It was not a good week for FirstGroup. The FTSE250 transportation firm locked horns with New York based activist investor Coast Capital, with the chairman Wolfhart Hauser ousted in the skirmish.

The clash wasn’t just being watched by investors, analysts and journalists. In a distant data centre, Irithmics’ deep learning algorithms were analysing the behaviour, views and expectations of FirstGroup’s institutional investors.

Rumblings of activist hedge fund Coast Capital’s dissatisfaction were first heard several months ago. However, the first indications of real trouble for chairman Wolfhart Hauser began with renewed vigour when, on the 13th of May, the hedge fund released a statement passionately questioning the management of the FTSE250 company and outlining their demands for the replacement of six board members, including the substitution of Hauser.

Two days after the reverberating appeal from Coast for a radical management shake-up, the neural networks began recognising increasingly fragmented behaviour amongst the FTSE250’s shareholders – something had startled them.

Following this, on the 17th of May, the neural networks reclassified how investors sensitivity to news, or additional information, had changed – it had become negative. The patterns in market behaviour recognised by the artificial intelligence (AI) suggested confidence in previous views and expectations of some institutional investors appeared to be weakening as they began anticipating uncertainty. It would seem Coast had their attention.

Strike One – May 30th: Coast Capital continued to add pressure and openly attacked FirstGroup’s freshly announced breakup plans, describing them as “half baked” and “the height of folly”. The criticism could not have come at a worse time for FirstGroup – the neural networks identified a measurable shift in investors’ views of longer-term strategic allocations. Coast’s message struck a chord, and knowing they had the attention of shareholders, Coast confidently continued to present their case.

Strike Two – June 3rd: Coast scored another goal as it followed the previous hard-hitting statement with another and capitalised on the climate of uncertainty. An interview by Coast Capital’s CEO James Rasteh reiterated their doubts of the FirstGroup’s strategy and highlighted their assessment of the company’s faults.

Rasteh’s interview coincided with two important changes in the dynamics of FirstGroup’s shareholders. The first, similarly to the shift in views of longer-term strategic allocations, there was a measurable shift in investors’ views of shorter-term tactical allocations; the second and more worrying, Irithmics’ AI flagged increased vulnerability to systematic selling or shorting. While not deterministic, the increased vulnerability to selling is rarely an encouraging sign for the board of any company.

It would appear investor appetite was evaporating just as FirstGroup reacted to activist demands and announced an EGM on June 4th to decide the future of six directors.

Strike Three – June 18th: Coast Capital continued the assault, heaping further criticism on the board and gaining backing. Within days, Schroders and Columbia Threadneedle (collectively owning around 19% of FirstGroup) announced their support of Coast’s opposition to the re-election of Hauser.

Out: On Tuesday June 25th, with 30% of investors calling for his replacement, chairman Wolfhart Hauser bowed to their demands and FirstGroup announced he would not stand for re-election. However, Coast failed to convince fellow shareholders to replace the current, and in their opinion ineffective, management with its representatives. The board may breathe a sigh of relief, they might even open a bottle from the onboard refreshments trolley. However, despite Coast’s defeat and their lack of support from other investors for a complete company transformation, the clash provides insights which should prompt corporates to take note.

The use of deep neural networks to classify the dynamics and market behaviour of institutional investors reveals just how susceptible and vulnerable shareholders are to the criticism and interrogation of a company, its management and strategy. Applying this to FirstGroup shows how Coast delivered an attack with almost surgical precision timing. If it was the hedge fund’s intention to disrupt shareholders and obtain their attention, they succeeded – from the AI’s perspective, their timing was perfect.

By contrast, FirstGroup’s responses and rebuffs appear to have had little measurable effect on restoring the stability of shareholder views.

Coast may, as FirstGroup described, be an “opportunistic, self-interested player that is only focused on short-term gains”, but their understanding of the UK-listed company’s shareholder dynamics is significant. Coast described the resignation of Hauser as “a constructive first step in the direction of finally achieving full value for shareholders”. The strong words from the active investors suggest they aren’t yet finished.

Clashes between management and activist investors are likely to raise the race for mastery of shareholder dynamics. Coast has demonstrated an activist’s ability to identify when shareholders are more susceptible to criticism. They have also demonstrated a willingness to use this knowledge.

At Irithmics, the neural networks continue to recognise patterns of investor behaviour for over 14,000 listed corporates across North America, Europe and Australia. Its deep learning algorithms quietly measure, monitor and anticipate how institutional investors build portfolios, manage and mitigate risk. It continues to identify changes in the views and the stability of a corporate’s shareholder base. It continues to identify vulnerabilities. It continues to raise flags.

To learn about Irithmics Research, just get in touch.