I want to be straightforward about what this article is.
It is not a case study. I have not run a controlled local SEO experiment on a Kathmandu business over six months and I am not going to present invented numbers as if I had. Almost every "case study" in this market is a screenshot of a rising line with no baseline, no control and no method, and those are worth nothing.
What this is: the experiment design. What to record before you touch anything, what to change, how long to wait, and what result would actually count as evidence. If you run it on your own business, you will know something that no agency report can tell you.
Why the baseline is the whole thing
Here is why nearly every local SEO claim is unfalsifiable.
Work gets done in March. In June, rankings are better. Was it the work? Nobody knows, because nobody recorded what June looks like in a normal year for that business. Kathmandu businesses have strong seasonality — trekking, festivals, the academic calendar, wedding season. A trekking company's visibility rises every September whether or not anyone touches the website.
If you have no baseline, you cannot separate your work from the calendar. So you record the baseline first, before changing anything, and you record it for long enough to see the shape.
What to record before you start
Spend one week on this. Do not skip it, and do not change anything while you do it.
From Google Business Profile, the previous 90 days: searches that showed your profile, split into direct and discovery; profile views; calls; direction requests; website clicks. Discovery searches are the number that matters most — that is people who found you without knowing your name.
From Search Console, the previous 90 days: total clicks and impressions; your top 20 queries with average position; the same filtered to queries containing a location word.
Ranking positions, recorded by hand. Pick ten queries a real customer would type. Not "best dentist Nepal" — queries like "dentist chabahil", "dentist ko fee kathmandu", "emergency dentist kathmandu". For each, record where you appear in the map pack and in the organic results.
Do this in an incognito window, on mobile, with location set to your actual area, and write down the date and time. Search results are personalised and vary by location and device; a position recorded on your logged-in laptop is not a measurement.
The business number. New customers per month, and how many mention finding you online. If you record nothing else, record this — it is the only number that pays for anything.
What to change, in one batch
There is a trade-off. Changing one thing at a time tells you what worked; changing everything at once gets results faster. For a real business, do it in one batch — you are trying to improve the business, not publish a paper.
Google Business Profile. Correct category, which is the single highest-impact field and frequently wrong. Real opening hours including festival closures. Photographs taken this year. Every service listed. Address and phone matching your website exactly.
On the website. Name, address and phone in the HTML as text, not inside an image. LocalBusiness schema with real coordinates and hours — in my audit of 34 Nepali sites, only six had any local business markup at all. One H1 per page stating what the page is. A page for each service, not one page listing all of them.
Consistency off-site. Same name, address and phone format on Facebook, directories, anywhere you are listed. Inconsistent details are a common and entirely avoidable cause of weak local visibility.
Reviews. Ask recent satisfied customers, in person, once. Do not buy reviews and do not offer incentives — both violate Google's policies and both are detectable.
How long to wait
Longer than you want to.
Google Business Profile changes can show within days. Website changes need recrawling and reassessment, typically weeks. Any change in actual customer volume needs a full cycle to distinguish from noise.
Measure again at 30, 60 and 90 days, using exactly the same method as the baseline — same queries, same device, same location, incognito. If you change how you measure, you have destroyed the comparison.
Ninety days is the minimum for a believable read. Six months is better.
What would actually count as evidence
Be strict here, because this is where people fool themselves.
Weak evidence: one query moved up three places. Traffic rose 15 per cent. Impressions increased. All of this happens by itself.
Stronger evidence: discovery searches in Google Business Profile up substantially; several of your ten tracked queries improved together; direction requests and calls up; the improvement holds at 60 and 90 days rather than spiking and fading.
The evidence that matters: more customers, and more of them saying they found you online.
And be honest about the failure case, because it is common. If nothing moves in 90 days, the conclusion may be that local search was not the constraint on this business. That is a genuinely useful finding — it means the money should go somewhere else, and knowing that is worth more than another quarter of hopeful retainer.
The confounders to write down
Keep a note of anything else that happened, because you will forget by month three:
- Seasonality — trekking season, Dashain and Tihar, exam periods, monsoon
- A competitor opening, closing, or starting to advertise
- Any paid advertising you ran
- Press, a viral post, a mention by someone with reach
- Google algorithm updates in the period
- Anything that changed in the business itself — prices, staff, a renovation
Any one of these can produce a bigger swing than your SEO work. A result you cannot separate from a confounder is not a result.
Why almost nobody does this
Because it can fail, and a proper baseline makes the failure visible.
Without a baseline, every outcome can be narrated as progress. With one, some quarters show clearly that the work did not move the business. Nobody selling a retainer wants that chart to exist.
But if you own the business, that chart is exactly what you need. It is the difference between knowing and hoping, and it costs one week of recording numbers.
If you run this on your own business, I would genuinely like to hear what happened — including if it did not work. Send me the numbers. With enough of them there would be something real to publish about local search in Kathmandu, which is more than exists today.