Capital Markets CRM: What a Sell-Side Desk Actually Needs

A capital markets CRM is a client relationship system built for the sell side — for equity research, institutional sales, sales trading, corporate access and investment banking — where the unit of value is a continuously covered client relationship rather than a deal that opens and closes. 

It differs from general-purpose CRM in what it holds. Every CRM holds contacts and opportunities. A sell-side desk also needs research readership, trading flow, holdings, interactions, corporate access meetings, entitlements and client feedback, all joined to a single client record that every desk reads from. 

That last clause is the entire argument, and it is the part that is hard. 

Why generic CRM breaks on a sell-side desk 

Horizontal CRM is built around a pipeline. A lead becomes an opportunity, the opportunity has a stage and a close date, and when it closes the system records a win and starts the next one. Every object in the data model exists to serve that arc. 

A sell-side client relationship has no such arc. There is no close. There is coverage — continuous, multi-desk, running for years, with dozens of touchpoints in a week rather than a decision every quarter. The client does not sign anything. Revenue does not arrive as a booked win; it arrives as commission the client allocates, after the fact, based on their own assessment of the value they received across research, sales, trading and corporate access. 

Three consequences follow, and each one breaks something structural in a generic system. 

Attribution runs backwards. On a sell-side desk you do not know what you earned from what until the client tells you — through the broker vote, or through allocations that reflect it. That means every interaction has to be captured in a form that can be re-scored months later against a revenue outcome nobody knew at the time. A pipeline CRM captures activity as a step toward a known goal. The sell side needs activity captured as evidence for a judgment that has not been made yet. 

The entity model is wrong. Generic CRM models Account, Contact, Opportunity. The sell side needs the firm, the funds or strategies inside it, the individuals, and the mapping between them — with the same person covered simultaneously by a research analyst, a salesperson and a sales trader, each with a different view of the relationship and a different claim on it. Flatten that into Account-and-Contact and you lose the structure that makes coverage decisions possible. 

Data entry cost decides adoption. A salesperson has fifteen client conversations before lunch. Any system that asks them to log those by hand will not be used, and a CRM nobody uses is worse than none, because now the reports are confidently wrong. This is why capture has to be automatic — pulled from email, calendar, chat, phone systems, entitlement platforms and trading records rather than typed. Adoption on a sell-side desk is not a training problem. It is a plumbing problem. 

What a single client record has to contain 

“Single client record” is used loosely enough by enough vendors that it has stopped meaning anything. Here is the concrete version. To be useful on a front-office desk, one client object has to carry all of the following, joined and current: 

• Readership. Which notes this client opened, which analysts they read, how long they spent, what they downloaded, and what they are entitled to receive in the first place. 

• Interactions. Calls, meetings, emails, chat, conference attendance, roadshow meetings, corporate access sessions — captured automatically, attributed to the right person and the right desk, with the desk that owns the relationship visible on the record. 

• Trading flow. What they traded, with which desk, direction, size relative to their normal activity, and how that has moved over time. 

• Holdings. Current and historical positions from filings and third-party data, so that a research read or a corporate access request can be interpreted against what the client actually owns. 

• Corporate access history. Which corporates they met, when, through which event, and what they said afterwards. 

• Feedback and ratings. Broker vote inputs, client feedback, and the qualitative record of what this client values from the franchise. 

• Revenue. Allocated commission and the path back from it to the activity that earned it. 

• Entitlements and permissions. What this client is allowed to receive, and — separately — what each internal user is allowed to see, which is not a preference setting on a sell-side desk but a control. 

Nothing in that list is exotic. Every firm has all of it. The difficulty is never acquisition; it is the join. 

Each of those datasets lives in a different system, updates on a different clock, and identifies the client differently — the entitlement platform’s identifier is not the order management system’s, which is not the CRM’s, which is not the one the corporate access team uses. Reconciling them is work, and because it is work, most firms do it periodically rather than continuously. That periodicity is the actual product problem. A picture assembled weekly is a record. A picture assembled continuously is something a person can act on. 

What each desk needs from the same record 

The reason this has to be one object rather than five well-integrated ones is that the desks do not consume it separately. They consume each other’s data. 

