SPA and SHA: Signing, the Price Bridge and Deal Protections

What the SPA does and what the SHA does

A share purchase agreement (SPA) and a shareholders' agreement (SHA) are often mentioned in one breath, as if they were two names for the paperwork of a deal. They are not. The SPA is the instrument of a single event: it moves ownership of shares from a seller to a buyer, fixes what is paid for them, and allocates the risk that the company is not what it was said to be. The SHA is the instrument of a continuing relationship: it governs how the people who own the company after the deal will run it together — who decides what, who may sell, what happens in deadlock. One is a transaction; the other is a constitution for the years that follow. Treating the SHA as a schedule to the SPA, or the SPA as a chapter of the SHA, is the first structural mistake, because the two documents answer to different bodies of law and fail in different ways.

This page is written to English-law practice, because the English-law SPA is the working template for cross-border private deals and its terms have been tested repeatedly in court. The mechanics travel — an enterprise-to-equity bridge and a completion-accounts true-up look much the same in most jurisdictions — but the remedies do not: whether a statement is a warranty or a representation, what "fairly disclosed" means, and whether a limitation clause is read narrowly are all questions on which the governing law decides the outcome. The tax layer of an exit — share deal versus asset deal, participation exemptions, exit relief and the taxation of deferred consideration — sits in the separate review of selling the business; what follows is the deal-mechanics layer that the tax topic does not reach.

Three features of this layer are worth stating before the detail. First, the price is a moving object until a defined date: either it is locked at a past balance-sheet date, or it is trued up after completion against what the accounts actually show, and the choice between those two shifts real money and real risk. Second, protection against a bad company is not one right but a graded set of instruments, each matched to a kind of risk — warranties for the unknown, indemnities for the known, escrow for the collectable — and each capped, floored and time-barred. Third, a foreign buyer of a business in a sensitive sector needs regulatory clearance that stands between signing and completion, and in some regimes a deal closed without it is not merely penalised but void.

Two instruments, two jobs

The SPA does its work once. It identifies the shares, states the price and the mechanism for arriving at it, lists the conditions that must be satisfied before completion, records the seller's warranties and indemnities, and sets out how disputes about all of that are limited and resolved. After completion and the release of any retained sums, the SPA is largely spent. Its afterlife is litigation: a warranty claim, an earn-out dispute, a price adjustment.

The SHA does its work continuously. It sits alongside the company's articles and governs the relationship between the shareholders: reserved matters on which no decision may be taken without a defined majority or a particular investor's consent; board composition and information rights; pre-emption on new share issues and on transfers; drag-along, so a majority can force a minority into a sale, and tag-along, so a minority can join one; provisions for deadlock and for the exit of a shareholder. Its subject is not a transfer of title but the exercise of the rights that title carries.

The distinction is not merely descriptive; it is enforced by company law. A company cannot bind itself not to exercise its statutory powers — for example, its power to alter its own articles. In Russell v Northern Bank Development Corporation Ltd [1992] 1 WLR 588 the House of Lords held that an undertaking by the company itself not to increase its capital without consent was unenforceable, because it fettered a statutory power; but the identical undertaking given by the shareholders to each other, as to how they would vote, was valid and binding between them. That is why substantive governance restrictions live in the SHA between shareholders rather than in the constitution: under the Companies Act 2006 the articles bind as a statutory contract (s.33) and can be altered by special resolution (s.21), so a protection that must survive a hostile majority is placed in a shareholders' agreement, which ordinary resolution cannot touch. An SHA is therefore not a synonym for an SPA and not an appendix to it: it is the document that decides, after the buyer has become an owner, what being an owner is worth.

Signing, conditions precedent and the gap before completion

In a simple deal signing and completion happen together: the parties sign, the money moves, the register is updated, the SPA is exchanged and completed on the same day. In most deals of any size they are separated, because something has to happen in between that is outside the parties' gift — a regulatory clearance, a third-party consent, a bank's release of security, a shareholder vote. The events that must occur before either party can be compelled to complete are the conditions precedent.

A condition precedent is not a warranty and not a covenant. It is a switch: until it is satisfied (or waived by the party entitled to insist on it), the obligation to complete does not arise, and if it becomes incapable of satisfaction by a long-stop date the SPA usually allows either party to walk away. Typical conditions precedent include antitrust or foreign-investment clearance, regulatory change-of-control approval where the target holds a licence — covered for financial firms in change of control and buying a licensed company — the consent of a counterparty under a change-of-control clause, and the accuracy of key warranties as at completion.

