App review: Receipt Bank

It isn’t often that one of the apps that I use on a regular basis attracts a large round of funding but it happened recently with Receipt Bank, a London based started which “makes your bookkeeping, faster, easier and more efficient.” Last month, Receipt Bank received a Series B investment worth $50 million from New York based Insight Venture Partners.

Receipt Bank, which started in 2010, targets accountants, bookkeepers and small businesses. It offers them an online platform through which users can submit their invoices, receipts, and bills by taking a picture and uploading it through Receipt Bank’s mobile app (see Fig. 1), desktop app (see Fig. 2), or an email submission. Receipt Bank’s system then automatically extracts relevant data, sorts and categorises it. Apart from viewing your processed expenses online, Receipt Bank also publishes everything to the user’s accounting software of choice, FreshBooks or Xero for example.

Fig. 1 – Screenshot of Receipt Bank iOS app

 

 

Fig. 2 – The entry in Receipt Bank for one of my receipts

Given that I’ve been using Receipt Bank for a while now; instead of just reviewing existing functionality, I’ve also had a think about how I’d use a $50m war chest to further build out the Receipt Bank product:

  1. Faster! Faster! Faster! – When I started using Receipt Bank last year, I emailed the customer support team enquiring about the wait between submitting a picture of a receipt and it being “ready for export”. I got a friendly reply explaining that “we ask for a maximum of 24 hours to process items, but we are usually much faster than that.” The customer support adviser also explained that “the turnaround time also depends on the number of items waiting to be processed by the software and also their quality.” I’m sure Receipt Bank uses some form of machine-learning, algorithms to automatically interpret and categorise the key data fields from the picture of a receipt. As the field of Artificial Intelligence continues to evolve, I expect Receipt Bank to be able to – eventually – process receipts and invoices within seconds, with no need for the user to add or edit any info processed. Because I envisage machine learning to be the core driver of Receipt Bank’s proposition, I suggest spending at least half of its latest investment on AI technology and engineers specialised in machine learning.
  2. Not just tracking my bills and invoices – Yes, everybody is jumping on the chatbot wagon (and some of the results are frankly laughable). However, I do believe that if Receipt Bank can learn a sufficient amount about its customers and their spending and accounting behaviours, it will be able to provide them with tailored advice and predictions. For example, if I pay my supplier in China a fixed amount per month to keep my stock up, I’d like to ask Receipt Bank’s future “Expense Assistant” how my supplier payments will be affected if there’s massive volatility in the exchange rate between the British Pound and the Chinese Yuan. Similarly, when I look at most of today’s finance departments, the people in these teams seem to spend on matching the right payments received to the relevant invoice(s) sent out. I realise that the machine learning around multiple invoices wrapped into a single payment is easier said than done, but I don’t think it will be impossible and the $25m investment into AI (see point 1. above) should help massively.
  3. What if the days of paper bills are numbered!? – Now that I’ve effectively spent $25m on AI technology, I’ve got $25m left. The first thing I’d do with this remaining money is to prepare for scenarios where invoices or receipts are no longer issued on paper but provided orally. At the moment, capability like Alexa Expense Tracker is mostly used for personal expenses, but I do envisage a future where people use Alexa or Siri to add and track their expenses. Given that voice technology is still very much in its infancy, I suggest restricting Receipt Bank’s investment into this area to a no more than $1m.
  4. Integrate more (and please don’t forget about Asia) – If I were Receipt Bank I’d probably use about $10m of the remaining fund to enter new geographies and integrate with additional systems. For example, I like how Sage’s Pegg hooks into any expenses you record on your mobile, whether it’s via Slack, Facebook, Skype, WhatsApp, etc. I don’t know whether Receipt Bank is looking to enter the Asian market, but I feel there’s great opportunity to integrate with messenger apps like WeChat and Hike, without spending more than $2m on development and marketing. Also, integrating with payment processors, like Finsync did recently with Worldpay, is an integration avenue worth considering! 
  5. But don’t forget about the current product! – I feel Receipt bank would be remiss if it were to forget about improving its current platform, both in terms of functionality and user experience. For example, I can’t judge how well Receipt Bank does in retaining its customers, but I feel there are a number of ways in which it can make the existing product ‘work harder’ (see Fig. 3 below). In my experience, some of my proposed improvements and features shouldn’t break the bank. By spending about $1m on continuous improvements over a number of years, Receipt Bank should have at least $20m left in the bank, as a buffer for difficult times and any new opportunities that might arise during the product lifecycle.

