Start a project Sign in
Home/Results
Results and measurementExample data

A decision is only worth what the outcome proves.

Specicon connects every decision to what happened next — and keeps influenced, attributed and incremental results clearly apart.

Proof, not adjectives

A decision is only valuable when the outcome is better.

Every recommendation carries an identifier, so the purchase, the return and the repeat all attribute back to it.

The decision journey

Ten points where a decision can be judged.

ten judgement points, four phases

Two holdout modes, named on every number. A pilot runs 50/50 — half the traffic sees guided decisions, half sees the status quo — because a four-week window needs a large control to reach significance. In production the control drops to 10%: enough to keep measuring, small enough that you are not withholding a working decision from half your customers.

phase 1 Need
01 — Customer need the question behind the visit
02 2.84M / mo Decision delivered one option, one reason
phase 2 Engagement
03 96.2% Option viewed the answer was actually seen
04 41.8% Compared held against alternatives
05 3.82% Accepted the customer took it
phase 3 Outcome
06 3.61% Completed order placed, service booked
07 −1.8pp returns Retained kept, not returned
08 4.6 / 5 Rated the customer judged it
phase 4 Learning
09 +11.4% lift Future decisions improved the outcome re-weights the next one
10 v14 → v15 Rules refined a new rule version ships only after it beats the holdout
points 1–2 are ours to control3–5 are the customer’s choice6–8 are the business result9–10 are what compounds
Eight views of performance

Whose problem got smaller?

View 01

Decision performance

  • Decision acceptance
  • Confidence distribution
  • Explanation engagement
  • Alternative selection
  • Exclusion reasons
View 02

Customer performance

  • Time to decision
  • Satisfaction
  • Confidence
  • Suitability feedback
  • Complaint reduction
View 03

Operational performance

  • Manual effort reduced
  • Faster routing
  • Better qualification
  • Lower decision friction
  • Improved data readiness
View 04

Financial performance

  • Assisted conversion
  • Influenced revenue
  • Incremental gross profit
  • Return impact
  • Repeat business
View 05

Data readiness

  • Option DNA readiness
  • Information completeness
  • Missing-data reduction
  • AI readability
  • Content coverage
View 06

Trust and governance

  • Reason coverage
  • Human override rate
  • Audit requests answered
  • Rule changes rolled back
  • Policy exceptions
View 07

Learning performance

  • Outcome signals captured
  • Feedback-to-tune cycle
  • Holdout lift
  • Drift alerts resolved
  • Rule versions promoted
View 08

Team performance

  • Advisor-assisted decisions
  • Time to answer
  • Escalation rate
  • First-contact resolution
  • Specialist time saved

Model-performance decomposition is not published. What is shown is the business result and the action it suggests.

Attribution

Three words that are not synonyms.

Widest

Influenced

Revenue where a Specicon decision appeared somewhere in the journey. The widest number, and the least conclusive one.

Narrower

Attributed

Revenue where the decision was the last meaningful step before the outcome. Useful for comparing surfaces against each other.

Conclusive

Incremental

The measured difference against a comparable group that did not receive the Specicon decision. The only number worth putting in a board pack.

Influenced revenue is not incremental revenue. We will not report it as though it were.

Experiments and holdouts

How we tell Specicon apart from a good month.

Specicon can compare customers who receive a Specicon decision with a similar group receiving the existing experience. That comparison is what determines whether the improvement came from Specicon rather than normal business fluctuation.

Control group

Your existing experience, unchanged, for a comparable slice of traffic over the same period.

Specicon group

The same journey with explainable decisions in place, on the surface you chose to start with.

Measurement period

Agreed before launch, long enough to cover your normal purchase or decision cycle.

Confidence

Reported with the result. A difference we cannot yet distinguish from noise is reported as exactly that.

Business conclusion

A plain answer to one question: keep it, expand it or stop.

Experimentation methodology and assignment logic remain proprietary.

Results dashboardExample data

What the monthly review looks at.

Decision performance · demo merchant · 30 daysSample data
Decisions served48,210 
Decision acceptance61%+4.2 pts vs control
Time to decision−38%vs control
Return rate−11%vs control
Strong fit54%
Moderate fit31%
Low confidence — held back15%

Figures update as outcome data arrives over the agreed measurement period.

Outcomes by industry

Success is not the same word everywhere.

5 outcomes

Commerce

  • Purchase
  • Gross profit
  • Return
  • Repeat purchase
  • Satisfaction
4 outcomes

Education

  • Enrollment
  • Completion
  • Skill attainment
  • Employment outcome
4 outcomes

Real estate

  • Qualified lead
  • Property visit
  • Transaction
  • Satisfaction
4 outcomes

Travel

  • Booking
  • Trip completion
  • Satisfaction
  • Repeat booking
4 outcomes

Insurance

  • Suitable product selection
  • Process completion
  • Retention
  • Human-reviewed outcome
4 outcomes

Services

  • Appointment
  • Completion
  • Satisfaction
  • Repeat usage
4 outcomes

Marketplace

  • Seller match quality
  • Listing conversion
  • Dispute rate
  • Repeat transaction
4 outcomes

Telecom

  • Plan fit
  • Activation
  • Churn avoided
  • Support contacts
4 outcomes

Media

  • Watch or read start
  • Completion
  • Subscription retained
  • Recommendation trust
4 outcomes

Automotive

  • Configuration completed
  • Test drive booked
  • Finance accepted
  • Service return
4 outcomes

Hospitality

  • Room or table fit
  • Booking completed
  • Stay rating
  • Direct rebooking
4 outcomes

Banking

  • Product suitability
  • Application completed
  • Default avoided
  • Human-reviewed outcome
Next move

Pick one decision. Agree how it will be judged.

We will write the measurement plan before the build starts, including what would count as a failure.

Design a measurement planSee the evidence so far