The gap between signing and completion carries its own risk: the business may deteriorate. Two devices address it. Interim covenants require the seller to run the business in the ordinary course and not to do listed things (declare dividends, take on debt, change key contracts) without the buyer's consent. A material adverse change (MAC) condition lets the buyer refuse to complete if something seriously damaging happens; under English practice these are read narrowly and are hard to invoke on ordinary trading downturns. The order of the calendar matters: the price mechanism, the warranty position and any clearance all attach to defined dates in this window, and a deal that closes before a mandatory clearance is obtained can be exposed to the sanction described in the national-security scenario below.

Two systems, one clause, the same answer. English law reads a MAC clause against the party invoking it. In Grupo Hotelero Urvasco SA v Carey Value Added SL [2013] EWHC 1039 (Comm) Blair J held that a material adverse change in a company's financial condition had to be shown primarily from its financial statements, had to be significant rather than temporary, and could not be founded on circumstances the party invoking the clause already knew about when it entered the contract — a standard no ordinary trading downturn meets. Delaware asks the same question and has answered it for a buyer only on extreme facts: in Akorn, Inc. v Fresenius Kabi AG (Delaware Court of Chancery, 1 October 2018, C.A. No. 2018-0300-JTL, affirmed by the Delaware Supreme Court in December 2018) a collapse in earnings combined with pervasive regulatory data-integrity failures let a buyer walk for the first time in that court's history; in AB Stable VIII LLC v MAPS Hotels and Resorts One LLC (Delaware Court of Chancery, 30 November 2020, affirmed 2021) the pandemic itself fell inside a carve-out, so there was no material adverse effect at all — but the seller's emergency response to it breached the ordinary-course covenant, and the buyer was released on that ground instead. The pattern holds in both systems: the MAC condition almost never wins, and the interim covenant often does. A buyer that wants a real exit from the gap negotiates the covenant, not the adjective.

Which clearances, and at what numbers

A condition precedent is only as good as the regime behind it, and the regimes carry hard numbers rather than judgement calls. Under the National Security and Investment Act 2021 the trigger is the acquisition of control of an entity, defined by percentage steps: a holding of shares or voting rights crossing 25%, 50% or 75%, voting rights that let the holder secure or block any class of resolution, or the acquisition of material influence over the entity's policy (s.8). Where the target operates in one of the seventeen sensitive areas specified by regulations made under s.6, notification is mandatory (s.32), completion without approval is void (s.13(1)), and the business that completes anyway faces a penalty of the higher of 5% of worldwide turnover and £10 million (s.41(1)(a)). Deals outside the mandatory sectors are not safe by default: the call-in power runs for five years after the trigger event and for six months after the Secretary of State becomes aware of it (s.2). The list is about to move — the Government's response to its consultation on the notifiable-acquisition regulations, published in March 2026, brings large water undertakers into mandatory notification, takes "off-the-shelf" artificial-intelligence systems out, and gives semiconductors and critical minerals their own categories, with the secondary legislation to be laid later in 2026; until it is made, the 2021 list governs. The substance of the regime is at national security review.

UK merger control moved as well. It is a voluntary regime — no obligation to notify, no statutory standstill — but the Competition and Markets Authority can review and unwind a completed deal, and since 1 January 2025 the target-side turnover threshold is £100 million, raised from £70 million by Schedule 4 paragraph 2(2) to the Digital Markets, Competition and Consumers Act 2024 amending s.23(1)(b) of the Enterprise Act 2002. The 25% share-of-supply test survives (s.23(3)–(4)), and a new acquirer-focused threshold, inserted by the same Schedule with effect from the same date, catches a deal where the acquirer already supplies at least 33% of goods or services of a description in the United Kingdom and has UK turnover above £350 million, provided the target has a UK nexus (s.23(4C)–(4F)). Above the EU thresholds the position is stricter still: a concentration with a Union dimension under Article 1(2) or 1(3) of Regulation (EC) No 139/2004 may not be implemented before clearance (Article 7(1)), and implementing it early is fined up to 10% of worldwide turnover (Article 14(2)).

Read the table as a calendar rather than a checklist. Each row fixes the earliest date on which completion is lawful, and the long-stop date in the SPA has to sit behind the slowest of them; a long-stop set by reference to the parties' own convenience is the commonest way a deal dies twice — once when the clearance is late, and again when the walk-away right matures before it arrives. For a target holding a financial licence the change-of-control layer is set out in change of control and buying a licensed company.

