Why I sold now:

I sold all of my holding yesterday after the announcement of the downgrade expected in the 1HFY22 results and rest of FY outlook. Group revenue down 7%, with a weak global division, "underlying EBITDA" down to only $8.5 mil (69% down) and worst of all, NPAT loss of $9.4 mil. The company made it clear the outlook for the rest of FY22 is uncertain and are expecting lower revenue and profits than previously thought. The only glimmer of hope someone could take from the announcement is the continued growth of the China segment with first half revenue up 141% PCP to $18mil, however, if you actually compare this to H2FY21, the result is only up from $17.2 mil so half on half a terrible result considering this segment was looking to be a real growth engine. I get a strong feeling results will get worse again before they get better in the future. Note all $ figures in USD.

My thesis was based on a turnaround from the bump in the road that was COVID for Appen, with the potential for growth to be restarted after the initial COVID hit. This turned out to be incorrect with the opposite occurring, results are continuing to degrade with no positive move in sight. Assuming that the old results could be a baseline was incorrect. I would currently give a valuation/target price of 15x times a $25mil profit for FY23 + $50 mil cash. This gives a valuation of $3.44 a share. This valuation is optimistic at this point but is still below the current market price. After previous updates I could still justify holding (obviously wrongly now) with a lower personal valuation, but those new valuations were still above the market price at the time. Needing to lower my valuation should have been another warning sign. 

Post Mortem - Summary of mistakes and lessons learnt from owning Appen:

• The first day you bought, you bought because they were up a few percent and a recent high. One day doesn't show any momentum. Over the next month the shares were down 35% at one point. There was no momentum. Always buy with some real momentum, otherwise wait.
• You bought when you knew results were imminent. Always wait for results to confirm your expectations. It is all about risk adjusted returns. The risk of unexpected results is much higher at this point in time than the likely returns from jumping in early in the hope of a quick buck from better-than-expected results. Ie, I would rather miss the first 20% of a multi bagger potential company than jumping in early in the hope of a quick jump but turns out, results are poor. 
• Compounding the point above, I bought a smaller batch again before results, this was after some management guidance was released because I thought the market was wrong and now there was great value because the shares were down 25%, I hadn't changed my view and management had confirmed my view. Don't catch a falling knife...
• The momentum of expectations, results and the company in general always continue to get worse and worse than expected. Very rarely are bad results and expectations turned around quickly unless there is an obvious factor. There was a bit of hope in this thesis (I guess every thesis really requires some hope otherwise you can't invest). I did not appreciate how hard it is to change momentum, not share price momentum but the momentum of everything to do with a company including the time it takes for changes made by a company to take effect. As a result of the holding Appen I have appreciated having a watchlist and taking my time sitting on the sidelines waiting and watching. At some point the market might give you the opportunity you are looking for when the outlook and momentum becomes positive then is the time to buy. If this is at a higher price so be it, you will receive better returns due to reducing the risk of permanent capital loss from a poor decision. In many cases you could be picking up a company on the upswing from the lows, therefore, maximising returns. 
• No clear buying plan was set out at the start. Potentially shows how important this step can be because it acts as a guide and limits the impulse buy due to changes/announcements. A strong justification would be required to change the original plan. I didn't write down any notes on the second parcel of shares bought, this purchase was just after the initial purchase in the same month and still before results! A buying plan is a must do in the process now.
• Buying a full position must be done over a decent amount of time, generally at least 6 months for longer-term holds. You went to a full position very quickly. This was always viewed as a long-term buy not a shorter-term trade. Therefore, there was no rush to create a full position so quickly. If I were following the new rules I have implemented, the losses would have been much lower. The importance of buying over time is that you learn more about the company over the time you own and watch the company. What you originally thought you knew about company changes and your conviction should change accordingly. 
• In my original thesis notes it was time to sell if revenue after the initial covid impacts didn't return to 20%. While it was maybe fair to say covid impacts lasted longer than expected, if this exit point was followed more closely, my capital could have been saved.
• Hard and painful lessons learnt. Many of the points above have already been integrated into my processes and checklists to stop the same mistakes from occurring again.
• Market sediment/opinion does matter. It is not just about what your convictions and belief of the fundamentals are. You need to consider the market and not think you are smarter than it, to a point of course. 
• When there is negative momentum and deteriorating fundamentals, it is better to just get out early and sit on the side-lines and if positivity of the story or momentum changes you can get back in. Holding because you see the potential for change or the long-term story being different isn't enough because it is too late when you find out you are wrong and capital is lost permanently. I have found when I have exited other failing positions it often provides me with better clarity as it reduces some of the bias you have when you hold a position.  
• I was too confident in the turnaround story that really had no basis besides management saying things should get better but in reality, it was disappointment after disappointment. 
• Conviction did decrease. I should consider a rule that this means cutting or trimming a position as a result. My concept is for long term holds, that over time of holding I should be becoming more confident and have a better understanding of the business. If you don't want to buy more (if you could with no consideration for portfolio weighting or available capital) then should you be selling.
• Just because you think a stock is cheap doesn't mean it won't get even cheaper! Appen is now an example of a stock where it was down 80% and halved from there! Ie down 90%! Luckily, I didn't get in at that point!
• Don't believe management words but the undertones of the message. When a downgrade is given it tends to end up worse. This was a great example of the need to not listen to closely to the positive messages of management, about how they plan for things to get better. Rather read the underlying message and take a cynical view of management comments.

