Try the TIMi Suite
Test the TIMi Suite for free. Click here
to get a full 3-months trial
of the TIMi Suite.
TIMi Added Value
The unique key advantages of the predictive analytic engine TIMi are:
- TIMi removes all the technological risks linked to predictive analytics projects.
Creating predictive models used to be expensive and risky. You had to hire statistians with 15 years of training in the most intricate softwares. Even if you employed highly trained statisticians, they could still fall in some dangerous pitfalls such as overfitting problems, dimentionality problems, intolerable computing time problems,... and after months of analysis of your data you get nothing (except large bills from the statistician team)... In other words, creating predictive models was a technological trip rather than a business trip.
Still companies all over the world are investing in predictive modeling technology. Why? Because it's the most profitable way to generate money out of your data. There are many ways to analyse your data (creating hyper-cubes, fancy reports, nice segmentations, creating strange Key Performance Index, performining data quality analysis,...) but none of these techniques gives you any direct Return On Investment (ROI). With predictive modeling and the "Profit Optimizer" of TIMi, you can directly see, in real-time the exact amount of Euros (or Dollars) that a predictive model will generate.
Even more than that! With TIMi, you completely remove the risk that a predictive analytical project will fail. If you have a datawarehouse, in only a few hours, any junior dataminer is able to test, with TIMi, if your data contain some potential to generate some good return. Based on this preliminary analysis, you can decide to proceed. With TIMi, there is no risks anymore: no more technological difficulties, no more versatile and expensive staff to recruit, no more uncertainty about the quality of your data.
- TIMi has a very low cost of ownership.
There is no need to hire expensive staffs anymore to do predictive analytics: Any junior informatician or junior statistician is able, after a two days training, to use TIMi. Since you are now able to produce yourself your own predictive analyses, you stay fully in charge of your own strategy and let nobody else take "clever decisions" for you. With TIMi, you "stay in control".
- TIMi makes you independent
You are not anymore dependant on one unique team member because everybody in your team can easily use TIMi.
- TIMi is user-friendly.
TIMi has 3 different user-interfaces, adapted to the degree of technicity of the user.
This is unique. These 3 different user-interfaces are:
- Wizard-based interface: In a few mouse-clicks, you can create all common predictive models (propensity-to buy, cross-selling, upselling, churn, etc.). To create some un-common "special models" (such as "Predictive models for Strategic & Tactic decisions" or "Time-Series models"), you need to use the Advanced interface.
- Advanced interface: this interface allows you to have more detailed control on common modelization parameters: you can filter-out some rows of your database, remove some columns, create "on-the-fly" new targets, etc.
The Advanced Interface is 100% safe to play with: In the "Advanced interface" you cannot deteriorate the accuracy of the predictive model by changing some low-level-algorithmic parameters that nobody typically understand.
- Expert interface: This interface give you access to all the low-level parameters of all the algorithms that are inside TIMi. When you play with these parameters, you can sometime slightly increase the accuracy of your predictive models. TIMi is not a black-box: All algorithms and all algorithmic parameters are accessible and documented.
- TIMi ensures high quality results.
- The predictive models built with TIMi are very accurate: Thanks to a unique technological and mathematical approach, TIMi guarantees you to obtain the best model with the highest ROI and the highest accuracy on any given real-world dataset (the lift is high, there is no overfitting, the model is robust). This is unique.
- With TIMi, the deliverables are always of high quality. For example: the reports are nice, printable, and actionable. TIMi automagically produces user-friendly colorful reports in MSWord, MSExcel and dynamic HTML. Inside the MSWord reports, you will find a "dahsboard" that give you a short, concise and precise description of your target. This intuitive "dahsboard" is unique.
- TIMi is insensible to errors inside your data. No more expensive data-cleaning required!
Y ou don't need to lose several months cleaning your data before using TIMi. If you want, you can use directly TIMi on the database containing your operational data to directly obtain highly accurate, directly actionable predictive models. This unique feature of TIMi allows you to start creating new predictive models instantaneously and at very low cost (No more expensive data-cleaning required!)
- TIMi is able to connect to a vast variety of databases.
TIMi can directly connect to a vast variety of databases through its standard OleDB or ODBC connector. There is no need to "extract" your database into giant cumbersome and annoying flat file anymore! All the databases that are supporting the ODBC or the OleDB technique are accessible to TIMi.
- TIMi reads natively datasets stored in compressed text files.
You don't need to decompress your dataset-text files anymore! TIMi supports .ZIP, .RAR, .GZ, and .LZO compression. This unique ability of TIMi allows you to manipulate easily what's used to be large, cumbersome dataset files. This is important when working with really big datasets. For example: you can easily analyze with TIMi a text-file dataset of 5 GB (that had the un-manageable size of 50 GB before compression!)
- TIMi creates predictive models that any business-user (even without any training) can understand.
- Predictive models created with TIMi are incredibly simple and yet extremely accurate.
