Test attribution with fake traffic in test mode

1. Point a script at test mode, and only test mode

Create an sk_test_ key in the dashboard and put it in its own variable. Every test script should refuse to run with any other key. A seed script pointed at a live key would write hundreds of fake visits and orders into your real numbers, and there is no undo button for that.

// scripts/_vevee-test.ts
import { createClient } from "@vevee/sdk";

const key = process.env.VEVEE_TEST_KEY ?? "";
if (!key.startsWith("sk_test_")) {
  console.error("Refusing to run: VEVEE_TEST_KEY must start with sk_test_");
  process.exit(1);
}

export const vevee = createClient({ apiKey: key });

2. Probe each call once

Before seeding anything, send one of each: an anonymous $pageview with UTMs, an identify with isSignup and firstTouch, an order and a refund. Each should succeed, and the identify response should say firstTouchStored: true. If it says firstTouch: "ignored", the call was missing isSignup: true.

import { randomUUID } from "node:crypto";
import { vevee } from "./_vevee-test";

const visitId = randomUUID();
const userId = "probe_" + Date.now();
const ua = "Mozilla/5.0 (iPhone; CPU iPhone OS 18_0 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/18.0 Mobile/15E148 Safari/604.1";

await vevee.analytics.capture({
  event: "$pageview",
  properties: { path: "/", visit_id: visitId, utm_source: "facebook", utm_medium: "paid_social", utm_campaign: "probe", landing_path: "/" },
  context: { geo: { country: "IT", city: "Milan" }, userAgent: ua },
});

const id = await vevee.analytics.identify(userId, { plan: "free" }, undefined, {
  isSignup: true,
  firstTouch: { started_at: new Date().toISOString(), utm_source: "facebook", utm_medium: "paid_social", utm_campaign: "probe", landing_path: "/" },
  context: { geo: { country: "IT" }, userAgent: ua },
});
console.log("firstTouchStored", id.firstTouchStored); // expect true

const orderId = "probe_ord_" + Date.now();
console.log(await vevee.revenue.order({ distinctId: userId, orderId, amount: 499, currency: "EUR", billingReason: "purchase" }));
console.log(await vevee.revenue.refund({ orderId, refundId: orderId + "_r1", amount: 499 }));

3. Sign up once from a real browser

Run a local build with your test key, open it on a URL like /?utm_source=facebook&utm_campaign=test, and sign up with a fresh account. In the dashboard (switched to test mode) the new person should have first_channel set to paid. While you are there, open the browser's storage panel: there should be no Vevee cookie and no Vevee key in localStorage or sessionStorage. Keys your own app writes for other purposes, such as a dashboard tour counter, are not analytics and never leave the device.

4. Seed believable fake traffic

A dashboard with three events tells you little. Seed enough visits that every chart and split row has something in it: several sources, real browser user agents, a handful of cities, some signups, some activations and some orders in more than one currency. Use real user agents on purpose: Vevee filters bot traffic out, so a made-up user agent can make seeded visits disappear. The study app seeded 90 visits from eight sources across nine cities.

import { randomUUID } from "node:crypto";
import { vevee } from "./_vevee-test";

const SOURCES = [
  { utm_source: "facebook", utm_medium: "paid_social", utm_campaign: "oct-pricing" },
  { utm_source: "newsletter", utm_medium: "email" },
  { referrer_host: "chatgpt.com" },
  { referrer_host: "www.google.com" },
  { referrer_host: "www.reddit.com" },
  {},
];
const CITIES = [["IT", "Rome"], ["IT", "Milan"], ["ES", "Madrid"], ["US", "Austin"]] as const;
const UAS = [
  "Mozilla/5.0 (iPhone; CPU iPhone OS 18_0 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/18.0 Mobile/15E148 Safari/604.1",
  "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/130.0.0.0 Safari/537.36",
];
const pick = <T,>(a: readonly T[]): T => a[Math.floor(Math.random() * a.length)]!;

for (let i = 0; i < 90; i++) {
  const visit = { visit_id: randomUUID(), landing_path: "/", ...pick(SOURCES) };
  const [country, city] = pick(CITIES);
  const userAgent = pick(UAS);
  const context = { geo: { country, city }, userAgent };

  await vevee.analytics.capture({ event: "$pageview", properties: { path: "/", ...visit }, context });
  if (Math.random() > 0.3) continue;

  await vevee.analytics.capture({ event: "signup_completed", properties: visit, context });
  const userId = "seed_" + i;
  const { visit_id: _drop, ...firstTouch } = visit;
  await vevee.analytics.identify(userId, {}, undefined, {
    isSignup: true,
    firstTouch: { started_at: new Date().toISOString(), ...firstTouch },
    context: { geo: { country }, userAgent },
  });
  await vevee.analytics.capture({ distinctId: userId, event: "signed_up" });

  if (Math.random() < 0.3) {
    await vevee.revenue.order({
      distinctId: userId,
      orderId: "seed_ord_" + i,
      amount: 999,
      currency: Math.random() < 0.5 ? "EUR" : "USD",
      billingReason: "subscription_create",
    });
  }
}

5. Cross-check the visit funnel against Traffic

Open the visit to signup funnel, split by channel, and the Traffic page for the same days. Visit counts per channel should match row by row, because both read the channel from the earliest event of each visit and both count by the app's local day. If they do not match, look at the time zone first, then at events that were sent without visit fields.

Symptoms and fixes

Issues a study app hit while testing, with the cause and the fix for each.

• Traffic showed 0 visits for today: the app time zone was unset, so data landed on the previous UTC day. Set it under Settings, Time zone.
• Revenue showed 0 after seeding EUR and USD orders: the day's exchange rate had not loaded yet. Rates load daily, and unconverted orders are counted in a "not converted" note until then.
• An activation step stayed at 0: the event was saved as its name plus a trailing dot. Removing the dot fixed it.
• Signups failed on preview deploys: the secret key was missing from the Preview environment. Add it there, or test against a production build in test mode.
• A visit funnel row looked empty: the funnel used the person filter together with a visit split, which always gives an empty table.

Test data is temporary

Test mode keeps its own people, events and orders, apart from live. Test data older than 30 days is deleted, so re-run the seed script when you need fresh charts. Test calls still count toward your workspace plan, the same as live ones, so keep seed runs modest.

Example run (mock data)

One example seed run for a study app (made-up numbers): 90 visits from eight sources, 27 signups, 9 orders in EUR and USD. With the funnel split by channel, each of the eight rows matched the Traffic page for the same day, and every seeded person had a first_channel matching their source.

Related guides and docs

Set up the integration first with "Add cookieless traffic attribution to a Next.js app", then build the funnels with "Build the two attribution funnels from JSON". Reference: /docs/guides/traffic-attribution, /docs/methods/capture, /docs/methods/identify and /docs/methods/revenue.