Answers, before you have to ask for them.
Nineteen questions we get in almost every first call — answered the way we would answer them on the call.
Product What does Specicon actually return?
One decision per request — the chosen option, a score between 0 and 1, and a short reason assembled from the dimensions that produced that score. The reason is part of the response, not a tooltip generated afterwards.
How the decision core works →Product Is this a recommendation engine?
It ranks options, so the output can look similar. The difference is that a recommendation is returned only when the system can say why; below the confidence floor it returns nothing rather than a guess. Roughly 2% of requests are held back for that reason.
The seven stages →Product Do we have to replace our commerce platform?
No. Specicon reads from the systems you run and answers over one endpoint. The first surface usually goes live behind a feature flag without touching the catalogue of record.
Integrations →Product What are the six modules?
Catalogue Intelligence, Product Intelligence Score, Content & GEO Studio, AI Visibility Monitor, Conversational Commerce and Revenue Intelligence. They all read the same Product DNA, so a fact fixed once improves every one of them.
Commerce Intelligence →Pricing & plans How is Specicon priced?
A monthly platform fee for your tier, plus metered decisions beyond the included volume, plus any module add-ons. Seats are never charged — adding console users costs nothing.
Tiers and rates →Pricing & plans What counts as one billable decision?
One request that returns a recommendation. Held-back decisions are not charged. Retries inside a single request count once. Console browsing and reporting are free.
See the ledger →Pricing & plans What happens if we exceed the included volume?
Nothing stops. The overage is metered at your tier rate and appears as a separate invoice line. Above roughly two months of consistent overage, moving up a tier is usually cheaper — the console says so directly.
Metering and forecast →Pricing & plans Can we start smaller than an annual contract?
Yes. A pilot is one scoped decision on one surface, measured against a holdout for a month. Most customers convert to an annual term after the first readout.
Scope a pilot →Data & privacy What data do you need, and what do you never store?
Consented behavioural events, catalogue attributes and business context. We hold a pseudonymous dimensional identifier — not names, addresses or payment details. Re-identification stays on your side.
Data Processing Agreement →Data & privacy Where is our data processed?
ap-south-1 (Mumbai) by default, with eu-central-1 and us-east-1 available on enterprise contracts. Residency is contractual, and a migration window is agreed in writing before anything moves.
Sub-processors →Data & privacy Do you train shared models on our data?
No. Your catalogue, signals and outcomes improve decisions for your account only. Nothing crosses tenant boundaries.
Trust & Security →Data & privacy How do we delete our data?
From Settings in the console. Deletion requests are queued for seven days and confirmed by email; decisions already invoiced keep their audit record, which is a legal requirement rather than a preference.
Data controls →Implementation How long does implementation take?
Ten working days is typical: keys and catalogue feed in days one to three, Product DNA and guardrails by day six, one surface live behind a flag by day eight, holdout split and first readout by day ten.
The twelve steps →Implementation What do we need on our side?
One engineer for the integration, one merchandiser or category owner for the decision logic, and read access to the catalogue and event stream. No data warehouse project is required.
Quickstart →Implementation Can we test before going live?
Yes — sandbox and staging environments with their own keys, and a replay mode that runs historical requests through the current logic so you can compare before shipping.
Environments and keys →Support & SLA What are your response targets?
Severity 1 in 15 minutes with a page to the on-call engineer, severity 2 in two hours, severity 3 in one business day. Enterprise contracts add a named engineer and 24/7 coverage.
Support and SLA →Support & SLA What uptime do you commit to?
99.95% monthly on the Expand tier and above, with service credits when we miss. Launch runs on the same infrastructure but carries no contractual SLA — we will not sell a promise a four-week engagement cannot honour. Measured availability over the last 90 days is not measured; the status page carries the full incident history and every credit we have issued.
Live status →Support & SLA Who do we talk to when something looks wrong?
The people who built your implementation — a solutions architect and a customer success lead named on your account, plus on-call engineering for anything urgent. No routing queue.
Talk to us →Data & privacy Why does the holdout change size after the pilot?
A pilot runs 50/50 because a four-week window needs a large control to reach significance. In production the control drops to 10% — still enough to measure, small enough that you are not withholding a working decision from half your customers. Every reported number names the mode it came from.
How results are measured →