What completion actually moves

Completion is not the moment the money lands; it is the moment the register changes. Under the Companies Act 2006 a person becomes a member of a company when their name is entered in the register of members (s.112(2)), and a company may not register a transfer unless a proper instrument of transfer has been delivered to it (s.770(1)(a)) — in England and Wales a stock transfer form under the Stock Transfer Act 1963 — after which the company must register it or give the transferee reasons for refusing within two months (s.771(1)). Between execution of the transfer and entry in the register the buyer holds the beneficial interest and the seller remains the legal owner; the company is not required to look behind the register, because no notice of any trust may be entered on it (s.126). That is why a completion agenda ends with a board minute and a written-up register rather than a handshake, and why the buyer takes the share certificate, the executed transfer and the board resolution as one bundle. Stamp duty at 0.5% of the consideration is payable on the transfer, and a company should not register an instrument that is not duly stamped (Stamp Act 1891, s.17); the tax treatment of the price, the retention and the deferred element belongs to the exit review at selling the business.

One consequence is worth isolating because buyers meet it late. A share sale does not move the employment contracts anywhere: the employer is unchanged, so the Transfer of Undertakings (Protection of Employment) Regulations 2006, which transfer employees automatically on a business or service transfer (regs 3 and 4), do not engage at all. Restructure the same deal as a purchase of the business and assets and they do, together with the information and consultation duties that come with them. The choice of share deal or asset deal is usually made for tax reasons; it changes the employment position, the consent position under contracts and the price mechanism at the same time.

The same deal, in Germany. English completion mechanics tolerate an exchange of signature pages. German ones do not: under §15(3) of the GmbH-Gesetz the assignment of a share in a GmbH requires a contract recorded by a notary, and under §15(4) so does the agreement creating the obligation to assign, a defect in the latter being cured only by a notarially recorded assignment. An SPA over GmbH shares signed by exchange of PDFs is void for want of form (§125 BGB), and the buyer's position towards the company depends on entry in the shareholder list filed with the commercial register (§16(1) GmbHG) — the civil-law analogue of the register of members. The commercial architecture of a deal is portable; the completion formalities are not, and they set the signing calendar. The constitutional and registry side of a company's own life — articles, resolutions, filings and dissolution — is at the company lifecycle.

From enterprise value to the price actually paid

Enterprise value (EV) is what the business is worth as an operating whole, independent of how it happens to be financed — commonly a multiple of earnings. Equity value is what the shares are worth, and it is the equity value that the buyer pays. The route between them is the price bridge, and it exists because EV is quoted on a cash-free, debt-free basis and assumes a normal level of working capital. Real companies hold cash, carry debt and have more or less working capital than normal on any given day, so the EV is adjusted to reach the equity price (ICAEW, Completion Mechanisms, Corporate Finance Faculty guideline).

The adjustments are: add cash, deduct financial debt, and adjust for the difference between actual and target working capital. Debt is deducted pound-for-pound because the buyer will have to repay it; cash is added because the buyer acquires it; working capital is normalised so that a seller cannot flatter the price by stripping out stock and collecting receivables early. The contentious part is classification — which liabilities are "debt-like" (deducted pound-for-pound) and which belong in working capital — because every item argued across that line moves the price directly.

The worked example below uses invented figures to show the arithmetic; the numbers are illustrative, not a market benchmark.

The point the bridge makes is the one stated at the top: a business valued at an EV of £120m changes hands for an equity price of £95m, because net debt of £22m and a £3m working-capital shortfall are the buyer's to bear. Enterprise value is a description of the business; it is not the cheque.

Locked box or completion accounts

There are two ways to settle the bridge, and they differ in when the price is fixed and who carries the trading risk in between.

Under completion accounts, an estimated price is paid at completion and then trued up: after completion the buyer draws up accounts as at the completion date, the actual cash, debt and working capital are measured, and the price is adjusted pound-for-pound against the estimates, with disputes referred to an independent accountant. The economic risk and reward of the business pass to the buyer at completion. This mechanism is accurate but slow and dispute-prone, because the adjustment is settled after the buyer already controls the books.