Fig. 3 – Suggestions to make Receipt Bank’s existing product work harder:

  1. Some touches of gamification – I’d argue that the longevity of the relationship between Receipt Bank and an individual user is determined by how often the users uploads bills onto the platform. I assume that most users will most probably not view managing their expenses as fun, I think it would be good to look at ways to make the experience more fun. For example, I could get a gold star from my accountant once I’ve successfully synced my month’s expenses into my accounting system. I feel that there’s plenty of room to reinforce the current gamification elements that Receipt Bank uses. For example, the message that Receipt Bank managed to save 27 minutes of my time doesn’t really do it for me (see Fig. 4 below). Instead, the focus could be on the productivity gain that I’ve made for billable work (if I’m a freelancer for example).
  2. Better progress and status updates – Even if it does continue to take up to 24 hours. to categorise and process my expenses, it would be great if Receipt Bank could make its “in progress” status more intuitive and informative.
  3. Clearer and stronger calls to action – For example, I can see that I’m not making the best use of my Receipt Bank subscription (see Fig. 5 below). However, there are no suggestions on specific actions I can take to get more value from my Receipt Bank plan.

Fig. 4 – Screenshot my Receipt Bank usage

Fig. 5 – Screenshot of my Receipt Bank “Usage summary”

Main learning point: Having thought about Receipt Bank’s current product offering, and my understanding of their target market, I suggest investing a good chunk of the recent investment into optimising the machine learning algorithms in such a way that both processing speed and accuracy are significantly increased. By doing this, the customer profile and behavioural data generated, will create additional opportunities to further retain customers and offer adjacent products and services.

Related links for further learning:

  1. http://uk.businessinsider.com/receipt-bank-raises-50-million-from-insight-venture-partners-2017-7
  2. https://venturebeat.com/2017/07/20/receipt-bank-raises-50-million-insight-venture-partners/
  3. https://itunes.apple.com/gb/app/receipt-bank-business-expense-scanner-tracker/id418327708?mt=8
  4. https://play.google.com/store/apps/details?id=com.receiptbank.android&hl=en_GB 
  5. https://www.forbes.com/sites/bernardmarr/2017/07/07/machine-learning-artificial-intelligence-and-the-future-of-accounting/#49bb42ac2dd1
  6. https://hellopegg.io/
  7. http://uk.pcmag.com/cloud-services/87846/feature/23-must-have-alexa-skills-for-your-small-business
  8. https://www.accountingweb.co.uk/tech/accounting-software/case-study-receipt-banks-rapid-growth
  9. https://www.finextra.com/pressarticle/70263/finsync-connects-with-worldpay-us
  10. http://www.bankingtech.com/520502/symitars-episys-core-system-integrated-with-amazon-echo-baxter-cu-an-early-taker/

App review: Toutiao

Fig. 1 – Screenshot of www.toutiao.com/ homepage 

When I first heard about Toutiaou I thought it might be just another news app, this coming one from China. I learned, however, very quickly that Toutiaou is much more than just a news app; at the time of writing, Toutiao has more than 700 million users in total, with ore than 78 million users reading over 1.3 billion articles on a daily basis.

Toutiao, known officially as Jinri Toutiao, which means “Today’s Headlines”, has a large part of its rapid rise to its ability to provide its users with a highly personalised news feed. Toutiao is a mobile platform that use machine learning algorithms to recommend content to its users, based on previous content accessed by users and their interaction with the content (see Fig. 2).

