Case Study: Simply Wall St Dashboard Feed

Case Study: Simply Wall St Dashboard Feed

This is a short overview of the “Dashboard Feed” project I led in my role as Content Product Lead at Simply Wall St, a fintech startup with circa 30 employees (at the time). This project was ultimately highly successful – far more than I’d anticipated at the time – so I want to look at:

• Why we built the dashboard feed
• How we determined what to build & tested this with users
• How I positioned the team to deliver
• The end result

TLDR: We built a dashboard feed that increased activation and retention significantly, and as 2024, the feed remains the largest single predictor of paid user retention.

(As with other case studies, I will omit commercially sensitive information. Thoughts and feedback welcome).

Simply Wall St’s primary acquisition tool for a long time was algorithmic/automated news articles distributed through a variety of channels. They contained insights that you simply couldn’t get on the platform and for a long time there was a slightly discordant experience where if someone enjoyed the analysis in an article, they would then click through to the platform and not find any of those insights.

Additionally, users would be dropped into a complex app (see below screenshot) with minimal guidance/direction. The result of this was that out of several million visitors, only around 5% became active users of the platform (measured by visiting once in a 14-day period) and a tiny fraction of that converted to paid. At the time, customer retention was also low and while the business enjoyed SaaS-like economics, it had retention closer to a newsletter product than a long-customer-lifetime B2B SaaS.

“News” – coming back to get new information – has always been a huge part of the SWS ethos and the CEO, Al, had determined that the way to improve both retention (of existing paid & free users) and activation (of new free users) would be a dashboard for users to track updates about their portfolio. It was an elegant solution as it would simultaneously solve several problems:

• Give existing users a reason to come back to the platform (“maybe there is some news on XYZ stock I hold”)
• Give new customers a clearer “hook” to engage with (“I can see the history of events at XYZ stock”)
• Allow us to share article/news insights in a more natural way
• Expand the existing Portfolio feature as currently it just showed dollar values

Quan Do (designer) and Mauro Ferrante (PM) led the dashboard product team designing the dashboard page itself. The current version of the dashboard looks like this, but the early one was very similar:

(Mock portfolio)

It became obvious in about five minutes that simply putting links to articles into the dashboard feed would not be ideal, as these were optimised for volume and also for attention-grabbing headlines like “Here’s why analysts are upgrading XYZ stock”. This meant that customers were getting many duplicate articles, and also the headlines “forced” people to click through to the article and read it (taking you to another website) rather than getting the information in the dashboard itself.

The headlines were also optimised for intrigue rather than communicating clear information. We considered making the articles native, which was not technically difficult – simply exporting already-stored text into another place on the site – but the articles would still have all the same issues of not communicating clear information, and it was far too much effort to recreate them in an ideal format. Likewise changing all of the headlines was not really appealing. (It would have meant creating and maintaining two types of articles, one for internal users and one for external readers/prospective customers – months or years of work which could not be spent improving the articles or improving other aspects of the experience).

However, the core analysis of the articles (code sorting companies into categories/outcomes) was very easy to recreate and could be done in days. We decided to create bespoke updates tailored for easy consumption in the dashboard format.

What do customers want?

So what should we update people on? What do customers care about? There are several answers to this, (and it is a lot more difficult than you would expect) but there is one overarching principle you must keep in mind:

Principle #1: People only care about whether the stock goes up or down and why.

Yeah they might look at the balance sheet, or new products, or revenue growth, or what Jim Cramer says, or [blah blah blah] but at the end of the day, everything boils down to “is this investment going to go up or down”, or, in limited cases, “will it keep paying dividends”.

For a subset of more sophisticated and enlightened customers, they might take a more longer-term view of “well I don’t mind if it goes up or down in the short term as long as it goes up more in the long term.” Everything we needs to relate back to that.

Principle #2: People don’t know all of the things.

Simply Wall St sells tools to investors. The amount of time people have for research is finite, and the number of things that could become relevant is almost infinite. As a result we cannot simply “build things that people like” because there are many more things that could be built that they don’t know about. A diagram:

Consequently, there is a big push-pull dynamic between:

• Most data is irrelevant most of the time, but it becomes acutely relevant at certain times, and
• There is more involved in investing than any typical user could possibly know or keep track of
• People are learning how to invest, so you must help decide what to draw their attention to

For example, if a company director retires, this happens all the time – directors are usually old – and doesn’t really have any predictive information. If, however, five directors leave within a short period, well that’s a yellow flag and something any investor should look at before deciding whether to buy or hold the company’s stock.

