPC (and sometimes email if the email user-flow is relevant for your brand) traffic and resulting sales. We are then able to measure the relative importance of each impression channel and re-distribute those sales back to the channel, source and campaign that influenced the downstream click landing.
How do we track or attribute YouTube video views?
As above, we model Youtube impressions, which is when a user sees the ad. Total impressions are used, representing the number of people your video has been shown to.
Is there a way to exclude boosted posts that run via Paid Social accounts from the modelling?
If your boosted posts are managed in the same manager account as the rest of your Paid Social activity, then sadly there is no way for us to exclude them from the modelling.
Back to top
How do you track influencers or other channels that we share discount codes for?
We can incorporate signals for how your influencers are performing through our post-purchase modelling, provided you are running specific influencer discount codes or give people the opportunity to say they found you via an influencer in the post-checkout survey. If you would like to set up post-purchase attribution into your Fospha model then please get in touch via the chat window!
Do you connect with our CRM platforms for attribution, such as Ometria for Emails and Attentive for SMS?
The Ensemble Model output will include attribution to Email and SMS wherever you have correctly coded your UTM parameters on landing URLs from this activity. Don’t forget you can use our handy UTM guide, found here!
Where conversions get reattributed for Brand PPC campaigns, why are 'FA Conversions' still visible but the CAC figure is not?
'Brand PPC' is a drain channel used in the Fospha Ensemble Model, as for new customers there is an ad interaction before they are aware of the brand to be able to search for it by name. We reattribute conversions coming via Brand PPC on a click basis back to the impressions activity that caused the customer to perform the Brand PPC search in the first place.
We see varying levels of Brand PPC being drained by the model from client-to-client, and because it is a paid channel, there are some scenarios that the reduced conversions post-impressions modelling cause the CPP, CAC, and ROAS to show incredibly inflated/deflated values. Imagine that you've spent £3000 on Brand PPC, and the conversions are showing as 0.25 after reattribution - this would mean the CPP would be £12,000, and can skew visualizations in the dashboards. For this reason, we have chosen to hide the calculated performance metrics for Brand PPC.
If you include Brand PPC performance in your overall Google performance, then yes, you are able to calculate the CPP, CAC, and ROAS using the data exports.