The accountant's answer is nearly final. The independent accountant who resolves a completion-accounts dispute is normally appointed as an expert rather than an arbitrator, and English law treats an expert's determination as binding even if it is wrong. In Jones v Sherwood Computer Services plc [1992] 1 WLR 277 the Court of Appeal held that where the expert has done the thing he was instructed to do, his figure stands notwithstanding a mistake, because the parties contracted for his answer rather than for the right one; Veba Oil Supply & Trading GmbH v Petrotrade Inc [2001] EWCA Civ 1832 narrowed the escape route to a material departure from instructions, materiality being assessed without asking whether the departure changed the outcome. A completion-accounts clause therefore does two jobs at once: it sets the accounting policies and the order of priority between them, and it writes the expert's instructions — and it is the second that a losing party will have to attack.

Under the locked box, the price is fixed by reference to a historical balance sheet — the "locked box date" — and does not adjust afterwards. Economic risk and reward pass to the buyer from that past date, so the buyer gets the profits (and losses) generated since. The seller warrants the locked-box accounts and undertakes that no value has leaked out of the company to the seller side between the locked-box date and completion; "leakage" (dividends, non-arm's-length payments) is indemnified pound-for-pound, while "permitted leakage" (agreed items such as defined management fees) is carved out. The seller is often compensated for the profit earned in the gap by an agreed daily accrual, an "equity ticker" (ICAEW guideline, above; EY, Locked box vs. completion accounts).

The choice is not neutral: a locked box gives price certainty and a clean break but demands trustworthy recent accounts, while completion accounts chase accuracy at the cost of a post-completion argument the seller conducts from the outside.

Warranties, disclosure and the two measures of loss

A warranty is a contractual statement about the company — that the accounts are accurate, that there is no undisclosed litigation, that the company owns its assets. If a warranty is untrue, the buyer has a claim for breach of contract. The measure of damages is the difference between the value of the shares as they would have been had the warranty been true and their actual value — a diminution-in-value measure, subject to the ordinary contract rules of causation, remoteness and the duty to mitigate. The buyer must prove not just that the statement was wrong but that the shares were worth less because of it.

Why this matters turns on a distinction English courts enforce strictly: a warranty is not automatically a representation. In Sycamore Bidco Ltd v Breslin [2012] EWHC 3443 (Ch) Mann J held that statements labelled as warranties in the SPA were warranties only, not representations, so no claim lay under the Misrepresentation Act 1967 — which mattered because the misrepresentation measure (restoring the buyer to the pre-contract position, potentially unwinding the whole price) was far larger than the warranty measure (the value shortfall). Idemitsu Kosan Co Ltd v Sumitomo Corporation [2016] EWHC 1909 (Comm) reinforced the point: without express words making the seller "represent" as well as "warrant", and against an entire-agreement clause recording non-reliance, warranties give a contract claim and nothing more. The label is not decoration; it fixes both the cause of action and the size of the recovery.

Warranties are qualified by disclosure. The seller delivers a disclosure letter (with a data room or disclosure bundle) setting out matters that are exceptions to the warranties; anything fairly disclosed cannot found a warranty claim, because the buyer bought with knowledge of it. What "fair disclosure" requires is a matter of the words the parties chose. In Infiniteland Ltd v Artisan Contracting Ltd [2005] EWCA Civ 758 the Court of Appeal read the contractual disclosure standard as the parties had written it rather than importing a general test; in Triumph Controls UK Ltd v Primus International Holding Co [2019] EWHC 565 (TCC) the court held that, on the wording used, disclosure had to reveal the nature of a matter but not its full extent or scope, so documents deemed disclosed through the data room satisfied the standard — and damages for the surviving breach were measured as the reduction in the price attributable to the flawed projections. The lesson for a buyer is that a warehouse of documents dumped into a data room may or may not amount to fair disclosure, depending entirely on the disclosure standard the SPA sets.

The limit on a non-reliance clause

The entire-agreement and non-reliance drafting described above is powerful but not untouchable. Section 3 of the Misrepresentation Act 1967, as substituted by s.8 of the Unfair Contract Terms Act 1977, provides that a term excluding or restricting liability for a pre-contract misrepresentation, or the remedies available for it, is of no effect except in so far as it satisfies the requirement of reasonableness in s.11(1) of the 1977 Act. The acknowledgement that proved too much. In First Tower Trustees Ltd v CDS (Superstores International) Ltd [2018] EWCA Civ 1396 the Court of Appeal held that a clause by which the tenant acknowledged it had not relied on any representation was in substance an exclusion of liability for misrepresentation and so caught by s.3 — calling it a "basis clause" does not take a term outside the statute where its effect is to remove a liability that would otherwise arise — and that it failed the reasonableness test, because it would have made the pre-contract enquiry process worthless in a case where the landlord had given an inaccurate answer about a contamination problem it knew of. On a negotiated share sale between advised commercial parties such a clause will usually be held reasonable. The point is that it has to be: a buyer is not automatically shut out where the clause operates to strip the remedy for an answer the seller knew was false.

