On February 14th, Google G Suite Business and Enterprise edition customers will begin to have access to Google Cloud Search - the first step in Google’s journey to replace the Google Search Appliance. You can read the official announcement here: https://blog.google/products/g-suite/introducing-google-cloud-search-g-suite/
SADA Systems has activated Google Cloud Search on our domain, so I would like to show you some details of what is included in the first release of Google Cloud Search. I have run a series of tests to show you how it works on our own content.
Homepage
To access Google Cloud Search (https://cloudsearch.google.com), you must be signed in as a member of a G Suite organization with Google Cloud Search activated. When you load the Google Cloud Search page, you are greeted by some helpful and relevant “assist cards”, even before running a search. My home page included a ‘Pick up where you left off’ card showing a few documents recently edited by me or by others. There is a short description to help you understand why certain documents are being shown. I have seen messages such as “You’ve edited this in the past week”, “Updated in the past day since your last edit” and “You’ve viewed this in the past day.”
Clicking “Show More” expands to show me a total of 8 “Pick up where you left off” documents.
Next, I was shown either a “Prepare for an upcoming meeting” or “Review a previous meeting” assist card, depending on the nearest event on my calendar. I noticed that like many features of Google Cloud Search, this assist card updated instantly (faster than I could change tabs and hit refresh) when something on my calendar changed. The card showed some basic information about the meeting, including the title, time, attendees and invitation statuses. Documents related to the meeting were also shown to help you prepare or review. Any documents specifically listed in the meeting description or metadata appeared, but I was also expecting it to show documents based on relevance to words in the meeting title or description. I could not get this to happen in my testing, despite trying several different titles and descriptions.
Clicking “Show More” opens an expanded view of the meeting’s metadata. Along the right side of the page is a timeline of my entire day’s meetings, and clicking on another meeting shows its detailed view and related documents.
The assist cards are very nice, but at the moment, it appears to require you to keep the Google Cloud Search homepage open and remember to check it periodically. I could see the cards becoming out of sight - out of mind for people that don’t run searches frequently. I suspect push-style notifications will come in the future.
One final detail on the homepage is a wonderful FAVICON for your browser tab. Conveying information in 32x32 pixels is hard. I like what Google came up with:
Search Box
Searching for content in Google Cloud Search is simple and fast. As with any Google search box, you simply type some keywords and hit enter to see nearly instantaneous results. I have been running searches all day and none have take more than 1 second to execute.
After typing as little as one letter in the search box, an autocomplete dropdown appears. The dropdown shows exact matches and profile pictures for people in the organization, as well as a list of suggested search terms:
After turning on Web & App activity tracking, the suggestions also included terms that I had searched for (since the point I turned on activity tracking). Otherwise, the suggestions appear to come from accessible content in my organization. For example, I searched for the number “8” and was shown a suggestion of “ticket #8686831”, which came from part of the subject of an email I had received. Searching for the numbers 6, 7 or 9 did not return any other ticket number suggestions. So, the suggestions appear to be limited to only my available content. I did notice some very common words (i.e. “administrator” or “documentation”) did not appear in the suggestion list, even though they appear frequently in my available content.
Search Results
Google Cloud Search defaults to searching all available content sources and shows the results in a heterogenous list. There are no controls for sorting the results (which appear to be in relevance order, as opposed to date order), and it does not show you the total number of results. There are simple next / previous controls at the bottom of the page, and you can only see 10 results per page. I am impressed with Google’s search result blending algorithm. I always seem to get a nice balance of different types of content on the first page of the results. That is not an easy feature to get right, but Google has done a great job.
There tabs to limit which sources are searched. For example, only Drive documents or only Calendar events. There are also filters to limit your search by date and by who created it (you vs. everyone).
Finally, there is a file type filter with a list of options. The list is not dynamic based on your current search results, though, and will allow you to select a file type with no available results.
Google has provided some convenient search operators that you can type directly in the search box (in addition to the standard AND, OR and NOT). These include source:, filetype:, owner:, from:, to:, before: and after:. Missing is a way to limit the query to just the title of a document, such as filename: or title:.
