One of my primary goals at work is to improve the search function in our knowledge base. There are a handful of tools in ServiceNow to make this happen, but one of the most powerful is using meta.
However, I haven’t found any documentation on how to really optimize your KB. It’s all “here’s how you add meta to an article,” or “here’s how to improve a single article’s rankings“. Alternatively, I’ll read about how the Zing search works to derive words with stemming, and feel like a total dummy for wondering how to apply it. It shouldn’t take a post-grad degree and fifteen hours to answer the question, “Does this mean I should be leaving off -ing, -s, -ed, etc., from root words when I add them to meta?”
In short, it’s all high-level developer stuff. Great for them, but I, your average IT Jane, still have a lot of questions. Or, as George’s Dad might say:
I got a lot of questions for you people, and now you’re gonna hear about it!
But anyway, I’ve had to tinker with things on my own. For a while, I was just jumping into an article and writing meta willy-nilly. I wasn’t sure how to build relationships between similar articles, though it was on my mind. I looked at our the search logs for potential meta, but I wasn’t consistent in how I applied them. I think it showed in the search results, which jumped up and down without much sense.
This week, I think I found a better way to build relationships between knowledge articles and improve our SEO, and I’m excited to share it today.
So, how can you logically and methodically use meta to create connections between your articles? My solution is to create meta maps to improve SEO, and building that meta through search log analysis.
How to build a great meta map for your KB articles
- Create spreadsheet for all the articles you want to review. These articles should be related in some important way–same service offering, similar topics.Here’s a snippet of what mine looked during a recent KB review. I’ve blacked it out, but the KB numbers run across the top. The row beneath is the name of the article. This forms the columns under which I add meta phrases into individual rows.
- Identify 1-2 keywords you want to use for the service offering as a whole. Add them to each article.For example, if you’re reviewing articles about multi-factor authentication, you may use “mfa” as one of your keywords for all related articles.
- Review each individual article for potential meta. Add them with abandon.
- What makes potential meta? Headings, and key terms or phrases are a good start.For an article about resetting MFA hardware tokens, you’d obviously use the meta “hardware token” or “reset hardware token”.
- Review the search log:
- You’re looking for common search terms. How you define “common” will depend on the robustness of your knowledge base. For a KB with less traffic, you may want to use search terms that show up at least 3-5 times.
- You’re also looking for potential synonyms to add to the synonym dictionary.In our MFA example, you’d want to set “mfa”, “multi-factor”, “two-factor”, and “2fa” as synonyms. If you’re using a particular company to provide MFA, like Duo or Protectimus, add those as well.
- You’re also also looking for odd duck search terms. If you’re looking for terms about resetting hardware tokens and you see “locked out mfa” multiple times, that should draw your attention. It might not be relevant to the article in question, but ask yourself, do you have an article in the KB that this search applies to? If not, it could warrant a new article entirely..
- Review any other search analysis sources you have, like Google Analytics. Perform the same kind of analysis as in step #4.
- Review your overflowing meta. Trim duplicates and lengthy phrases, check for stop words, stems, etc.
- Take your meta and stack it three times in the article. You don’t need to retype! In a Google Sheet, concatenate your rows into a single comma-separated cell that you can copy and paste into the meta field using the formula (where ##:## is the cells you want to join:
And voilà, now you have your own meta map!
One thing to note… During step 4 and 5, when you’re doing your search analysis, your focus is likely to spiral towards other articles, because it’s hard to look at the search log without noticing a dozen other potentially useful things. So keep in mind that when you reach step 4 and are focusing on a specific article, you may find other great meta for different articles. That’s a good thing! Add them to the relevant articles, but come back to the one you were working on.
Potential benefits of a meta map
- This allowed me to have a higher-level view of the articles in a single service offering, which I couldn’t get by jumping into individual articles before.
- I can use the find feature to track repeated meta across articles. I can eliminate or boost this to provide more articles to an individual search.
- Maps can be shared with coworkers for their contributions and knowledge.
- Systematic approach to writing meta, rather than the scattered approach I was using.
- Can compare this to future versions for analysis.
Potential limitations of a meta map
- Digging through the search log may lead you toward confirmation bias about search terms. Here’s what I mean. Our SN instance gets thousands of searches a day, and I can’t look through every search. Instead, I have to take a guess about the top two or three broad-level searches that most people will likely be doing. So I will search for logs containing the term “mfa” and create meta from them. But what if people aren’t searching for “mfa”, but using the product name? I’ll miss those searches.
- It’s hard to hold a lot of articles in your head at once to evaluate which terms should go where.
- Somewhat unwieldy spreadsheets, depending on how many articles you’re reviewing at once.
- Not all-comprehensive. Take this situation for example: In step 4 and 5, I notice there are a number of searches about “recording meetings.” Great, I think, I can add this meta to the article we have on playing and recording conference calls with WebEx. But is that what the customer meant? There are other services that allow customers to record audio, and more than one web-conferencing tool with recording abilities. The original search term wasn’t “recording meetings webex” so I can’t be sure this is the article that would solve their problem.
Well, that’s it! I hope that helps. If you have any suggestions for improving my process, I’d love to hear it. As you can tell, it’s a complex topic and Google ain’t sharing.
Photo by Sam Erwin on Unsplash