Where the buyer already knows the answer is false, the SPA has to say what follows, because English law supplies no default. Whether a buyer who discovers a breach before signing can still claim afterwards — the sandbagging question — turns on the words: Eurocopy plc v Teesdale [1992] BCLC 1067 left open that a buyer's actual knowledge might defeat a warranty claim despite a clause confining disclosure to the disclosure letter, and Infiniteland treated the answer as a matter of construing the particular agreement. A buyer that wants to claim on what it already knows buys that right expressly; a seller that wants to stop it excludes it expressly. Silence is not neutral, it is litigation.

Indemnities: pound-for-pound cover for a known risk

Where a warranty answers the unknown, an indemnity answers the known. An indemnity is a promise to reimburse the buyer, pound-for-pound, for a specified loss — typically an identified risk that due diligence has already surfaced: a pending tax assessment, a live piece of litigation, an environmental liability, a defective title. Because it is a promise to pay a defined loss rather than a statement that turned out false, an indemnity does not require the buyer to prove diminution in the value of the shares, and it is generally not subject to the mitigation and remoteness limits that constrain a warranty claim (Ashurst, Quickguide: Warranties and Indemnities). Crucially, a specific indemnity is given precisely for a matter that is known and often disclosed, so disclosure does not defeat it — the whole point is that the parties have priced and allocated a risk both can see.

This is why the two instruments are not interchangeable. A buyer who accepts a warranty where it needs an indemnity takes on the burden of proving loss and the discount of the diminution measure; a seller who gives an indemnity where a warranty would do converts a contingent, capped exposure into a direct one. Each protective term should be traceable to a particular risk: unknown and general — warranty; known and specific — indemnity.

Caps, baskets, de minimis and time limits

Seller protection is built from a standard architecture of limits, and each element answers a different concern. A cap sets the maximum aggregate liability, often expressed as a proportion of the price (a fundamental-warranty cap may be the full price; a general-warranty cap is usually lower). A basket or threshold stops small claims: the buyer cannot claim until aggregate claims exceed a floor — and the basket is either a true "excess" (only the amount above the floor is recoverable) or a "tipping" basket (once the floor is passed, the whole amount is recoverable). A de minimis excludes individual claims below a trivial figure, so the machinery is not triggered by pinpricks. Time limits bar claims not notified within a stated period — commonly shorter for general warranties and longer for tax and title.

Because these clauses cut down remedies the buyer would otherwise have, their construction is a live battleground, and English courts resolve genuine ambiguity in a limitation clause against the party relying on it. In Nobahar-Cookson v The Hut Group Ltd [2016] EWCA Civ 128 the notification clause required a claim to be served within a set period of the buyer becoming "aware of the matter". The Court of Appeal (Briggs LJ) held the phrase genuinely ambiguous and, applying a narrow construction to a clause that limited an important remedy, read it as awareness of a proper basis for a claim rather than awareness of the underlying facts — so a notice given after forensic advice was in time. The market practice of naming market percentages for caps and baskets is not reproduced here, because those figures vary by deal and sector and no single sourced benchmark governs them; what is fixed is the function of each limit and the way the courts read the words.

Two further points fix the calendar. The contractual limits are cutting down periods that would otherwise be long: the default limitation period for a claim on a simple contract is six years from breach (Limitation Act 1980, s.5), and twelve years where the SPA is executed as a deed (s.8(1)) — so a general-warranty period of eighteen or twenty-four months is a concession extracted from the buyer, not a fact of law. Tax covenants run longer for a matching reason: the corporation-tax assessment windows in Schedule 18 to the Finance Act 1998 run to four years ordinarily, six where a loss of tax is brought about carelessly and twenty where it is brought about deliberately, so a tax indemnity that expires before HMRC's own window closes leaves the buyer holding the exposure it thought it had sold back.