The search results themselves are color-coded and show different information for different types of content. Here are a few examples:
Docs:
Calendar:
Email:
Groups:
Images:
Videos:
The results are visually lacking, though. There are no thumbnails or previews, even for images or videos, making it very difficult to locate that type of media. Metadata and content summaries are shown below the results, but only a few sentences worth (dynamically selected based on your specific query). The descriptions are a mashup of content and metadata with no formatting. You cannot view any more content or metadata without opening the full document.
People Results
If your query contains the exact name of a person (first and last), a People card is shown to the right of the search results. It contains shortcuts for starting a Hangout or composing an email, as well as some basic metadata for the person, including title, email address and a clickable list of direct reports. I have only seen telephone numbers and physical addresses for some users, which is probably because of missing data in our directory.
People Card:
Show More (direct reports):
Indexing Speed and Change Detection
In most cases, I have been extremely impressed with how fast Google Cloud Search picks up new items or changes to existing items. If I change the name of a document in Google Drive, the new name is searchable and displayed in the search results within 1 or 2 seconds. The same is true for Calendar meetings and Gmail messages.
However, I have noticed that recent changes to the contents of a Google Doc were not searchable immediately after editing the document. I waited more than half an hour with no luck. I finally decided to try renaming the document. That change, and the recently edits, were now searchable just 1 second later. There seems to be a longer delay for Google Cloud Search to detect certain types of changes. Note: the Google Drive search within Google Cloud Search seems to be exactly the same as the standalone search within Google Drive - in that if Google Cloud Search isn’t finding something, the standalone Google Drive search doesn’t either, and vice versa.
Optical Character Recognition
One of the more interesting features for me is the additional processing and analysis that Google Cloud Search can apply to incoming content. Images with text, for example, are processed with Optical Character Recognition technology, and the results are added to the metadata of the image. The text is both searchable and displayed in the result listing.
Original Image:
Search Result:
I continue to be extremely impressed with Google’s OCR capability. It recognized all of the text on this textured, non-white background, with multiple fonts. Very impressive.
This feature seems to share some legacy with Google Drive. I have a suspicion that it is because Google Drive is performing OCR on the documents that Google Cloud Search can search the text. I have not seen evidence of OCR being performed on non-Google Drive content. So, we will have to wait and see if this feature has any benefit for third-party documents in the future.
I did see some lengthy delays in getting Google Cloud Search / Google Drive to perform the OCR. I got it to work on one PNG image within about 1 minute, and then several other attempts failed to produce any searchable text. I tried some PDFs (scanned paper documents using an app on my iPhone) and the OCR was not available until the following morning. Once the OCR was performed, it was superb, capturing all of the text on the page. I read somewhere that Google Drive OCRs the first 10 pages of PDFs, but I haven’t officially confirmed that.
In addition to OCR, we have been hoping for other visual and audio processing, such as the recognition of objects within a picture (i.e. cake, horse, grass, stop sign, etc.) using the Google Cloud Vision API, or transcription of audio files using the Google Cloud Speech API. I hope these features are on the TODO list.
Conclusion
In general, there is a lot to be happy for in the first release of Google Cloud Search. It appears to be very fast and accurate in my initial tests. The assist cards are clever, and unified search across all the different G Suite sources is extremely useful.
I am not sure how the searching and ranking algorithms compare to the Google Search Appliance, though. How well will it search through 20 million documents and sort them to show you the best matches? I plan to run some more tests on larger sets of documents, comparing the results side by side with the GSA.
I have also not seen the signs of the machine learning and analytics, yet. It might be there under the hood, but it’s impossible to tell. Is it giving more weight to documents that are viewed more often or more recently? Is it analyzing what people search for and what they click on? I have no idea. I understand that it is “Google Magic”, but I ultimately have to sell this tool to companies that will want to know a little bit more about what’s going on under the hood. I hope that some visual feedback is shown, similar to the indications under the “Pick up where you left off” recommendation.
Finally, we will have to wait a bit longer for the ability to index custom content sources and third-party data. Once available, that will open up countless possibilities using GSA-like connectors, as well as direct integrations negotiated by Google (like the Box integration previously announced).
For now, Google Cloud Search provides some very nice advantages over the individual searches present in each of Google’s apps, but it’s not enough to be full-blown enterprise search, yet. But it is a great start, and I know those additional features are on the way.