Understanding how your predictive models compute their prediction is absolutely VITAL. You can never use "blindly" your predictive model: it's the open door to major error that could jeopardize all your marketing campaigns. You NEED to understand what's your model is doing. Only TIMi is giving you the required tools and reports to properly understand how your predictive models are working. These reports are also instantaneously giving some very precious "insights" about your business. This is unique.
- Predictive models created with TIMi are NO "Black-Box". The ability to obtain predictive models that are, at the same time, very accurate (more accurate than any other solution) and very easy to understand is unique.
- Predictive models created with TIMi are incredibly simple and yet extremely accurate.
- TIMi ensures an easy and fully automated industrialization of your analytical solutions.
You can use in 2 mouses clicks all your predictive models and apply them on your whole customer database. For still easier industrialization of all your predictive models, you can export your predictive models to standard SQL code (Oracle, MySQL, SQL server, etc.), to
Teradata-Optimized SQL code, to SAS code, to PMML code, to C++ code, to dynamic HTML pages. You can also use all your predictive models directly inside any data-transformation-graph inside the Anatella-ETL. The same breadth of functionality is nowhere available (or it costs you a lot!).
- The TIMi scoring engine is FAST. The scoring engine of TIMi is between 50 and 100 times faster than the best scoring engine available in other tool. Other tools are so slow when scoring a database that the users are forced to "simplify" their predictive models by removing some important variable out of their model (this greatly reduces the accuracy of the predictive model and thus your ROI) to be able to apply their predictive model in a reasonable time (i.e. less than 24h of continuous computation during which their operational database "freezes"). The TIMi scoring engine is unique because of its unmatched speed.
- When you re-apply your predictive model on an updated version of your customer database, TIMi performs several checks:
- If your predictive model is outdated, an alarm is generated. This is unique.
- One very common and recurrent problem that nearly always arises, sooner or later, when applying a predictive model on any dataset is that the IT team actually did a mistake during the "update" of the dataset and inverted the name of 2 columns of the dataset (or some similar error). In such situation, all the other datamining softwares (other than TIMi) will produce completely erroneous predictions without giving you any warning or any messages that would allow you to detect the error. In opposition, each time you apply a predictive model built with TIMi, TIMi check the correctness of the dataset and generate an error message if the dataset is "odd" (for example: 2 column's names are inverted). This very important functionality is unique to TIMi. There exists no other datamining tool that proposes such an ultra-important, crucial functionality. Without this functionality, you can never be sure that you updated correctly your predictions/ranking. Only because of this functionality, I would personally never use any other datamining tool than TIMi. This is a unique functionality.
- You can use in 2 mouses clicks all your predictive models and apply them on your whole customer database. For still easier industrialization of all your predictive models, you can export your predictive models to standard SQL code (Oracle, MySQL, SQL server, etc.), to
- TIMi ensures consistent quality.
- Since TIMi is 100% automated, there is no risk that an inattentive (or unmotivated) employee produces poor accuracy models. TIMi ensures that all your predictive models stay accurate, always.
- With a few other datamining tool, you can also easily industrialize your predictive models: i.e. you can regularly "apply" your predictive models at each updates of your customer database.
...but TIMi goes one step further: With TIMi, you can automagically re-create completely new predictive models based on the latest updates of your customer database. Why continue to use an old, outdated (and now inaccurate because it's obsolete) predictive model? With TIMi, all the parameters required to update your predictive models are saved and you can use them to re-create automagically new, updated predictive models.
Other solutions are not able to create automatically new predictive models and you very often ends-up using obsolete &; inaccurate models. This is unique.
- TIMi still offers you some more unique technical abilities!
Those unique abilities are: datasets of any size, confidence intervals directly on the lift to know the stability of your model, unmatched speed, realtime campaign optimizer, customer profile directly actionable from a business perspective, close integration between segmentation analysis and prediction analysis, ensemble modeling, full Unicode support (i.e. the ability to process databases that contain, inside the same table, Cyrillic characters, Greek characters, Chinese Ideograms, etc), ...
- TIMi increases your productivity.
When using TIMi, the time required to complete one full datamining project is usually divided by two compared to other classical datamining softwares. Or, in other words, you can do twice more work in a given time when using TIMi! To know how TIMi is able to dramatically increase your productivity, read this paper. This is unique.
- TIMi gives you the "shortest time to market".
No other predictive datamining tool is able to produce and industrialize an analytical solution as fast as TIMi. This is unique.
- TIMi is currently the only datamining tool that is really scalable and accurate.
No other tools is able to process the vast amount of information currently available inside enterprise-level-databases.
More precisely: The whole idea of using a larger quantity of data is to obtain better, more accurate predictive models. This purpose is defeated by other modeling tools that are forced to "switch to" simplified or approximate (but fast) algorithms in order to be able to process the vast amount of data in a reasonable amount of time. In opposition, TIMi is able to "digest" very large learning dataset (with millions of rows) and still use no approximation. This means that, when you feed more data to TIMi, it's really able to improve the delivered models (in opposition to other tools).
TIMi is the only Analytical Tool that is scalable to the size of commonly available enterprise-level databases. This is unique.
