Choosing sales and marketing software for accountants is usually framed as a product comparison. Feature grids, pricing tiers, review scores. That framing is why so many firms end up with a subscription nobody logs into by month four.

The useful question is different. What specific work do you want the system to carry, and who in the practice is accountable for it? A firm that wants faster response times to inbound enquiries needs something quite different from a firm running structured outbound to a list built from registry data. Both are legitimate. They are not served by the same configuration, and often not by the same product.

We build acquisition and onboarding systems for accounting firms, which means we spend a lot of time inside these platforms, migrating data out of ones that did not work and configuring the ones that do. The patterns are consistent. Firms rarely fail because they picked the wrong logo. They fail because they bought software before they had defined the process it was meant to run.

This guide sets out a selection framework: define the workflow, map the data, test integration honestly, check the compliance layer, and price the whole thing rather than the licence.

Before you look at any vendor, write down the journey an enquiry takes through your practice today. Not the idealised version. The actual one, including the parts that depend on someone remembering.

A typical firm's real journey looks something like this. Someone fills in a website form or calls. The enquiry lands in a shared inbox. A partner sees it that evening, or the next morning if it arrived after five. They reply, suggest a call, wait. If no reply comes, follow up happens or it does not, depending on how busy that week is. If the prospect books, notes live in the partner's head or in an email thread. If they sign, someone starts a manual onboarding checklist.

Turn the journey into requirements

Each break in that chain becomes a requirement. Slow first response means you need automated acknowledgement and routing. Inconsistent follow-up means you need sequenced reminders with owner assignment. Notes living in a partner's head means you need a single record per prospect that any team member can read.

Write these as statements of work, not features. "Every enquiry receives a response within fifteen minutes and is assigned to a named owner" is a requirement. "AI-powered lead scoring" is a feature that may or may not serve any requirement you actually have.

Then rank them. Three or four requirements will genuinely change your numbers. The rest are conveniences. When you sit in a demo, the sales team will steer you toward what their product does well. A ranked requirement list is the only thing that keeps the conversation on your terms.

The most expensive mistake we see is not a bad product choice. It is two products doing overlapping jobs with no agreement on which one is authoritative.

A firm signs up for a CRM to manage the sales pipeline. Separately, someone runs email campaigns from a marketing tool. The website form posts to a third place. Three months later nobody can answer a basic question: how many enquiries did we receive last month, and what happened to them.

One record per prospect

Pick one system that holds the authoritative record of every prospect and every interaction. Everything else feeds into it or reads from it. This is a governance decision more than a technical one, and it needs to be made before purchase, because it determines which product needs to be the strongest.

In practice that usually means the platform holding your pipeline should also handle your outbound messaging, or at minimum receive a complete record of it. If a prospect gets four marketing emails and then calls the office, the person answering should see those four emails on the record.

Where practice software fits

Your practice management or bookkeeping software is not a sales and marketing system, and trying to make it one usually ends badly. The natural boundary is the point of engagement. Prospects and enquiries live in the sales system. Once a client signs and is onboarded, the client record lives in practice software. The handover between those two points is where onboarding automation earns its keep, and it is worth specifying that handover explicitly during selection.

Integration is where accounting firms get burned most reliably. ICAEW's Mastering Mid-Tier Technology research identified integration as the second largest obstacle firms face when adopting specialised software, and the frustration is well documented across the profession.

The distinction that matters was put well by John Toon of Beever and Struthers, who describes a real integration as data passing between products without the user having to touch a file. On that definition, an export and import is not an integration. It is a manual task with extra steps.

Questions that separate real from marketed

• Does data pass one way or both ways, and which fields specifically?
• What is the sync frequency? Real time, hourly, or on manual trigger?
• What happens when a record exists in both systems with conflicting values?
• Is the connection native, built through an API by one of the two vendors, or dependent on a third party middleware subscription?
• If the middleware provider changes its pricing or shuts down, what breaks?

