Platform · Marketing attribution
One deal, six SAGARIS models, six answers
Multi touch revenue attribution across email, calls, meetings, SMS, LinkedIn and microsites. Every closed won deal is scored under all six models at once, so you can finally settle which channel earned the money. It covers the revenue your team creates through those channels rather than advertising spend.
Included at $499 per seat per month
One deal, scored six ways
credited revenue
One closed deal, five touches, six models. LinkedIn opened it, so it earns the whole 60,000 under first touch and nothing under last touch, from the same five rows. Every row still adds up to the deal.
How the credit is computed
Every model here is arithmetic you can check.
Linear is where the dashboard starts, because it takes the fewest positions. The one we added is meeting multi touch, a W shape: thirty percent to the first touch, thirty to the last touch before the meeting, twenty to the meeting itself, ten across the middle and ten across everything after. All six are this explicit, and you can read every one of them off the same deal.
Meeting multi touch on a real journey
role weight
The weights are summed per touch and then divided by their own total, so an empty bucket cannot leak revenue. A journey with no middle touches, or no meeting at all, still distributes the whole deal rather than most of it.
Six channels, one writer
Fourteen production paths record touchpoints and every one of them goes through a single shared emitter. It never throws, so attribution can never break a send or a dial, and it checks for the row before inserting one, so a retried webhook cannot double count a touch. The window is the last ninety days by default and it is a parameter on the request, not a setting anyone has to rebuild.
- Calls
- Meetings
- SMS
- Microsites
Time decay, seven day half life
weight per touch
days before the deal closed
A touch a week before the close counts half as much as one on the day, a fortnight before it a quarter, three weeks before it an eighth. That is the whole model. It is why a long cycle looks so different under time decay, and seeing both is how you pick the one your board will accept.
Why this is different
It reads the rows the dialer wrote, not a copy of them.
Attribution here is not a reporting layer bolted over somebody else's data. The dial, the send, the reply and the booked meeting each write a touchpoint as they happen, and the attribution reader queries those same rows. There is no export, no ETL, no nightly sync, and no window where the data has not landed yet.
Reporting on a copy
answers arrive tomorrow
The work happens in one system, is copied into another overnight and becomes a report in the morning. Four of the five stages are transport, and each one is somewhere the answer can go quietly stale.
SAGARIS
answers arrive now
The call that ended at nine is a row the report reads at nine oh one. Same table, same request, nothing in between to fall behind or drop a day.
All six, side by side
One request scores the same deals under all six models at once. You can see how much of your answer comes from the data and how much comes from the model you picked. That is the difference between a number you present and a number you can defend in the room.
Credited revenue, not activity
Both readers start from closed won opportunities and split that deal's actual value across the touches on it. A channel's number is dollars it was credited with, not emails it sent, and every model conserves the total: the rows add up to the deal.
Credit survives a mailbox handoff
When a conversation moves from the outbound mailbox to a corporate one mid journey, the touches before the move keep their credit under every model. A test derives every write path from the source tree and fails if any of them starts keying on which mailbox holds the thread. Your history stays whole.
The questions buyers actually ask.
- MOST ASKED
Outbound revenue, across email, calls, meetings, SMS, LinkedIn and microsites. The six models take two inputs: the touches recorded against a deal, and that deal's closed value. So you can see which channel your closed revenue actually came from. That settles the argument about whether the calls or the emails did the work.
Six: first touch, last touch, linear, time decay, U shaped, and a meeting multi touch model that is W shaped. All six are rule based arithmetic you can check line by line. Linear is where the dashboard starts. Anyone on your team can reproduce a split by hand. That is what makes the number arguable in a pipeline review instead of taken on faith.
U shaped gives forty percent to the first touch, forty percent to the last and twenty percent split equally across everything between. Two touches make it fifty fifty and a single touch takes all of it, so the degenerate cases are defined rather than left to a divide by zero. Time decay uses a half life of exactly seven days: a touch a week before the close carries half the weight of one on the day, a fortnight before it a quarter, three weeks before it an eighth.
A microsite is generated for one named prospect, so a view on it becomes a touchpoint on that contact alongside the call and the reply rather than an anonymous session. The visitor itself is stored as an opaque salted hash that cannot be turned back into a person. Attribution starts from a contact you already know and a deal that already exists, so every credited touch has a name behind it.
Closed won opportunities. Both readers query deals that reached the closed won stage inside the window and split each deal's actual value across the touchpoints recorded against it, so a channel's figure is dollars it was credited with rather than a count of activity it produced. Open pipeline contributes nothing until it closes, which is why this number is smaller than an influenced pipeline number and worth more.
Fourteen production paths record touchpoints and every one of them goes through a single shared emitter. It checks for an identical touch on the same deal, contact, channel, type and instant before inserting, so a retried webhook or a replayed job cannot double count. It also never throws: if attribution cannot record something, the send or the dial still completes and the failure is logged rather than surfaced to the person doing the work.
Ninety days by default, and it is a parameter on the request rather than a setting somebody has to rebuild. A long enterprise cycle and a fast transactional one can be looked at over the same data without changing anything about how it was collected.
Yes, and nothing is recomputed destructively. Every model reads the same touchpoint rows every time, so switching one changes the answer and never the record. That is also why the comparison can hand you all six at once: they are six readings of one ledger, not six pipelines somebody had to configure in advance.
Bring one closed deal. We will score it six ways.
Pick a deal your team has already argued about and we will run its real journey through all six models in front of you. You will see exactly where the models disagree and why. That is the argument settled, on your own numbers, in one meeting.
In this topic
CRM, pipeline and revenue data
- The SAGARIS CRMA pipeline that updates itself from your calls, email and SMS.
- CRM documentationContacts, leads, opportunities and stage moves.
- Signals and memory documentationSignals derived from a contact's own history, nothing bought or guessed.
- Data enrichment with provenanceFour lookups per contact, and every field records where it came from.