Fig. 2 – Screenshot of Toutiao iOS app

I identified a number of elements that contribute to Toutiao’s success:

  1. AI and machine learning – Toutiao’s flagship value proposition to its users, having its own dedicated AI Lab in order to constantly further the development of the AI technology that underpins its platform. Toutiao’s algorithms learn from the types of content its users interact with and the way(s) in which they interact with this content. Given that Toutiao users spend on average 76 minutes per day on the app, there’s a wealth of data for Toutiao’s algorithms to learn form and to base personalisations on.
  2. Variety of content types to choose from – Toutiao enables its users to upload short videos, and Toutiao’s algorithms of will recommend selected videos to appropriate users (see Fig. 3). Last year, Ivideos on Toutiao were played 1.5 billion times per day, making Toutiao China’s largest short video platform. Users can also upload pictures, similar to Instagram or Facebook, users can share their pictures, with other users being abel to like or comment on this content (see Fig. 4).
  3. Third party integrations – Toutiao has got strategic partnerships in place with the likes of WeChat, a highly popular messaging app (see Fig. 5), and jd.com, a local online marketplace. It’s easy to see how Toutiao is following an approach whereby they’re inserting their news feed into a user’s broader ecosystem.

Main learning point: I was amazed by the scale at which Toutiao operate and the levels at which its users interact with the app. We often talk about the likes of Netflix and Spotify when it comes to personalised recommendations, but with the amount of data that Toutiao gathers, I can they can create a highly tailored content experience for their users.

Fig. 3 – Screenshot of video section on Toutiao iOS app 

Fig. 4 – Screenshot of user generated content feed on Toutiao iOS app

IMG_4954

Fig. 5 – Screenshot of Toutiao and WeChat integration on Toutiao iOS app

Related links for further learning:

  1. https://www.toutiao.com/
  2. https://www.crunchbase.com/organization/toutiao#/entity
  3. http://technode.com/2017/06/05/podcast-analyse-asia-187-toutiao-with-matthew-brennan/
  4. https://www.technologyreview.com/s/603351/the-insanely-popular-chinese-news-app-that-youve-never-heard-of/
  5. https://www.forbes.com/sites/ywang/2017/05/26/jinri-toutiao-how-chinas-11-billion-news-aggregator-is-no-fake/#24d401d64d8a
  6. https://en.wikipedia.org/wiki/Toutiao
  7. http://lab.toutiao.com/
  8. https://www.liftigniter.com/toutiao-making-headlines-machine-learning/
  9. https://techcrunch.com/2017/02/01/chinese-news-reading-app-toutiao-acquires-flipagram/
  10. https://lotusruan.wordpress.com/2016/03/20/cant-beat-giant-companies-on-wechatweibo-try-toutiao/
  11. https://www.chinainternetwatch.com/tag/toutiao/

 

 

 

App review: Grip

 

Grip is a London based startup that specialises in “smart event networking software”. That sounds like a relevant problem to solve, because don’t we all have a (secret) love-hate relationship with ‘networking’ at events!?

Yes, I’d love to meet with interesting people at events but I hate approaching people randomly.

Let’s have a closer look at how Grip is looking to solve this problem:

My quick summary of Grip (before using it) – I expect an app that uses clever algorithms to suggest relevant people to meet during events.

How does Grip explain itself in the first minute? – The Grip homepage describes the tedium involved in networking at events, with attendees often failing to make the connections they’d hoped for. Grip’s value proposition is to remove this tedium by unlocking “valuable connections at your event, saving attendees time and hard work. We use advanced algorithms to recommend the right people and present them in an easy swiping interface that your attendees will love.”

Getting started, what’s the process like? – Grip uses natural language processing to connect event attendees based on interest, needs and other things they’ve got in common. I liked Grip’s ability to tell an attendee not just who, but also why they should meet someone, in the form of Reasons To Meet.