A notice that named nothing. The specificity a notification clause demands is a separate trap from its deadline. In Teoco Ltd v Aircom Jersey 4 Ltd [2018] EWCA Civ 23 the SPA required notice setting out reasonable details of the claim, including the grounds on which it was based and the buyer's good-faith estimate of the amount. The Court of Appeal held that the buyer's letters, which described the underlying foreign tax problems but did not identify the particular warranties and indemnities relied on, failed that requirement, and the claims were struck out as not validly notified — so a substantial claim was lost without any court ever deciding whether the warranties were true.

Escrow, holdback, earn-out

Three devices bridge the gap between a price agreed and a price safely collectable. Escrow places part of the price with a third party for a defined period, to be released to the seller unless a warranty or indemnity claim is made against it; it turns a promise to pay damages into a fund the buyer can actually reach. A holdback is the same idea kept in the buyer's hands rather than a stakeholder's, against a specific anticipated liability. Both trade seller liquidity for buyer security, and both need a clear release mechanic and a defined tax treatment, because a retention taxed as proceeds at completion is tax paid on money that may never arrive — the point developed in the tax layer at selling the business. Warranty and indemnity insurance, which can move the warranty exposure off the seller entirely, is covered there as well.

An earn-out defers part of the price and makes it contingent on the business hitting agreed post-completion targets. It is the sharpest source of later disputes, for two reasons. First, in English tax law a right to unascertainable deferred consideration is treated as a separate asset, so it is not simply "more price later". Second, and more dangerous, the buyer now controls the business whose performance sets the payment. English law does not imply a general duty on the buyer to run the business so as to maximise the earn-out; the seller's protection is the express wording it negotiated. In Porton Capital Technology Funds v 3M UK Holdings Ltd [2011] EWHC 2895 (Comm) the buyer discontinued the product on which the earn-out depended; the court examined the express consent and endeavours provisions and held that, while a party could weigh its own commercial interests, a conflicted decision-maker's exercise of a contractual discretion is open to scrutiny and cannot be arbitrary, capricious or irrational. The practical conclusion is constant: an earn-out is protected by express conduct-of-business covenants, endeavours obligations, consent rights over decisions that would gut it, and information rights — not by trust.

Forfeiture, or price adjustment? A seller whose deferred consideration is cancelled on breach will argue that the clause is a penalty and unenforceable. In Cavendish Square Holding BV v Talal El Makdessi [2015] UKSC 67 the Supreme Court reformulated the test: the question is whether a secondary obligation imposes on the contract-breaker a detriment out of all proportion to the innocent party's legitimate interest in enforcing the primary obligation. On the facts, clauses that withheld the unpaid instalments of the price and gave the buyer a call option over the seller's remaining shares at a price excluding goodwill, both triggered by the seller's breach of restrictive covenants, were not penalties at all: they were price adjustments protecting the buyer's legitimate interest in the goodwill it was paying for. The same reasoning governs a bad-leaver call option in a shareholders' agreement, discussed below. What survives is a clause that measures a genuine loss of value; what falls is a clause that punishes.

Two practical points follow from the fact that a deferred price is an unsecured promise. A seller can ask for security for the earn-out or the deferred instalment — a charge over the shares sold, a guarantee from a substantial parent — which moves the seller into the position analysed in private credit and security interests instead of that of an ordinary creditor. And if the buyer fails before the deferred consideration is paid, an unsecured seller ranks with the trade creditors, while a genuine escrow held by a stakeholder is not the buyer's property to distribute at all; the ranking and recognition questions that decide how much either seller recovers belong to cross-border insolvency.

Four scenarios

Warranty and indemnity insurance, and the seller's side of the bargain

Every mechanism described so far assumes a seller who will still exist, and still be solvent, when a claim is made. Often it will not: a fund at the end of its life distributes the proceeds and winds up, a founder emigrates, twenty-three individual sellers each refuse joint and several liability. Warranty and indemnity insurance answers that problem by moving the warranty exposure off the seller and onto an insurer, which is why it is now the ordinary way an institutional seller gives a clean exit rather than an exotic one.

The usual form is a buy-side policy: the buyer is the insured and claims against the insurer as though it were suing the seller, but the policy is not a copy of the SPA. It carries its own de minimis and its own retention — an excess the buyer bears before the policy responds, which on many deals steps down after a period. It frequently contains a knowledge scrape, stripping the seller's awareness qualifiers out of the insured warranties so that what is insured is wider than what the seller actually gave. And it carries exclusions that survive negotiation: matters actually known to the deal team, matters disclosed, secondary tax liabilities and transfer-pricing adjustments, pension underfunding, fines and penalties, and forward-looking statements. Subrogation against the seller is waived except in the case of the seller's own fraud, which is what makes it possible for the SPA to be signed with a nominal seller cap of a single pound.

