Work out whether the change made a difference
Compares how the changed pages moved against how the comparison pages moved over the same days, and tests that gap against a thousand reshuffles of which pages were which, within each matched group. A gap that a reshuffle produces easily is reported as flat, never as a small win.
POST
https://app.softsolz.uk/api/v1/services/seo/experiments/{experiment_id}/measureRecent Requests
Kept only in this browserTimeStatusTook
Send a request with Try it to see it here.
Path Params
experiment_idstringrequiredHeaders
AuthorizationstringrequiredBearer token: Bearer sk_live_... (or sk_test_... for sandbox). A login token will not work here.
Response
Response Body
verdictstringwin, loss, flat or inconclusive.p_valuenumber | nullNull when inconclusive.effect_clicksnumber | nullExtra clicks per changed page.effect_pctnumber | nullThat as a percentage.reasonsarray of stringsWhy it is inconclusive, when it is: not_enough_days, not_enough_controls (fewer than min_controls_per_page comparison pages per changed page) or no_treatment_pages.min_daysintegerThe shortest window this experiment reports on.min_controls_per_pageintegerComparison pages needed for each changed page.quasi_experimentbooleanTrue when the pages were matched rather than randomised.measured_overobjectThe before and after windows.notestringWhat this design can and cannot show.Request
curl --request POST \ --url https://app.softsolz.uk/api/v1/services/seo/experiments/{experiment_id}/measure \ --header 'Authorization: Bearer sk_live_your_key'Credentials
Sent as a Bearer token and used in the samples above and in Try it. Kept only in this browser tab until you close it. Use an sk_test_ key to stay in your sandbox.
Response · 200
{ "data": { "verdict": "win", "p_value": 0.04, "effect_clicks": 42.5, "effect_pct": 18.2, "reasons": [ "not_enough_days" ], "min_days": 21, "min_controls_per_page": 3, "quasi_experiment": true, "measured_over": {}, "note": "We compare the pages you changed with similar pages you left alone, picked by kind of page and by traffic, not at random. So a result shows what your changed pages did next to comparable pages, not proof that the change alone caused it." }}