Equity research needs to know who is reading, how deeply, and whether readership is concentrating or drifting — and needs that joined to trading and holdings, because readership on its own says interest, not value. An analyst preparing a marketing schedule is making a resource allocation decision, and the input to it is the whole relationship, not the download log. 

Institutional sales needs the full client context before the call: what they read this week, what they hold, what they traded, what they asked corporate access for, what they said last time and who else at the firm has spoken to them since. This is the desk where a fragmented record costs the most, because the salesperson is expected to be the person who knows everything about the relationship, and is usually assembling that knowledge across several platforms in the minutes before dialling. 

Sales trading needs current client context at the moment of pricing — the client’s recent flow, their axes, how this order compares to their normal size. Yesterday’s context is not a degraded version of today’s. On a block it is a different number. 

Corporate access needs demand signals to match investors to corporates — holdings, readership, prior meeting history, stated interest — and then needs the meeting, the attendance and the feedback to flow back into the same record, because those meetings are among the most valuable things the franchise provides and they are scored in the vote. 

Investment banking needs the institutional relationship map and deal context, under information barriers that are structural rather than cosmetic. A system that serves both the private and public sides has to enforce that separation in the data model, not with a hidden field. 

Each desk’s output is another desk’s input. That is the definition of a system that has to be one system. 

Doesn’t a capital-markets skin on a generic CRM solve this? 

This is where most evaluations go wrong, because the demo looks right. Custom objects, renamed fields, a sell-side page layout, a partner-built accelerator: the screen a salesperson sees can be made to look purpose-built in a matter of weeks. 

What cannot be skinned is the layer underneath. A configuration exercise changes what the interface displays; it does not change what the platform ingests, how often, or what it can join. If readership arrives nightly through a batch job, flow arrives from a separate warehouse on a separate schedule, and holdings come from a third source refreshed monthly, then the client record is a view stitched across three clocks — and it presents as one picture while being three pictures of different ages. 

The cost of that shows up in two places. First as latency: the full picture reaches the person who needs it after the moment to use it has passed. Second as a permanent integration bill — connectors to build, mappings to maintain, reconciliation to run, and a vendor list where every new data source is a project. “Best of breed, integrated later” is an architecture decision disguised as a procurement decision, and the later never fully arrives. 

The alternative is not more connectors. It is fewer systems that need connecting — CRM, market data, client intelligence and analytics operating on one client record rather than four products and an integration budget. That is what consolidation in capital markets technology is actually about, and it is why the consolidation conversation is an architecture conversation rather than a cost-saving one. 

How to evaluate a capital markets CRM 

Five questions that separate a purpose-built system from a configured one: 

1. What arrives automatically, and how quickly? Ask specifically which of readership, interactions, flow, holdings and corporate access history is captured without a human typing it, and what the lag is on each. The answer is the product. 

2. What is the client entity, and who resolves it? If matching identifiers across sources is the client’s job, the integration cost has been moved, not removed. 

3. Can the vote be reconstructed? Ask the system to show every input behind one client’s commission allocation over a year. If that takes a project, attribution is a spreadsheet exercise with a CRM attached. 

4. How is coverage modelled across desks? One client, several owners, different views — either the model handles that natively or the firm handles it in convention. 

5. What is real adoption? Not licences issued. Daily active users on the desk. Datazoic runs at 92% user adoption across deployed teams, and that figure is a plumbing outcome rather than a training one — people use a system that already knows things. 

The record is the precondition 

Everything a front-office team wants from its technology — better pre-call preparation, faster reaction to a client’s behaviour changing, defensible attribution, and increasingly agents that watch the franchise and surface what matters — depends on the same underlying object being complete and current. 

Anything built on top of a fragmented record inherits the fragmentation. That is worth being clear-eyed about before evaluating what gets built on top. 

—- 

Datazoic is an AI Native capital markets CRM and sell-side product suite for front-office teams at investment banks — covering equity research, institutional sales, trading, corporate access and investment banking in a single client record. See how Prism handles connected engagement tracking, what Pinsight does with contact intelligence, and how Consumption Hub reconstructs broker vote reporting.