So, there is judgement in what to show people and what to build. Building things and then relying on metrics/engagement also doesn’t work because there is a) too much data to do this, and b) you would deceive yourself. Probably what you would conclude is that “people aren’t interested in directors changing” when the reality is closer to “99% of the time it is not interesting but 1% of the time it is extremely interesting”. Understanding this helps you make a more informed decision about what exactly to notify people about.

Deciding what to build

Keeping these two principles in mind, this is how we decided what should go into the dashboard feed first. We used a combination of Jobs To Be Done (“JTBD” – more on this below), data available, reader interest, and judgement. In diagram form again:

While choosing JTBD is relatively straightforward, it’s also limited by the data we had from our provider. For example, at the time, data was provided on a lagged basis, so real-time updates were out. Similarly, we might have some interesting data (for example – auditor going concern flags) but it was only available in a handful of markets which may or may not have overlapped markets where we generated revenue.

So, once we’ve selected JTBD, and then narrowed them down to the jobs we can serve with data, they need to actually occur for companies that people care about OR be sufficiently interesting that they are worth creating because they are highly impactful. There’s no point having the world’s greatest update, but it only covers 10 companies in Malaysia where we have like 10,000 free users and 0 paid.

A note on “highly impactful” (which is a judgement call). Auditor going concern is a great example, as it’s a subset of our debt analysis that only covers maybe 100 companies a year. The vast majority of these are mining explorers or biotech speculators, but every once in a while it will cover highly indebted large cap like Ford or Tesla and is sufficiently important – “this company you have invested in might not exist next year” to notify people about.

A note on JTBD as this is important to understand:

Jobs to be done: Adam Hejl and Steve Jocum ran a JTBD research project where they interviewed >100 customers of all levels of experience and surveyed around 13,000 users to collect all of the possible “jobs” that someone could have when investing. There ended up being something like >400 jobs which clustered into five main categories. I won’t share these as it’s highly proprietary, but there was a category called “Monitoring” which revolved around users keeping track of companies they already owned, and this was the category we weighed most heavily for the Dashboard.

(Humblebrag: Using reciprocity principles, I wrote the best-converting text to convince users to complete the survey, which improved conversion by about 50%.)

After these steps we had the following information:

• Important jobs for users, that
• Could be served with data, and
• Covered companies that people cared about,
• At least somewhat frequently (>=2x a year)

While it’s not possible to know all events that will happen in the future, we did some sanity checks based on recent history to confirm that after we’d built the updates, >90% of all users and almost 100% of paid users would have at least 1 update for each of the companies in their portfolio currently. (There were a variety of limitations to this analysis, so the actual numbers were likely significantly lower, but we also knew we could improve this over time).

The last remaining step was the judgement step of determining what was actually relevant for moving the stock price.

Judgement – what updates are worth your precious time?

Simple example: When a company reports results, its earnings either go up, down, or are flat. Up and down are much more important than flat, but is flat still sufficiently interesting to notify somebody about? (It is, but we would test this).

To give a more complicated example: One director leaving is not interesting. Unless that director is the CEO, when it is very interesting. What about if it’s only a single director, but he was the chair of the audit committee? Two directors is probably not interesting. How about three? Four and more definitely is. Do people care if the board diversity changed (e.g. by adding a woman)? What about if a CFO leaves, but it’s the second CFO in a year? Maybe we should draw attention to frequent turnover. There are hundreds of possibles like this for most of the main jobs and it takes time to work through them.

In the end we settled on the following main updates to address user monitoring jobs:

• Company reported earnings
• Key personnel change
• Key persons buy/sell shares
• Meaningful analyst estimate change
• Change in estimated valuation
• Change in risk

Each of these had a range of variants, for example sometimes executives notify of an intent to sell stock before they sell. Sometimes companies report quarterly earnings but not half year, or only report annually.

…………….More to come later on:

• having decided on what updates to build, how we created mockups and tested with users
• deciding how to deliver (e.g. building in phases/ iterated delivery)
• managing timeframes
• measuring impact
• training the team to deliver
• people management & quality control throughout this process

 case study simply wall st