why oddly

Not the single-store crowd.

Most store tools install on your store and read your store. They can tell you your numbers. They cannot tell you whether your numbers are good, because they have nothing to compare them against.

oddly reads a live cohort of stores like yours. So a conversion rate, an average order value, an inventory-health score arrives with the one thing that decides whether it is a problem worth fixing: where you sit against everyone running a store like yours. That comparison is built from many stores, which is something a tool that only ever sees one store cannot do.

reads the whole cohort. shows you your store in it.

Start free See how oddly compares

The single-store ceiling

A tool installed on one store can only ever read that one store. It will happily tell you your conversion rate is 2.1 percent, your average order value is 48 dollars, your stock is turning every nine weeks. What it cannot tell you is the part you actually need: is 2.1 percent good for a store like yours, or is it quietly the reason last month was flat.

Without that context, every number is just a number. You end up guessing whether a metric is a problem or simply normal for your category, and you only find out you were wrong after a season of margin has already walked out the door. The ceiling is structural: the answer lives in other stores, and a single-store tool was never able to look there.

The cross-store benchmark

oddly groups your store with stores like it, by category and shape, then shows you where you land against the cohort: top quartile, middle of the pack, or behind. Your 2.1 percent stops being a lonely number and becomes a percentile, and a percentile is something you can act on. If you are behind the cohort on average order value, that is a concrete, ranked opportunity. If you are ahead on conversion, that is one less thing to worry about.

This is the part of oddly a single-store tool is structurally unable to copy. A cross-store benchmark is built by aggregating across many stores, and a tool that installs on one store at a time has only ever seen one. The benchmark is the data layer underneath oddly's diagnostics, and it gets sharper as the cohort grows.

It is built to be safe to take part in: aggregated, opt-in, anonymized, and auditable. The next section is exactly how each of those holds.

See where your store lands

How the benchmark stays honest

A cross-store benchmark is only worth having if a store can take part without giving anything away. oddly's is built on four rules, each enforced in the product rather than promised in a policy.

Opt-in. A store sees the cohort comparison only when it chooses to contribute its own aggregates. Opt out and you keep your own dashboard; you simply do not see the cohort half. Nothing is contributed by default.

Anonymized. Each contributor is reduced to an opaque hash before anything is aggregated, so the benchmark is counted without the math ever touching a store's identity. What goes into the cohort are summary figures, never your orders or customers.

Minimum cohort size. oddly will not show a comparison built from fewer than five distinct stores. Below that floor the cohort is held back, so no single store can be reverse-engineered out of the result.

Auditable. The benchmark is computed deterministically from a fixed method, not estimated by a model, so a given comparison can be traced back to the rule and the aggregates behind it rather than taken on trust.

More than a benchmark: the deterministic layer

The cross-store benchmark is the data moat. The reason you can trust what oddly does with it is the layer underneath. oddly is a deterministic accountability layer for commerce automation, and everything it surfaces or applies rests on four guarantees.

Deterministic, not inferred. Verdicts come from your store's own data and a fixed set of rules. The same situation always produces the same call, traceable to the rule and the signal. No generative model sits in the decision path. No AI guessing.

Decrease-only, by design. Reducing, pausing, or blocking spend is automatic; increasing, granting, or escalating always needs a human. On the paid ad-channel tiers oddly can only reduce or pause spend, never raise it. The asymmetry is hardcoded, not a setting.

Held-mutate. oddly dry-runs a change and waits for your explicit approval before applying it, and every change can be undone within 24 hours. It proposes and verifies; you dispose. On the free Watch tier it only watches and reports.

Audit as a contract. The event log is a product surface, not plumbing. Every action is recorded with the rule that fired and the signal behind it, so the record is inspectable rather than reconstructed later.

Read how oddly thinks for the long version of the four guarantees.

One engine, every signal

Money rarely leaks inside one tool. It leaks in the seam between them: demand rising in search for a product that is about to stock out, or ad spend still running against a listing that stopped converting. A tool that watches one source at a time never sees the seam. oddly reads them through one engine, so a finding can cross sources.

Shopify. Store health, inventory, margins, listings, orders.
Google Ads and Meta Ads. Spend, with a decrease-only hand on the dial.
GA4. Sessions and conversion behaviour alongside the store.
Search Console. The demand showing up before it reaches the store.

Because the same engine reads all of them, an opportunity can join the dots: capture the demand search is showing while the stock is still there, and stop the spend chasing a page that no longer earns it.

oddly against a single-store tool

A plain capability comparison. Not against any named product, against the shape of a tool that installs on your store and reads your store alone.

Capability	A single-store tool	oddly
Reads your own store	Yes	Yes
Tells you if a number is good for your category	No, it has nothing to compare against	Yes, against an opt-in cohort of similar stores
Verdicts you can trace to a rule	Varies; often a model or a chart you read yourself	Deterministic, traceable to the rule and the signal
Carries the fix, with your approval	Usually reports only	Held-mutate: dry-run, approve, reversible in 24 hours
Full audit log of every action	Rarely	Audit as a contract
Shopify, ads, GA4, and Search Console in one engine	One source at a time	Correlated, so findings cross sources

For head-to-head pages against specific tools, see the comparisons. For oddly's cross-store benchmark as a primitive other systems can call, see the platform.

See where you actually stand

Start free on Watch. Connect your Shopify store and oddly shows you what it catches, and once you opt in, how you compare against stores like yours. No charge to look, and nothing contributed until you choose to.

Start free See pricing

Common questions

What makes oddly different from a single-store analytics tool?

A tool installed on one store can only ever read that one store. It can tell you your conversion rate is 2.1 percent; it cannot tell you whether 2.1 percent is good or poor for a store like yours. oddly reads an opt-in cohort of similar stores and shows you where you sit against them, so the number comes with the context that decides whether it is a problem worth fixing.

How does oddly keep the cross-store benchmark private?

Three ways. It is opt-in: a store only sees the cohort comparison when it chooses to contribute its own aggregates. It is anonymized: each contributor is reduced to an opaque hash before aggregation, so the comparison never touches a store's identity. And it has a minimum cohort size: oddly will not show a comparison built from fewer than five distinct stores, so no single store can be reverse-engineered from the result.

Does oddly use a generative model to make decisions?

No. oddly's verdicts come from a fixed set of rules and your store's own data, so the same situation always produces the same call, traceable to the rule and the signal behind it. No generative model sits in the decision path.

What does oddly mean by one engine, every signal?

oddly reads Shopify, Google Ads, Meta Ads, GA4, and Search Console through one engine, so a finding can cross sources: rising search demand for a product that is about to stock out, or ad spend running against a listing that is no longer converting. Tools that watch one source at a time cannot see across the seam where the money usually leaks.