Run a proof before you commit

Ask for a trial that uses a sample of your own records rather than the vendor's demonstration dataset. Push twenty real enquiries through the full path: form submission, record creation, automated sequence, appointment booking, handover to onboarding. Integrations tend to hold up on clean examples and fail on edge cases, and your edge cases are the only ones that matter.

Where a native connection does not exist, an API can usually bridge the gap, but that is development work with a maintenance cost attached. Price it honestly rather than assuming it is a weekend job.

Any platform you use to contact prospects has to support your obligations under UK GDPR and the Privacy and Electronic Communications Regulations 2003. This is a selection criterion, and one that rarely appears on a feature comparison page.

The ICO's position is straightforward. Consent is one of six lawful bases for processing personal data, and it will not always be the right one. You must pick the basis that reflects the actual relationship and purpose. Separately, PECR requires consent for many types of marketing calls and electronic messages, as well as for most website cookies and tracking. Where PECR requires consent for a message, consent is in practice also your lawful basis under UK GDPR. Where PECR does not require it, legitimate interests may be available to you.

What to look for in the platform

• Consent capture recorded with a timestamp, the wording shown, and the source of the record
• Unsubscribe handling that applies across every channel, not per campaign
• Suppression lists that survive imports and cannot be overwritten by a bulk upload
• A cookie and tracking configuration you can actually control on your website
• Data residency and processor terms you can hand to your own compliance review

If you are running outbound to corporate contacts sourced from registry data, the rules differ from consumer marketing, and the ICO's guidance on direct marketing is the reference point rather than a vendor's blog post. Read the primary source at ico.org.uk and make your own determination.

Almost every platform now describes itself with AI language. AI-powered insights, intelligent automation, machine learning at the core. The underlying technology is frequently a set of rules with a good interface, and the gap between the claim and the delivery has commercial consequences when you have built a process around it.

Daniel Lawrence, writing on AI-washing in accounting software, makes the point that the thing a firm most needs to develop is not an understanding of AI but a clear statement of what it is trying to achieve, defined specifically enough that partners can agree whether it happened.

Useful questions in a demo

• What decision does this feature make on my behalf, and what happens when it is wrong?
• Can I see the logic, or is the output a black box?
• What data does it train on, and does my client data leave the platform?
• Can I override it, and does my override change future behaviour?

Some AI features are genuinely useful in a firm context. Automatic qualification of inbound enquiries against criteria you set is one. One of our clients uses a qualification layer on her enquiry form so that her calendar carries only opportunities worth her time, and the value there is measurable in hours reclaimed rather than in the sophistication of the model.

Judge these features by what they remove from someone's week. If a vendor cannot describe the outcome in operational terms, treat the claim as marketing and price the product on everything else it does.

The licence fee is the smallest and most visible part of the cost. The parts that catch firms out sit around it.

Costs to model before you sign

• Configuration and build time, whether internal hours or paid implementation
• Data migration from existing spreadsheets, inboxes and any incumbent system
• Training, including the second round six months later when the process has drifted
• Ongoing maintenance: someone has to own the system, and that time is real
• Usage-based charges for email volume, SMS, call minutes or contact records
• Contract length and what happens to your data at the end of it

Ownership is the variable most firms underestimate

Software with no named owner degrades. Pipeline stages stop reflecting reality, automations misfire, and within a year the team is working around the system rather than through it. Before purchase, decide who in the practice owns configuration and reporting, and how much of their week that will take. If the answer is nobody, either pick something simpler or arrange for the system to be maintained externally.

Judge on a measurable baseline

Record your current numbers before implementation: enquiries received per month, average first response time, proportion converting to a booked call, proportion converting to a client, and hours spent on onboarding admin per new client. Without that baseline, you cannot tell whether the software changed anything, and the renewal conversation becomes a matter of opinion.

This is the sequence we use when helping a firm evaluate sales and marketing software. It takes a few weeks rather than an afternoon, and it removes most of the risk.

Four patterns account for most of the failed implementations we are asked to clean up.