- TIMi allows easy collaboration inside your team of dataminers.
You can store all your datasets on a central remote computer that is shared by all your teams of dataminers and still have a minimal stress on the network of your company. This is possible because TIMi has a special I/O access pattern that minimizes I/O access when accessing the datasets to analyze. This is especially true when using compressed dataset formats such as the .RAR or .ZIP compressed text files.
It's important that all the dataminers of your company share the same datasets because, otherwise, you can obtain very serious inconsistencies between different analyses performed by different dataminers on different datasets. This can lead to very bad contradictory strategic decisions.
- TIMi is 100% integrated into SAS.
TIMi is fully integrated with SAS. It is able to read natively SAS dataset files, SAS variable labels and permanent SAS variables formats. You can directly run TIMi from the SAS environment, to create new models, to compute scoring, etc. The generated predictive models can also be exported in native SAS base code. TIMi can be seen as a true plug-in inside the SAS system.
Some of our customers are attached to SAS for legacy reasons. The integration of TIMi inside SAS still offer them the possibility to use the highly efficient modeling process from TIMi inside their SAS environment without modifying anything inside their current SAS-based operating processes.
- TIMi is using the powerfull Teradata engine to open the road to a "ubiquitous predictive analytics".
As already explained above, you can automatically export, in one mouse-click, all TIMi predictive models to Teradata-Optimized SQL scripts to compute all your scorings directly "in-database".
With classical modeling tools, the scoring procedure used to be a source of unreliability in Business Intelligence Solutions and it was certainly a big bottleneck. In opposition, with TIMi and Teradata, the scoring computations are entirely performed "in-database" by the highly optimized & parallelized Teradata engine. This allows us to reach very high speed and very low consumption of server resources.
When classical modeling solutions push a heavy load on you BI system, the combination of Teradata and TIMi is so light and agile that you can deploy predictive analytics quickly and easily for a large number of different problems. This opens the road to a "ubiquitous predictive analytics" in companies equipped with Teradata and TIMi.
Tests operated by a large telecom operator in France demonstrated that TIMi Models exported as Teradata-Optimized-SQL-Scripts are able to score about 60 million individuals in a few minutes. This is several orders of magnitude faster than any other analytical solution. This offers new application fields for the usage of predictive modeling. Predictive analytics is no longer restricted to a few highly-strategic projects and can be used on a wide variety of problems (especially since building a TIMi model takes only a few minutes and a few mouse-clicks).
- TIMi does not force you to do "bad things" that are reducing substantially the accuracy of your predictive models.
WARNING: This section is rather technical: You need some minimum background in predictive modeling to understand most of the concepts exposed here.
All other Predictive Analytic Tools are forcing you to do the following horrible things:
- Reduce the number of rows of your learning datasets (sampling).
This reduces the accuracy of your predictive model. If you don't do it, other Predictive Analytic Tools simply refuse to start, compute for several weeks or simply crash.
- Manually reduce the number of columns of their datasets using some "intuition".
If you don't do it other Predictive Analytic Tools simply refuse to start or simply crash.
It's very common to use some business sense to eliminate "a-priori" a large number of your variables. This is not a good idea. You should let TIMi decide if a variable must be removed or not. An another "bad habit" is to delete variables with many missings. Very often the target size is a few percent, in this situation, variables with 90% missings are perfectly legitimate.
- over-clean their learning datasets (to be sure that no outliers remains at all) removing so much information out of the dataset that the accuracy of the predictive model becomes very poor. Most "cleaning procedures" are removing precious information out of the dataset and are in reality decreasing the accuracy of your predictive models. But, if you don't clean your data, other Predictive Analytic Tools deliver predictive models with really poor accuracy.
- loose time by manually computing transformation of the original variables If you don't do it, other Predictive Analytic Tools deliver a predictive model with really poor accuracy.
- Reduce the number of variables inside your predictive model so that its possible to use it in production (because, otherwise, "it will be too slow")(this is VERY COMMON). Aaaaaargh! No comments!
- Spend countless hours tuning the hundreds of incomprehensible parameters available inside the dozens of inefficient different modelization algorithms (trees, neural nets, logistic regression...). If you don't do it, other Predictive Analytic Tools deliver predictive models with really poor accuracy.
- Spend countless hours trying to find at least one algorithm that works (even a little!). If you don't do it, other Predictive Analytic Tools deliver a predictive model with really poor accuracy.
- Drink too much coffee anxiously waiting for the analytical CRM tool to finish its computation and
suddenly decide to abort prematurely the computation because it "takes too much time" (and maybe losing a very good model in the process).
Other modeling tools are extremely slow compared to TIMi. We noticed on many different real-life industrial projets that TIMi&Anatella are providing a significant advantage in terms of modeling accuracy and in terms of running-time-speed over other off-the-shelves solutions.
All these concerns no longer exist when working with TIMi.
- Reduce the number of rows of your learning datasets (sampling).
Next: Anatella unique added Value compared to other ETL tools
Test the TIMi Suite for Free!