Lessons Learnt, Thesis Review and Comments on Appen

Some expensive lessons learned the hard way while holding Appen:

• There is no need to jump on a company just before results. Unless it is clear they will be positive. There was a highly likely possibility when I entered APX that results for FY20 could be negative and they were...
• So what did I do? Buy some more... If I liked it at $25, why wouldn't I like it at $18? Lesson learnt - don't catch the falling knife! Why enter a stock that has negative sediment and negative outlook on results even if I believe longer-term it is worth more than the current price? Things almost always normally get even worse before they get better. The negative momentum continues for much longer than you think. It makes it extremely hard to tell if you are wrong or it is just bad sediment is pushing the price down?
• This report should have been the time when I added Appen to the watchlist ready to jump in after the next report when they have two positive half-year results that match with the thesis. Expectations would have been reined in after a bit more knowledge watching the company. I think it will be an awfully long time before I see my original valuation of $32 again.
• It takes time for the turn around to happen... So why not wait for the actual turn around in case it doesn't come or takes too long?
• If most of the brokers are downgrading then good luck fighting that sediment... Brokers prices tend to be optimistic as well, so you got to take a factor off their target price. These are normally 1-year targets, so you must take some short-term pain as a minimum.
• One very strong positive day does not mean momentum has changed. Momentum changes over time, one day is just noise/reaction.

What has changed to my approach to investing as a result:

• Need to have positive momentum/sediment to make a purchase. There is no point buying otherwise. The positivity needs to be in the company’s results and the market sediment. I have been interested in buying Magellan (MFG) but haven't seen any positive news to even think about deep dive researching it yet. This has paid off as MFG is down significantly since I did a quick review to put it on the watchlist.
• In the book "Taming the Lion" by Richard Farleigh, he notes you should never buy unless the trend has begun, you may not get the lowest price but you have reduced the risk of loss as you are not fighting potential negative momentum. QAV's basic three-point trend-line is another example of this. I can't see a valid argument against this approach, yes, you may miss the maximum outcome but risk-to-reward ratio is swayed very strongly to the reward side comparatively. I am investing to make multi-bag returns. What's the issue with missing the first 10-20% to earn 200%+ and minimise potential losses on the way there?
• You need to sell if all signs are negative but still within the thesis. You can always jump back in. There was no guarantee 6 months ago the recovery may be starting in 2H as it did. When I have done this before with AVA, DUG, RBL it has been the right decision.

Onto the future of Appen and my holding:

The thesis for purchasing was based on Appen having 1-2 years of pain before re-joining the previous growth trajectory, not incorrect at this point but points out in hindsight how stupid it was to jump in at the point I did. The 2H result was positive with the resumption of growth, additionally the YTD revenue and booked work well ahead of the PCP. This is what I should have waited for.

