DataFocus · Journal
Search Tuning in OT Flash: Synonyms, Custom Keywords and Fiscal Years
OT Flash lets you teach Data Search your company's vocabulary: synonyms for column names and values, custom keywords, custom locations and a custom date such as a fiscal year. This tutorial shows where each setting lives and how to check it.
To make keyword search understand your synonyms and fiscal year in DataFocus OT Flash, use its search tuning settings. Synonyms map your own words to column names and column values, custom keywords add search terms of your own, custom locations define regions such as "East", and a custom date defines conventions such as your fiscal year. Table-specific terms are set on the table, and company-wide terms are set by an administrator under the Search tab of Configure. Tuning is usually the cheapest fix for a search that fails on company vocabulary, because you change a setting instead of retraining anything.
Why does keyword search miss company vocabulary?
OT Flash is built so that business users type column names and keywords, as they would in a search engine, and the system draws a chart. In Data Search you choose a table, type something like "Sales by Region", and the result is computed with a chart picked for it. That works when your words match the table's words.
Companies rarely talk that way. Finance says "turnover" where the column is called Revenue. Sales says "Southern region" where the data holds the value South. Planning says "FY" where the calendar has only years and months. And "East region" may be a management grouping that appears in no column at all.
If the input is not a standard keyword statement, OT Flash asks FocusGPT to parse it into a search statement, shows the parsed statement, and offers a Revert option to go back to what you typed. That fallback helps, but it adds a guess to every search. Search tuning removes the guess for the terms you use every day.
Consider a hypothetical mid-size distributor (illustrative, not a customer). Its Orders table has a column named Net Revenue and a Region column with the values North, South, East and West. Analysts type "turnover by Southern region this fiscal year". Without tuning, three of those words have no match. The sections below fix each one.
How do I add synonyms for column and value names?
There are two levels of synonyms, and they solve different problems.
Column synonyms. In a table's column settings you can set, per column, a Display Name, Data Type, Description, Column Type, Aggregation, data format, Geo Type, Synonyms, Index Type and a priority from 1 to 100, and whether the column is visible to the model. The Synonyms field takes extra names for the column, separated by commas. After you add them, typing a synonym in a search also finds the column.
For the hypothetical Orders table, the Net Revenue column would get "turnover, sales value" as synonyms. A search for "turnover by region" then resolves to the same column as "net revenue by region".
Value synonyms. On the table's Search extension tab you can add synonyms for column values. The manual's own example is a Region value of South that is given the synonym "Southern region"; after that, a search for "Southern region" also matches South. You add each one with the Add synonyms for column values button.
A practical rule for both levels:
- Collect the failed searches first. Ask three or four analysts which words they type that return nothing useful.
- Add one synonym per real phrase. Do not pre-fill every thesaurus entry; each extra synonym is one more way for a search to match something unexpected.
- Prefer the table's own Search extension for terms that only make sense in one table, and the system-wide setting for terms every table shares.
What are custom keywords and custom locations?
Custom keywords are search keywords you define yourself. They exist at two scopes: on a table's Search extension tab, where they apply to that table only, and under the administrator's Search settings, where they apply to the whole system. The manual describes them in one line and gives no worked example, so treat the scope as the main design decision. A keyword that carries one team's meaning of a term belongs on that team's table. A keyword every department agrees on belongs in the system-wide list.
Custom locations are defined by an administrator on the Search tab of Configure. The manual's example is a region such as East made up of particular states and cities. Once defined, the region can be used to narrow the scope of a search. The tab also offers an option to import system locations by country, which gives you a starting set before you add your own groupings.
For the hypothetical distributor, "East region" is a sales-management grouping. Defining it as a custom location means an analyst can search for "sales of each state in the East region", which is the same kind of question the manual uses to illustrate natural-language input, without anyone adding a helper column to the table.
How do I define a fiscal year as a custom date?
A custom date is how you tell search about your company's own date conventions. The manual's example is the fiscal year. It is set by an administrator on the same Search tab of Configure, next to custom locations and the system-wide synonyms and keywords.
The point of doing this in one place is consistency. If every analyst translates "FY" into a date range by hand, two people will eventually disagree about whether the year starts in April or July. Defining it once as a custom date moves that decision out of individual searches and into configuration that an administrator owns.
The manual states that the custom date exists and that the fiscal year is its example. It does not give a field-by-field walkthrough in the sections this post draws on, so follow the screens in your installed version's release documentation and the user manual at datafocus.ai/manuals/ot-flash/r57/en/human.html for the exact fields. Do not assume how a year that starts mid-calendar is expressed until you have seen the screen.
Search tuning or the FocusGPT knowledge base?
OT Flash has a second way to teach vocabulary. FocusGPT can be taught business definitions through My Knowledge Base, where an entry has content, keywords, a type, a category and an enabled flag, and can be imported singly or in batches. Administrators manage system-wide knowledge, FocusGPT's confirmed question-to-keyword mappings, personal knowledge and categories under Admin > Knowledge.
The two mechanisms act at different points:
| Search tuning | FocusGPT knowledge base | |
|---|---|---|
| Acts on | Keyword search itself | FocusGPT's interpretation of a question |
| Typical content | Synonyms, custom keywords, custom locations, custom date | Business definitions with keywords, type and category |
| Where | Table's Search extension tab; Search tab in Configure | My Knowledge Base; Admin > Knowledge |
| Best for | Short, exact terms: "turnover", "Southern region", "FY" | Definitions that need a sentence: what counts as an active customer |
A decision rule: if the term is a name for something that already exists as a column, a value, a region or a date range, tune search. If the term is a definition that needs explaining, put it in the knowledge base. Doing the first also helps the second, because FocusGPT's output is a search statement, and a statement built from terms that search already understands is easier to get right. The manual's own advice on FocusGPT results is to check key conclusions in Data Search, which is another reason to keep keyword search sharp.
How do I test tuned searches before announcing them?
Treat tuning like any change that other people depend on. A short checklist:
- Re-run the failed searches. Take the list you collected in the first step and type each phrase exactly as the analyst did.
- Read the SQL. Data Search shows the SQL and an analysis summary in the right rail. Confirm that "turnover" maps to the intended column and that "Southern region" filters on South, not on something broader.
- Check the negative cases. Search for a neighbouring phrase that should not change, such as "North", and confirm it still returns what it did before.
- Check the fiscal boundary. Compare a fiscal-year search against a number you already trust from a finance report for the same period. Pick a period that straddles a calendar-year change, because that is where a wrong definition shows up.
- Test with a limited role first. The menus and data a user sees depend on the roles and permissions the administrator assigned, so check at least one ordinary business-user account, not only an administrator's.
- Announce the vocabulary, not the feature. Send the team the list of terms that now work. A short list of real phrases is more useful than a description of the settings.
When is search tuning not the right fix?
Tuning helps when the words differ but the meaning is the same. It does not fix a table whose columns are ambiguous, two columns that both plausibly mean "revenue", or a metric whose definition is disputed between departments. A synonym applied to the wrong column makes the wrong answer easier to reach. Settle which column is authoritative first, then attach the synonyms.
Also keep ownership clear. System-wide synonyms and keywords affect everyone's searches, so assign them to a named administrator and keep a record of what was added and why.
Next step
Read the OT Flash user manual's chapters on tables and on the Search tab of Configure for the exact screens: datafocus.ai/manuals/ot-flash/r57/en/human.html. To install the free edition, go to the download page, where the installers and release notes are published.