Grip users will be able to tailor the real-time recommendations they get by setting their own matchmaking rules. I like the element of Grip not totally relying on machine learning, but also giving users the opportunity to feed their preferences into category rules into the Grip dashboard. This will influence the matchmaking engine in real-time and improve the future recommendations for event exhibitors, delegates and sponsors.

I can imagine that the data around users’ acceptance or rejection of Grip’s suggested matches, will help in further refining the app’s recommendations. This reminded me about the review that I did of THEO recently. THEO acts a ‘robo-advisor’ and uses machine learning to provide its users tailored investment advice.

Integrating the Grip API – Apart from the app, Grip have also got their own API, which makes it easier for companies to incorporate event matchmaking capability into their website or apps.

Main learning point: Grip is taking a significant problem for event attendees and exhibitors, and is using machine learning to solve this problem in a real-time and personalised fashion.

Related links for further learning:

  1. https://grip.events/handsake-event-networking/
  2. https://www.eventbrite.co.uk/blog/event-tech-adoption-at-events-ds00/
  3. https://grip.events/ai-event-matchmaking/
  4. https://grip.events/7-secrets-game-changing-event-networking/
  5. http://event-profs.com/world-first-artificially-intelligent-event-technology/
  6. https://marcabraham.com/2017/04/19/app-review-theo/
  7. https://www.eventbrite.co.uk/blog/event-tech-startups-2017-ds00/

 

App review: THEO

Fig. 1 – Screenshot of THEO – Taken from: http://fintechnews.sg/3137/roboadvisor/robo-advisory-services-asia/

I recently came across THEO, a mobile, Japanese investment service offered by Money Design. THEO acts as a ‘robo-advisor’; enabling users to invest using their smartphone, and applying machine-based learning to offer users investment suggestions. The service allows users to start investment from 100,000 JPY. By answering nine questions (see Fig. 2 below), Money Design’s proprietary robo-advisor’s algorithm selects an optimum combination from about 6,000 Exchange-Traded Funds (‘ETFs’) in about two minutes and provides discretionary investment management to the user.

Fig. 2 – Screenshot of questions asked to THEO users to create their investment profile 

The user’s answers will trigger THEO’s underlying algorithms to deliver the most optimal money management plan for the user (see Fig. 3). At this point, we’ll need to consider the artificial intelligence aspect of THEO. This is where the accuracy of the proposed plan, as generated by THEO’s algorithms, comes into play (see Fig. 3 below). As one Japanese investor commented: “I am an aggressive investor with a long timescale so I was surprised to see how conservative the allocation ended up.”

 

Fig, 3 – Screenshot of sample diagnosis results based on answering THEO’s questions

Main learning point: The key point with apps like THEO is going to be the accuracy and personal fit of the investment plan its algorithms will suggest to investors. I wonder whether any manual ‘tweaking’ is involved in assessing investment profiles and subsequent recommendations.

Related links for further learning:

  1. http://jftoday.com/THEO,+the+robo-advisory+investment+app,+exceeds+5,000+users+for+100days/
  2. https://www.bloomberg.com/news/articles/2016-07-12/hedge-fund-founder-turns-robo-adviser-for-japan-s-cash-hoarders
  3. http://fintechnews.sg/3137/roboadvisor/robo-advisory-services-asia/
  4. http://www.retirejapan.info/blog/japan-robo-advisor-theo
  5. https://theo.blue/
  6. http://www.investopedia.com/terms/e/etf.asp
  7. http://fintechnews.sg/3137/roboadvisor/robo-advisory-services-asia/
  8. http://www.theasianbanker.com/updates-and-articles/robo-advisors-poised-to-take-off
  9. http://uk.reuters.com/article/us-china-wealth-roboadvisors-idUKKCN10S2GT
  10. http://finovate.com/drivewealth-brings-robo-advisory-china-new-partnership-creditease/
  11. https://medium.com/@Mosaic_VC/trust-in-a-robo-advisor-world-62397cbe75fe
  12. http://www.wired.co.uk/article/how-ai-is-transforming-the-future-of-fintech
  13. https://en.wikipedia.org/wiki/Artificial_intelligence