Problem is, as I have learnt the lessons above, I have been holding APX shares. Do I sell out now when I see some positive company results which are aligned with my original thesis because the market sediment is still negative? I wouldn't be buying at the current point in time but it seems too cheap to sell and my thesis hasn't been proved wrong yet (the thesis revolved around picking up a downtrodden stock and holding long-term). Therefore, I will continue to hold but on a serious watch, ready to dump if any company results aren't positive.

In terms of the company, I still think Appen as a significant role to play in the AI space. Appen admits the big 5 tech giant's revenues will reduce proportionally over time. Management has a plan to diversify the business. Non-tech-giant companies (Boeing, Adobe, Siemens and Home Depot are examples of companies that use Appen's services) have unique data that will need to be labelled for their own applications of AI/ML, data-labelling as a service is Appen's core business.

AI/ML needs a high-quality training dataset, human verification of datasets must be a part of this loop to ensure accuracy. Without high quality data, as with all algorithms, rubbish in = rubbish out. For example, look at this meme video of Elon Musk promising self-driving cars every year since 2014, even with all the data collected so far by Tesla, they haven't produced a publicly available autonomous vehicle. AI/ML is hard to perfect!

I think the world is yet to discovery how many ways AI/ML can be used to improve productivity/profitability of businesses, the big tech giants have been integrating these systems for a while now but AI/ML is becoming a focus area of other enterprises, as shown by the large increases in AI/ML spending predicted into the future.

The "China" business (which is expanding into Japan and Korea) is a new high growth segment of the business which is performing well. The recent capex investments in new products, improving automation/productivity and restructuring do need to start showing results through increased revenue. The lack of capex previously was something missed in the initial thesis as a flag. Upon noticing this, I am not surprised over the past couple of years that growth slowed, compounded by COVID. Appen is in the innovation space so needs to continue to change over time, just like it has previously, moving away from its origins as a company that developed linguistic technologies.

FY21 Results Notes

General Notes

• Management aspirations for FY26:
• Double FY21 revenue.
• 1/3 Revenue from non-global customers.
• EBITDA margin of 20%.
• How management plans to meet above aspirations:
• Customer aligned organisation structure.
• Increasing investment in product and engineering to up to 10% of revenue.
• Team of data scientists to build and deploy machine learning models to pre-labelled training data.
• Quadrant unlocking point of interest market.
• Key financials:
• Revenue = $447.8m, up 8%.
• Statutory EBITDA = $72.9m, down 2%.
• Statutory NPAT = $28.5m, down 20%
• "Underlying" NPAT = $40.6m, down 10%.
• Dividend = 10c (AUD) - entire year. 50% franked
• Investment in product development of $30.2m.
• Use of underlying NPAT is roughly fair. Amortisation of acquisition-rated intangibles is a significant contributor to the difference between underlying and statutory results. $1.6m in restructure costs I would add back, companies are always restructuring.
• Underlying diluted EPS = 32.53c (US).
• Cash balance of $47.9m (no debt).
• "Free cash flow" (Operating activities - payments for intangibles - plant - lease) = 53.9 - 21.8 - 1.3 - 4.9 = $25.9m. Noting receivables are up $38m which may be the reason for the low number here but also contract assets down $21m, moved to receivables, therefore net $19m. Working capital increased $17.6m. Should see a pickup in operating cash flows next half.
• New customers included Salesforce, Boeing, Adobe and Bloomberg.
• New senior managers in 2H21: Chief Transformation Officer, SVP and GM of Enterprise, SVP and GM of Quadrant (Quadrant founder) and Chief Product Officer (from Microsoft). Good time for some change.
• Roadmap notes the expansion of the China business into Japan and Korea.
• Looking back at financials, Appen appears to have not spent much money on internal capex prior to 2019. Acquisitions was the main "capex" prior to 2019. This takes a few years to start filtering through. Development spend is: "increasing the range of products with more pre-labelled data sets, automated data labelling and model evaluation products".
• Product development is amortised over 3 years.

Positives

• 2H revenues were very strong compared to previous halves at $250m. Previous largest half was $212m. Global services revenue up 32% on 1H. Higher proportion of non-ad related projects.
• China revenue up 442% to $24.7m. This business is expanding into Japan and Korea.
• Revenue order book YTD plus orders in hand approx $190m this was $165m in PCP.

