How to Get Traffic to a Side Project: Our Own Funnel, Week by Week (2026)

How to get traffic to a side project: expect the first visitors to arrive direct, not from Google, and measure every step after the visit so you know where people drop out. On PageSeeds, 55 of 68 unique visitors in one 28-day window arrived direct, 3 came from Google and 2 from Bing. Search took months to produce even a handful of clicks.

This post is the honest version of that start. The numbers come from the PageSeeds program ledger, the growth log we keep for the product, with readings dated 2026-07-31 to 2026-10-16. It covers where traffic came from, what search impressions turned out to be worth, and where the funnel actually broke.

The starting point: impressions, almost no clicks

The first deep dive, on 2026-07-31, found about 4,455 Google impressions in 90 days at positions 50 to 85, and 4 clicks in total, all on the homepage.

Weekly Search Console totals across the summer looked like this (complete Monday to Sunday weeks):

| Week ending | Impressions | Clicks | |-------------|-------------|--------| | 2026-05-03 | 226 | 0 | | 2026-06-07 | 293 | 1 | | 2026-07-05 | 365 | 0 | | 2026-08-02 | 337 | 0 | | 2026-08-30 | 699 | 0 | | 2026-09-13 | 460 | 1 | | 2026-09-27 | 176 | 1 |

Impressions rose to a peak of 699 in the week ending 2026-08-30 and fell back to 176 by late September. Clicks never went above 1 in a week. Do not read a rising impression line as a sign that clicks are about to follow.

Where the visitors actually came from

The growth review on 2026-10-09 split visitors by channel for the previous 28 days:

| Channel | Unique visitors (28d to 2026-10-09) | |---------|-------------------------------------| | Direct | 55 | | Google | 3 | | Bing | 2 | | Other | 8 | | Total | 68 |

The "Other" row is the remainder: 68 − 55 − 3 − 2 = 8. Direct traffic was 55 of 68, or 81%. Direct means the visit had no referrer. The ledger does not say what produced those visits, and the launch posts planned in it had not gone out yet.

A week later, on 2026-10-16, unique visitors in 28 days had risen to 77 (prior 44), and the channel tape still showed 0 to 2 Google and 0 to 1 Bing sessions a week. Search Console for the same 28 days: 2 clicks on 1,184 impressions.

So the honest answer for an early side project: search is a slow second channel. Plan for your first visitors to come from distribution you do by hand.

Why the impressions did not turn into visits

Most of the impressions were the wrong ones. The growth review found the high-impression blog pages sitting at position 58 to 94 for "content writing service" queries. Those queries do not match the CLI offer: the people typing them want a writer, not a terminal tool. In the 28 days to 2026-10-09, all 4 clicks went to the homepage.

Two lessons for your own project:

1. Check which queries the impressions come from, not just the total. A total that grows on off-topic queries is noise.
2. Write for the person who would install the thing. Posts like SEO for side projects and Claude Code SEO are written for that reader.

One query did match: the homepage ranked about 7 for "page seed" with 73 impressions and 0 clicks in the 28 days to 2026-10-16. The growth review flagged it as a brand click-through item: the page is already found, so the next test is its title and snippet.

Where the funnel broke (it was not traffic)

Traffic felt like the problem. The ledger says otherwise. In the 28 days to 2026-10-16:

| Step | Count (28d) | |------|-------------| | Unique site visitors | 77 | | Unique /install visitors | 24 (prior 10) | | Install command copied | 0 | | Checkout started | 1 (most likely an owner test) | | License minted | 0 |

Visits to the install page more than doubled, from 10 to 24, and nobody copied the install command. The growth review on 2026-10-09 put it plainly: traffic is not the binding constraint, and the funnel breaks between install and checkout.

Worked example: what more traffic would have bought

Suppose PageSeeds had doubled its 77 visitors to 154 with the same funnel. The install page share was 24 of 77, about 31%, so roughly 48 people would have reached /install. With 0 install copies out of 24, the expected number of install copies from those 48 is still 0. Doubling traffic into a step that converts nothing doubles nothing.

That is why the next two experiments were not traffic experiments. One lets an unlicensed install finish one free weekly pass before the paywall. The other replaced a hosted waitlist that had 1 signup in total with a pilot application card. Both are open bets in the ledger with check dates in November 2026.

A traffic plan for a side project, in order

1. Instrument before you promote. Track visits by channel, visits to your install or signup page, and the step right after it. PageSeeds only added an install-copy event on 2026-10-09; before that, nobody could say whether visitors to /install tried the product.
2. Send the first visitors yourself. Launch posts, communities you already take part in, direct messages to people with the problem. On PageSeeds, search sent 5 of 68 visitors in 28 days.
3. Start SEO on day one, but judge it in months. Publish pages for the queries your future users type, then measure each page about 30 days after it ships. The weekly SEO workflow describes that loop.
4. Read the queries, not the totals. If impressions come from people who would never use the product, they are not progress. Re-point or prune those pages instead of chasing their rank.
5. Fix the step that converts nothing before buying more traffic. A step at 0% makes every extra visitor worthless.

If you want to see which page types earn search clicks once a project does get found, SEO for side projects breaks down the page types, and the results page has the full snapshot.

FAQ

Where does the first traffic to a side project come from?

Mostly from direct visits: people you sent a link to, launch posts and your own sharing. On PageSeeds, 55 of 68 unique visitors in a 28-day window arrived direct, against 3 from Google and 2 from Bing.

How long until search traffic shows up for a side project?

Longer than most builders expect. PageSeeds had 4 Google clicks in the 90 days to 2026-07-31 and 2 clicks on 1,184 impressions in the 28 days to 2026-10-16. Impressions come first, clicks much later.

Are search impressions a good sign on their own?

Only if they match what you sell. Most PageSeeds impressions came from blog pages at position 58 to 94 for content-writing-service queries, which do not match a developer CLI. Those impressions never turned into visits.

What should I measure first on a side project?

Visitors by channel, visits to your install or signup page, and the step after it. PageSeeds had 24 unique /install visitors in 28 days and 0 install-command copies, which showed the break was after the install page, not before it.

PageSeeds runs this measure-then-ship loop as a weekly pass from your coding agent. To try one pass on your own project, install PageSeeds; it runs from a terminal SEO CLI in your repo.