Because it is a contract of insurance, the policy answers to insurance law and not only to the deal. Under the Insurance Act 2015 the insured owes a duty to make a fair presentation of the risk (ss.3–7, in force for policies placed from 12 August 2016), and a breach no longer lets the insurer simply avoid the policy: the remedies in s.8 and Schedule 1 are proportionate and turn on what the insurer would have done had the presentation been fair. Section 9 prevents a representation from being converted into a warranty by a basis-of-the-contract clause, s.12 governs fraudulent claims, and s.13A — inserted by s.28 of the Enterprise Act 2016 with effect from 4 May 2017 — implies a term that sums due under the policy are paid within a reasonable time. In practice this makes the underwriting call, where the insurer questions the deal team about the diligence, a second disclosure exercise with its own legal consequences: what is said there is part of the presentation of the risk.

The mirror side: what the seller is buying. Most of this page reads from the buyer's chair, which makes the architecture look one-directional. It is not. The seller's own protections are a general cap that is a fraction of the price alongside a fundamental-warranty cap that is not; a disclosure standard satisfied by the data room; a generous disclosure letter; an anti-sandbagging clause; a sole-remedy provision routing every complaint through the warranty machinery and its limits; and control over the conduct of third-party claims, so that the buyer cannot settle a customer dispute generously at the seller's expense. The seller's exposures are the mirror image of the buyer's: an unsecured deferred price, an earn-out administered by someone else, a payment obligation guaranteed by a shell, and — where the seller rolls part of its holding into the buyer's equity instead of cashing out — a minority position governed by a shareholders' agreement it did not write. That last exposure is the subject of the next section.

Inside the SHA: pre-emption, drag, tag, deadlock and the leaver

The SPA ends; the shareholders' agreement begins, and it decides what the shares are actually worth to the person holding them. Five mechanisms carry almost all of the weight, and each has a place where it must live and a way in which it fails.

Reserved matters, and where a veto survives. A list of decisions that cannot be taken without a defined majority or a named investor's consent is the core of minority protection, and where it sits decides whether it lasts. The articles bind as a statutory contract (Companies Act 2006, s.33) but can be altered by special resolution (s.21), so a holder of 75% can rewrite them. Provisions may be entrenched under s.22, making them amendable only on stricter conditions, but entrenchment does not prevent amendment by agreement of all the members or by court order (s.22(3)). A shareholders' agreement is not vulnerable in that way at all, because it is a contract and a majority cannot vary it without the counterparty. The three votes that saved a director. The most economical illustration is Bushell v Faith [1970] AC 1099, where the articles gave a director-shareholder three votes per share on any resolution to remove him from the board. The House of Lords held the arrangement valid: the statutory right to remove a director by ordinary resolution — now s.168 of the Companies Act 2006 — says nothing about how the votes attached to shares are to be counted. Protection can therefore be engineered in the share rights, in the articles or in the agreement. What cannot be done is to make the company itself promise not to use a statutory power, which is where Russell v Northern Bank began.

Pre-emption, and the holes in it. On a new issue the Companies Act 2006 gives existing holders a statutory right of first refusal over equity securities offered for cash (s.561). The right is narrower than it looks: it does not apply where the consideration is wholly or partly non-cash (s.565) or to an allotment under an employees' share scheme (s.566), a private company may exclude it altogether in its articles (s.567), and it can be disapplied by special resolution (ss.570–571). A minority relying on the statute alone can be diluted by an issue for non-cash consideration; a minority relying on the agreement writes the protection to cover every issue, not only cash ones, and adds an anti-dilution price adjustment. On a transfer there is no statutory pre-emption at all — a right of first refusal exists only because the articles or the agreement create one, together with the valuation machinery that makes it work, and the directors' power to refuse to register a transfer must be exercised within the two months allowed by s.771.

