Loading Adrox
Preparing the operating view.
Loading Adrox
Preparing the operating view.
Multi-touch attribution
A buyer searched on Google, saw a Meta ad, chatted on WhatsApp, opened an email, visited the site — then booked. Switch the model and watch the credit move.
Position-based
The first and last touches get 40% each; the middle steps share the remaining 20%.
USE IT FOR
Valuing both the channel that found the buyer and the one that closed them.
BLIND SPOT
The middle steps — WhatsApp and email here — may deserve more than they get.
Incrementality · Holdouts
Pick a segment, choose how many people to hold back, and Adrox builds a random control group that receives nothing. After the measurement window, the difference is your true lift.
SEGMENT
Past enquirers, last 12 months · 60,000 contacts
After the window · worked example
CONFIDENTTREATED BOOKING RATE
3.6%
CONTROL BOOKING RATE
3.0%
INCREMENTAL CONVERSIONS
324
RELATIVE LIFT
20%
A balanced holdout: enough people held back to read the lift with confidence, while most of the segment still gets the campaign.
Example rates are for illustration. Your test reports your own measured rates.
How to read it
Incremental conversions — treated bookings minus what the treated group would have booked at the control rate.
Relative lift — how much higher the treated rate is than the control rate.
Confidence — how sure you can be the gap isn’t chance. Bigger control groups raise it; they also hold more people back.
Lift test design session
Margin ranking
Contribution margin takes off what return on ad spend ignores — discounts and offers, broker and partner commissions, the cost of servicing leads that never book. Switch the ranking and see the order change.
Ranked by contribution margin: WhatsApp nurture and Google Search rise to the top — low cost to serve, high-quality buyers. Portals and partners fall once commissions and lead servicing are counted.
Thick bar: current ranking. Thin bar: the other measure, for comparison.
Anomaly flags & forecasts
Every metric is watched against its own normal range. Anything outside gets a flag and a likely cause; forecasts come from your history — and where there isn’t enough, Adrox says so.
Meta lead forms fell well below their normal range while Google held steady. The likely cause: a lead form was disconnected during a creative update.
Offline attribution
Real estate still sells offline. Add offline activity by hand or import it from a CSV, and it sits in the same funnel and the same attribution models as your digital channels.
Enter the event once with its dates and cost. Leads captured at the stall carry the event as a touch, so it shows up in attribution and in cost per site visit next to your digital channels.
Works best with
FAQ
First touch, last touch, linear, time-decay and position-based. You can switch between them on any report and compare how credit moves.
Attribution shares out credit for conversions that happened. A holdout keeps a random control group away from a campaign, so you can see how many conversions it actually caused — treated minus control.
You choose from 1% to 99%; 10% is the default. Bigger control groups give clearer answers but hold more people back, and Adrox shows the confidence of every result.
From your CRM — LeadSquared, HubSpot or Salesforce — matched to the original click ID. Adrox reads the stages back; it doesn’t replace your CRM.
Yes. Add events, print and hoardings by hand or import them from a CSV, and they appear in the same funnel and models as digital channels.
Adrox says “insufficient history” instead of guessing. Forecasts and anomaly flags switch on once there is enough measured history.