Negatives

• EBITDA guidance was missed and no guidance given. This doesn't provide any confidence to the market.
• Most risks in the risk section of the report make note of an increase or equal level or risk going forward. Not a positive sign for the future.
• Risks note increasing pressure from competitors.

Has the thesis been broken?

• On serious watch. Not a buy but not a sell yet based on thesis expecting weakness during COVID. Revenues need to start a strong upwards trajectory from here. See overall notes above.
• Note: there is no denying I might just by putting my head in the sand again...

Valuation

• Ongoing NPAT = $55m, PE target = 20, Cash = $66m. Price target = $9.48. Conservative PE ratio used due to previously very poor market sediment. If APX grows valuation can be rerated. All calculations in AUD.

What are you expecting and what do you need to see over the next reporting season or generally into the future?

• Revenues must continue to increase and beat previous PCP.
• Cash flows increase to approximately match underlying NPAT (smoothing out over time required).
• Capex investments must start to show some improvements in terms of revenue and/or margins.
• Given Appen is capitalising the majority of the product development, EBITDA is not a fair measure of profitability going forward. Need to look at cash flows and profitability (NPAT/EBIT).
• Management "promises":
• Target of 35% growth from non-global customers.

Note all figures quoted in USD unless stated. Fiscal years are calendar years.

General Notes/Neutral outcome:

• Revenue up 12% to $600 mil. Relevance up 15% while speech & image down 10%.
• Underlying EBITDA $108.6m (+8%)
• 136 new customers.
• Guidance is positive with 18-28% growth but I don't think this is reliable.
• China and government are focus growth areas.
• 34% increase in projects with existing major customers. However, this is a result of a greater number of smaller projects.
• Company noted face-to face sales and customer engagement was down due to COVID restrictions. This is potentially critical to COVID recovery for Appen. They will be able to personally meet and sell their product again over the next year. Customers will be able to sell their product which uses Appen's datasets. To me this point is a strong reason as to why revenue has not matched previous expectations. The need to be physically present to make sales is underestimated in my opinion. Engagement will need to improve, I think this will happen naturally as companies can begin to reduce their focus on core business and look for opportunities.

Positive:

• Profit up 21.4% to $50.5mil
• Customer wins in the Q4 was strong after declining Q1-Q3.
• China growth.
• Around US$100m of ACV at the end of CY20. However, as of 1 Feb 2021 this had risen to $124.4mil. Is this customers moving to committed revenue over projects or a bullish sign for H1CY21?
• Relevance is the main driver of the business.
• Strong employee engagement.
• Overall cash flow positive with $78.4 mil cash on hand and no debt. Important to make it through the current downturn.
• Cash flow from operating activities up 39% to $93.5mil

Negatives:

• Revenue growth below long-term requirement for thesis.
• Speech & image is going backwards and growth isn't as spectacular as relevance. Still producing positive EBITDA.
• Appen now quoting annual growth rate of AI industry at 24%, I think this figure was previously 28%.
• Brokers downgrading due to the poor result causing downward price action.
• Current figures for revenue so far might be a bit weak. Purposefully hazy on year-to-date revenue whether its better or worse I can’t tell?

Has the thesis been broken?

• No, thesis was based on purchasing Appen at a weak point. The thesis requires growth to restart over the next 1-2 years. Will need to see signs of improvement over the next year. 
• Thesis is based on this point in time being a temporary stall rather than a negative inflection point. If a negative inflection point (lower growth in revenue/profitability) becomes evident = sell.

Appen at current prices appears to be a great opportunity with strong tailwinds behind it for at least the next 10 years. AI is one of Ark Invest's 5 innovative disruptive technologies. However, the investment case requires continual growth at 25%+ which will taper down to 20% by 2030. These are large numbers for a company already creating $500+ million in revenue.

Appen has a nice segment of the market. I found a figure that projects AI data will be 10% of all AI spending. Given Appen only has one other major competitor that I can find this gives Appen the opportunity to realise the revenue figures quoted above.