Drag and tag. A drag-along lets holders of a defined majority compel the rest to sell on the same terms, so that a buyer can be offered 100%; a tag-along lets a minority insist on being included in a sale the majority has negotiated. The statutory analogue exists only for public takeover offers: under ss.979–982 of the Companies Act 2006 an offeror that has acquired 90% of the shares to which the offer relates may compulsorily acquire the remainder, and s.983 gives the outstanding holders a corresponding right to be bought out. Private companies have nothing of the kind, which is precisely why a drag has to be drafted. Introducing a drag after the fact. In Re Charterhouse Capital Ltd, on appeal Arbuthnott v Bonnyman [2015] EWCA Civ 536, the majority amended the articles to permit the compulsory acquisition of a dissenting minority's shares and then used the new power. The Court of Appeal upheld the amendment: the test is whether it was made bona fide in what the shareholders considered to be the interests of the company, judged objectively, so an amendment stands unless no reasonable person could consider it to be for the company's benefit, or it amounts to oppression of or discrimination against the minority. A minority that has not negotiated a tag-along or an anti-dilution protection should not assume a court will supply one.

Deadlock. A 50/50 company with no casting vote and no tie-break is a company that can stop working. The contractual answers are mechanical: a chairman's casting vote, escalation to the shareholders' principals, a put and call at a formula price, or a Russian roulette in which one party names a price at which it will either buy or sell and the other chooses which side of it to take. The statutory answers are worse for everyone. A member may petition on the ground that the company's affairs are being conducted in a manner unfairly prejudicial to members' interests (s.994), with remedies under s.996 that typically take the form of an order that the shares be bought; or a member may petition to wind the company up on the just and equitable ground (Insolvency Act 1986, s.122(1)(g)). What "unfair" is not. In Ebrahimi v Westbourne Galleries Ltd [1973] AC 360 the House of Lords held that in a small company formed on the basis of a personal relationship — a quasi-partnership — equitable considerations can make it just and equitable to wind the company up even though the majority acted entirely within its legal powers. But O'Neill v Phillips [1999] 1 WLR 1092, still the only House of Lords decision on the unfair-prejudice jurisdiction, drew the boundary: unfairness means a breach of the terms on which the parties agreed the company's affairs would be conducted, or the use of rules in a way equity would not permit, and there is no free-standing right to be bought out simply because trust has broken down. An exit that was not written into the agreement is not supplied by the statute.

Who decides an SHA dispute. An arbitration clause in a shareholders' agreement reaches further than it appears to. In Fulham Football Club (1987) Ltd v Richards [2011] EWCA Civ 855 the Court of Appeal held that an unfair-prejudice dispute is arbitrable: its subject matter is the parties' rights against each other, so an arbitrator can grant relief between them, even though he cannot wind the company up or make an order binding third parties. Where the shareholders sit in different countries, the choice between that clause and a court is the choice analysed in cross-border disputes and enforcement.

Leavers and vesting. Where sellers or founders stay on, part of the equity is usually made contingent on their staying: shares vest over time, or are subject to a call option exercisable when the holder leaves, at a price that turns on whether the departure was a good one or a bad one. A bad-leaver price at nominal value invites the argument that the clause is a penalty, and the answer is the one Cavendish gives — a clause that adjusts the price to reflect a legitimate interest, here the continued involvement the equity was granted to secure, is a primary obligation and survives, while a clause set at a level out of all proportion to that interest does not. The tax layer is separate and unforgiving: shares subject to forfeiture or to a leaver discount are restricted securities under Part 7 of the Income Tax (Earnings and Pensions) Act 2003, and the joint election that takes the future growth outside the employment-income charge must be made within fourteen days of acquisition (s.431) — the tax treatment is at selling the business. Where a holding changes hands later without a whole-company sale, the transfer restrictions in this section are what a buyer of that stake actually inherits, which is the subject of secondary share sales.

Read the third column first. Almost every failure in this table is a failure of placement rather than of drafting: a protection put where a majority can remove it, or left out of one document because it was assumed to be in another.

Remedies map

Every protective term is matched to a risk and carries a limit. Read across: this is what protects the buyer, against what, with what recovery, and what caps it.

The last two rows sit outside the purchase agreement altogether, and that is the point the map makes about the SPA's reach. Nothing in it governs how the surviving owners behave to one another afterwards, or answers for a seller that has ceased to exist. A buyer taking a minority stake through an immaculate SPA, with no reserved matters, no pre-emption and no tag-along, has excellent protection against the company being misdescribed and none against being outvoted, diluted or locked in. The relationship layer keeps working long after the transaction layer is spent: the mechanics set out above, together with the lock-up that can restrain a shareholder from selling — see lock-up periods — belong to it.