Building trust is our top priority

As we’ve been building Thematic over the last few years, we’ve been asked by customers “how do you know your indexes are accurate?” As a new entrant to the space, this is a perfectly valid question, as we don’t have a long history or reputation to fall back on to answer this. We’re big believers in “showing, not telling,” and being transparent about our approach to product development so that index creators and investment product builders like you can develop institutional-quality custom products with full confidence.

By benchmarking Thematic against traditional index providers like S&P, MSCI, and Solactive, we can confidently state that our calculations are on par with the traditional providers. As you can see here, our index calculations are already as accurate as - and identical to - one of the category leaders, Solactive.

In this post, we’ll bring you behind the scenes to understand the challenges we encountered and how we solved them to achieve our current level of index calculation accuracy.

An exponentially growing challenge

On its surface, calculating an index value at a moment in time looks pretty straightforward. Once you factor in corporate actions - dividends, mergers, acquisitions, special events, and even symbol changes - ongoing index calculation gets more complicated. Then add in factors such as international listings, currency conversion, corrections to pricing data, different interpretations of corporate actions - all these factors introduce innumerable variables that create potential data quality challenges.

Then, factor in the challenges of operating at scale. For any index calculation system, the dimensions of scale are 1) the number of indices, 2) the number of instruments that can be added to indices, and 3) time (managing historical financial data as well as new financial data that is created every day). With a universe of over 78,000 securities spanning over 110 exchanges, working across all 3 dimensions, the challenge of maintaining index calculation accuracy becomes exponentially harder every single day.

Achieving accuracy at scale

In the indexing space, the good news is that every index publishes its methodology. It’s like having the teacher’s guide before the quiz, you just have to translate and understand it - and for us, create the index calculation logic in code. We leveraged existing methodologies to get started, and have built an automated comparison process to ensure our calculation accuracy stays at par or better going forward. Here’s how it works.

Step 1: Benchmarking our calculations

One way we’ve successfully done so is to run our calculation system and performance history side-by-side with other index calculations - be it from a traditional vendor like Solactive, or compared to a customer’s own index history. It is trivial for us to replicate an index on Thematic with the same list of instruments, query the prices for each instrument starting from a point in the past. For example, here’s an example of a comparison with an index calculated by Solactive: The Lux Health+Tech Index (LUXHT).

It’s important to note that the goal of the matching and comparison process is not to create a 100% match. The goal is to understand, to a very high degree of confidence, that we understand the cause of the differences.

As we run comparisons, we run a robust automated process augmented with human optimization to identify, investigate, and address the most common causes for differences in index calculations across platforms:

Corporate actions. For all corporate actions relevant to an index, we inspect the corporate actions data and effective dates, checking against a variety of sources.
Rebalancing periods. For each rebalancing, we match the timing and check the rebalance logic against the published index methodology.
Historical price differences and currency conversion. Both of these are a result of using different underlying data sources for price data and currency conversion rates. Once we’ve exhausted the potential causes in the first two buckets, the remaining differences are likely due to these factors.

At times, after running an investigation, we believe our calculations to be correct over the comparison, so our values may intentionally not always line up with others.

Step 2: Scaling up our systems and processes

What we did for LUXHT is an example of a straightforward comparison process. The key here is to run this process consistently and at scale, across a variety of published indices from the traditional providers. This is our early warning detection system for potential issues that our operations team investigates.

As a technology-first company, we very rarely rely on manual operations to detect and fix issues. Our team has deep experience in building high transaction volume systems with built in quality and reconciliation checks at Airbnb, Google, Brex, and Flexport, and we’ve brought those software and process design principles to Thematic. Our software is modular, so each type of corporate action is its own piece of logic that is easy to test, replicate, and doesn’t have a cascading effect on other pieces of logic. We’ve set up extensive alerting and monitoring systems on our data providers as well as our index calculation systems as described above. Without this, we wouldn’t have the ability to scale to hundreds of indices built on our platform in a short amount of time, covering over 48,000 instruments for nearly 20 years of historical data.

In conclusion

We believe that the systems and processes we’ve established put us on par or better than more established players in the industry when it comes to index calculation accuracy.

On top of that, our technology advantage makes it possible for us to make index calculations rapidly, giving index creators instant backtests and the ability to create an unlimited number of indices.

If you’d like to learn more, build your own custom index, or have feedback on our product, we’re all ears! Please feel free to drop us a line at hello@gothematic.com.