How to run a vendor search

Editorial — Buyside Fintech's guidance

Before you look at any vendor

The most common way a selection goes wrong is starting with products. You end up comparing feature lists, every vendor has every feature, and the decision falls to whoever demonstrated best. Start instead with these four questions, and write the answers down where the whole team can see them.

1. What breaks today? Be specific. "Reconciliation is painful" is not a requirement. "Two people spend every morning until 11am clearing breaks between the administrator and our book, and month-end takes four days" is.
2. What must be true in three years? A system chosen for the firm you are now will be replaced by the firm you become. If you plan to add a strategy, a jurisdiction or an asset class, that belongs in the requirement now.
3. What will you not compromise on? Usually two or three things. Everything else is negotiable, and knowing which is which before you see a demonstration is what stops you being sold.
4. Who has to live with it? Name them. The people who will use the system daily should be in the room for the demonstrations, not shown the result.

Work out which category you are actually buying

This sounds trivial and frequently is not. Order management and execution management are different systems; portfolio accounting and fund accounting answer different questions; a fund administrator is a service and fund accounting software is a product. Buying the wrong category wastes a quarter.

The category reference defines each of the 41 categories and states, for each, what it is most often confused with. Read the one you think you need and the one it warns you about.

Build a long list, then cut it

Aim for eight to fifteen names before you cut to a shortlist. Fewer than that and you are choosing between the vendors you already knew, which is how firms end up with the market leader they never tested.

• Filter the directory by category, and by more than one category if your requirement spans them.
• Ask peers at firms of your size and strategy. Size matters more than strategy: a system built for a fifty-billion manager will not fit a two-hundred-million one, whatever the sales team says.
• Ask your administrator, custodian and auditor who they see working. They integrate with everyone and have no stake in your choice.
• Include at least one vendor you have never heard of. The reason this directory records a product's generation is that newer platforms are the hardest to find and are frequently the ones that fit a smaller firm.

Then cut to four to six on hard criteria only — asset class coverage, deployment model, jurisdiction, whether they serve firms your size. Do not cut on impressions.

Ask everyone the same questions

The single highest-value thing you can do is put the same written questions to every shortlisted vendor, and get the answers in writing before you see a demonstration. Written answers are comparable, attributable and hard to walk back. Demonstrations are not.

Our annotated question set explains what each question establishes and what a weak answer looks like. You can also run the whole process here, which sends one brief to the vendors you pick and returns the answers side by side.

Run the demonstrations properly

• Send your own scenario in advance and require them to demonstrate it. A scripted demonstration shows you the system's best half-hour, which is not the half-hour you will spend in it.
• Use your own data if they will let you, even a small extract. Systems that look identical on clean sample data diverge sharply on a real book with odd instruments in it.
• Make them show a failure. A trade booked wrong and corrected after reporting. A break that will not reconcile. A corporate action with an election. The happy path is the same everywhere; the difference is entirely in what happens on a bad day.
• Have the daily users drive for part of it, rather than watching a sales engineer.

Check references, and check the ones you were not given

Every vendor will supply three delighted clients. Speak to them, and then find someone the vendor did not nominate — through peers, your administrator, or a user group. Ask all of them three questions:

1. What did implementation actually cost, in time and in your people, against what you were told?
2. What do you know now that you wish you had asked?
3. Would you buy it again?

Before you sign

• Price the whole thing. Licence, implementation, data, integration, training, and the annual increase. Ask what the fee is in year five, in writing — the number that matters is rarely year one.
• Ask how you would leave. What you can export, in what format, how long it takes and what it costs. A vendor who has not thought about this has told you something.
• Agree what "live" means before implementation starts, in testable terms, and tie payment to it.
• Name the people. Who specifically is on your implementation, and what happens if they leave.