Monday, August 03, 2009

Youcalc: On-Demand Analytics Without Stored Data

Summary: Youcalc is an on-demand analytics vendor with 130 prepackaged applications primarily for sales and marketing reporting. Unlike its competitors, youcalc it reads data directly from other Software-as-a-Service systems rather than loading it into its own database. This saves money and simplifies installation but has some drawbacks too. Still, it's an intriguing alternative to the standard approach.

Youcalc is fundamentally different from other on-demand analytics vendors like Birst, Cloud9 Analytics, Gooddata and Pivotlink: while those vendors all query data stored in their system, youcalc queries the source data directly. That is, youcalc provides analytical applications that read from an existing system, typically a Software-as-a-Service vendor like or Google AdWords.

Although this sounds like a subtle difference, the implications are huge. It means that youcalc doesn’t need the infrastructure to build and store client databases, thereby reducing its costs dramatically.

It also means that youcalc can to give each new client immediate access to standard applications, since there is no need to adjust for differences in their data. Although this is possible with prebuilt applications at other on-demand analytics vendors, youcalc has made it more central to their business model. In fact, youcalc extends this to related community concepts such as user-contributed enhancements, forums, tagging and rating of popular applications.

I’m intrigued by the youcalc approach but do see some disadvantages. One is that data integration capabilities are limited: the current version of the system can only combine data sources that already share common keys or are linked with an existing cross reference table. I suppose it’s technically possible to allow more sophisticated data matching, but any processing will still be limited by the need to repeat it each time the data is read from its sources and loaded into memory.

A second, more fundamental limitation is that the system can’t access historical data, such as point-in-time snapshots of information which is not retained in operational systems. At best, youcalc could point to an externally-built data warehouse as a source – but now you’re back doing all the database development that youcalc is supposed to avoid. No free lunch here, folks.

Still, there are those cost savings. Youcalc is priced at an astonishingly low $19.95 per user per month, which gives access to 130+ prebuilt applications for products including, SugarCRM, Google AdWords, Google Analytics, MailChimp and 37SignalsBaseCamp (project management) and Highrise (contact management). There’s also a free version that is excludes some of the more powerful applications. The full set is available for a 30 day free trial.

Unfortunately, these prices may not last. CEO Rasmus Madsen told me the company plans eventually to charge higher fees for applications linked to higher priced source systems.

None of this would matter if the youcalc applications and underlying technology weren't worth having. But I found them quite impressive.

Applications can contain multiple objects such as charts and lists. They can also contain drop-down selection boxes to filter components and select alternative chart dimensions. A single application can have multiple pages linked by menus. Users can embed images, text notes and external URLs, and have control over style details such as type fonts and background colors. Although the presentation is nowhere near as advanced as products like Tableau or TIBCO Spotfire, it is competitive with other on-demand analytics systems.

Most current youcalc applications display a single chart from a single data source, such as “Time-Day Distribution for Google Analyzer”. But users can change the contents by selecting different dimensions (e.g., date range) and metrics (e.g. visits, new visitors, bounces, etc.). Some applications combine multiple data sources, such as the “AdWords Campaign ROI Overview for” that compares cost from Google AdWords with revenue from

Users can modify these applications or create their own from scratch (although all the existing applications were built by youcalc). Development is done with Java-based desktop software that runs on Windows, Mac or Linux PCs. The interface involves dragging different components onto a whiteboard and then configuring and connecting them. There are two different whiteboards, one to show the actual application and another to display the flows used to construct each object. These flows begin with connection to an external data source and then send the data through functions to apply formulas, convert formats, create summaries, and perform other tasks. Parameters of each function can be edited during the set-up or connected to objects like drop-down menus for end-user interaction. A completed application can be saved as a stand-alone Web page, a mobile phone Web page, embedded within an external page, or deployed as a widget on an iGoogle home page.

None of this requires actual programming, and basic tasks should be easy enough for a skilled spreadsheet jockey. More demanding activities, such as connecting to an in-house data source, take considerable technical understanding. (The system doesn’t query in-house resources directly; rather, it sends a message to a “listener” on the in-house system, which runs the specified query and transmits the results back as an XML data stream.) Connections for standard sources such as are very simple since they’re prebuilt: users just enter their log-in credentials and the system does the rest.

If youcalc has an Achilles heel, it will turn out to be data volume. The system accesses standard sources (, AdWords, etc.) through their APIs, which often limit the number of records that can be pulled at once. Youcalc connectors can submit new calls until all the data has been read, but this is still awkward and will probably be slow for large volumes.

In addition, the data must be loaded into system memory during each user session. This also imposes some practical limits—we’re probably talking in the multi-gigabyte range—even though youcalc runs in the Amazon data cloud, which gives it access to very large servers. Madsen says the largest current installation works with data for 150 users.

Youcalc was launched in its current form at the end of 2008, although the company has been working on its core technologies since 2003. Madsen said that 4,000 accounts were created in the first six months since launch, and there are currently more than 7,000 application sessions per week. Most are from small businesses, which makes sense for any number of reasons including price, functionality and ease of deployment. The company hopes eventually to attract larger firms as well.


Rasmus said...

David - thanks for covering youcalc

Re. point-in-time snapshots:
We see SaaS vendors (like sfdc) storing more and more historical data, keeping logs on more fields. They face the same problem of not being able to offer historical reporting with their built-in reporting tools so they have to solve it. It is a simple storage issue and storage is getting cheaper and cheaper. Surely, having a BI vendor store data because the SaaS vendor does not do it, is a short term opportunity only.
Alternatively, we see pure database/DW on demand vendors catering for that market – why not use Amazon S3 for the purpose, or ? (SFDC could partner up wiothg Amazon an offer history storage as a premium service) – clients would have more faith in those platforms that a storage offering from an small BI vendor.

We believe analytics vendors will end up sticking to what they are good at, i.e. analyzing data, not storing it.

CEO - youcalc

Traveler said...


I remember from the days you used to write about Marketing Automation. As usual, you get to the real technical issues quickly.

There is one more issue with querying data directly from the source. If all the analytics can given based on source schema, then companies like would have given analytics to begin with. Apart from performance, real-analysis requires looking at data differently than how it is represented in the OLTP schema. This does not necessarily require physical transformation, but so far, I haven't seen a credible approach that works on the source system itself.

Rasmus said...


I would be interested in learning about your thoughts on real-time analytics on the source system and what you believe are required features.

If you are interested in sharing your thoughts, please drop me a line at ram at youcalc dot com