Customer value proposition:

Appen's value proposition to it's customers is the removal of project risks and provides price certainty. AI requires large and accurate datasets to train AI. For the customer to collect, process and use the data this expands the internal scope of the project. Questions customers have to ask themselves is: how many people to hire, who are the experts and how to hire them short time, what to do with them after collecting the data if it is a short project, how many working hours are required, do we have the correct systems to create and store the data, will  we do this right? By purchasing the AI training data from Appen, all of these risks are removed and the cost is known at the start. The time saving alone could be well worth it in the fast moving tech world.  To use the popular jargon I see Appen as an "AI data/training as a service or AI data/training on demand" company.

Positives:

• Appen is a growth company that is strong cashflow positive and profitable.
• Net cash
• Strong momentum with growing revenues and profitability. 5 year ROE = 21% CAGR
• "Picks and shovels of AI"
• Management has significant skin in the game:
  • Chairman (founder) owns 7.61%
  • CEO and Bill Pulver have high 7 figures of wealth in the company (at current prices)
  • All other directors have at least $1m invested
• MD pay in comparison to other companies:
  • Base salary is 8th percentile
  • STI is 33rd percentile
  • LIT is 80th percentile
  • Show me the incentive I'll show you the outcome... MD is paid lower for short term and higher for long term outcomes when compared to other similar companies.
• Unless the case was for a strong bear case. I am yet to see a price estimate or valuation under $20. No sell analyst ratings only buy and hold.
• Strong network effect having built up over 1 million workers to assist in the processing of data. This would take time to reproduce.
• China expansion with strong protections of Appen IP ensured. This part of the business is only just getting started. China doesn't appear far of the US in terms of AI tech. A potential area for even more growth.
• Long term chart - bottom left to top right. Any graph of revenue, profits, cash flows is the same. A great indictor of a company that exceeds in my view. Winners keep winning.
• AI datasets need continued training. Over time the dataset is outdated or degrades.
• Price appears to have a strong floor at around $20-22 since Feb 19. Dipped well below a long term 3 point trend line (on log scale).
• Industry growth expected to be around 28% a year.

Negatives:

• Concentration of customers - Top 5 clients provide 88% of revenues.
• Increasing AUD/USD will reduce profitability
• Recent announcements of weak figures. Is this a one off?

Risks:

• Big tech giants take AI data work in house due to the amount of data they require it may become more economical or this puts pressure on margins.
• Growth stops and is not as strong as predicted.
• I can't find many negative reports on Appen therefore research and thesis could be based on confirmation bias.
• Current figures are weak. Company says this is COVID related. What if this is not a one off as widely believed.
• My figures are way to optimisitic.
• Losses to other competitors.

When to get out:

• When growth stops or significantly reduced.
• Share prices in significant gains for a long time and doesn't discount any risk.
• Margins are reduced causing a substantial change in valuation that no longer justifies the risk. Appen has had margins in a tight range historically.

Expected outcomes:

• 30% Worst case - growth stops and Appen can only maintain profitability at current levels. No future growth from here. I use this as a base case given the expected growth in field. PE of 15 + cash = $8.6 per share
• 60% Expectated outcome - 2030 PE of 30, Revenue growth 25% from 2021 to 2025 then taper to 20%. NPAT = 11% and discounted back at 15% = $32.44 per share.
• 10% Best case: Same as above but 30% p.a revenue growth for every year to 2030. $82 per share.
• Overall approximately a 60% downside risk for a 10 year return in the order of 20% p.a. compounding. Target price based on probabilities = $30.22

I see Appen as a strong "value" growth buy. I have been sitting on the fence due to the weak EBITDA expectation. Why only release the expected EBITDA figure? Are the revenue and profit figures worse?  I understand the companies explanation. At a granular level my guess is teams are all working from home in the US and its probably getting to the point where teamwork is dropping off and new projects are less important than keeping to your known strengths.

I will be looking to buy in small chucks to build a position given the current outlook but don't won't hold off if others start to realise the value I see here.

I would be very interested in a forum conversation with anyone who would like to point out what I have got wrong with this thesis! As mentioned I couldn't find many negatives...