Multi-Site Operations: A Routine for Running Many Client Websites
Running fifty client websites does not take fifty times the effort of running one. It takes a routine, and most of that routine can run while you sleep.
This is the routine, split into what runs every night, what you look at once a week, and what you send once a month. It also says plainly which work will never be automated.
What multi-site operations means for an agency
Multi-site operations is the work of keeping many client websites healthy after they launch. Building is the fun part. Operations is the part that never stops, and it is where a small shop either stays small or gets its evenings back.
The good news is that the work splits cleanly. Some of it grows with every site you add. Some of it does not grow at all once you set it up.
Five checks that should run every night without you
These are the checks that pay for themselves. Write each one as a script, because a check you have to remember is not a check.
- Rebuild the sitemap. Every site gets a fresh list of its own pages, so a new page is never left out.
- Read every page back. Fetch each page the way a visitor would and look at what came back, not only the status code. A page can load normally and still be wrong.
- Find broken internal links. Catch the link that points at a page that moved, or at a previous client's site.
- Send a real test message through each form. A form can return a normal page and still reject every message. We found exactly that while building the BellyUp! sites, and a status code never would have caught it.
- Fail loudly. If anything breaks, the job writes it to a log, keeps checking the rest, and finishes with an error. One early failure should never hide behind a later success.
The 160 domains case study shows this job running across a whole portfolio. The errors worth catching turned out to be the quiet ones, pages that returned normally while being wrong.
The 30 minute weekly look at your client websites
Once the nightly job exists, your week gets short. You are not hunting for problems. You are reading the list the job already made.
- Read the failure log. Open the nightly log and fix what it flagged.
- Check Search Console for drops. Look for any client site whose clicks or impressions fell. It is free and it covers every site you add.
- Carry one fix back to every site. If you fixed something on one site this week, put the same fix on every site that has the problem. A fix is not done when it works on one site.
- Answer the client requests that piled up. This is the part software cannot do for you.
The monthly client report
The monthly report is what justifies the monthly fee. It should answer three questions in plain words: what happened, what it means, and what you are doing next.
Keep crawler visits and human visits apart when you report. A bot reading a page is not a person reading it, and blending them makes a number that looks better and means less. The FAQ on what to send a client every month goes through what to include.
How much time multi-site operations takes at 10, 25 and 50 sites
Here is a pretend shop, Pine Road Web, to show the shape. The minutes are made up for the example, not measured.
Suppose a site checked by hand takes 45 minutes of attention a month. Now suppose the same site, with the nightly checks in place, takes 10 minutes because you only look at what the job flagged.
| Client sites | Checked by hand, 45 minutes each | Nightly checks in place, 10 minutes each | Hours saved a month |
|---|---|---|---|
| 10 | 7.5 hours | 1.7 hours | 5.8 hours |
| 25 | 18.8 hours | 4.2 hours | 14.6 hours |
| 50 | 37.5 hours | 8.3 hours | 29.2 hours |
At 10 sites you can live with doing it by hand. At 50, doing it by hand is a second job. The checks cost the same to write whether you have 10 sites or 50, which is why they matter more the bigger you get.
The work that stays human
None of this removes the work that was always the real work. Writing, judgment and client relationships cost the same per site at 160 sites as at 6.
Software cannot tell you whether a page was worth writing. It cannot tell you whether a restaurant on a client's site is still open. In the BellyUp! build, that check still depends on someone reading the newest dated posts, and it costs the same in city thirteen as in city one.
Where to start if you run many client sites by hand today
- Pick one build and repeat it. Checks only work across sites that are built the same way. Twenty sites built twenty ways cannot be checked together.
- Write the first check. Start with broken internal links. It is the easiest to write and the one clients notice first.
- Make it fail loudly. Send the result somewhere you will see it before a client does.
- Add one more check each month. After a year you have twelve, and none of them cost you an evening.
Related reading: the twentieth client site problem explains why the work grows faster than the income, managing many client websites without a growing bill covers the cost side, and the multi-site management glossary entry has the definition.