Capability

Excel Solver in a web app, for people who do not have Excel

You built an Excel Solver model and it works. The problem is that running it means opening the workbook, and the people who need the answer are not the people who own the model. This page is about closing that gap, and about the one condition attached.

Yes, Solver runs server-side on your published model Your users click a button, the optimisation runs against your workbook on the server, and the results come back into the interface. They never see the model, never install anything, and do not need Excel at all. The algorithm uses multi-start with random seed generation, so it is searching for a global optimum rather than settling on the first local one. Documentation
One condition: not on the public cloud tier Solver recalculates your model repeatedly until it converges, which makes its cost unpredictable on shared infrastructure. It runs on Private Cloud and on server licences, both of which are quoted rather than self-serve. If that rules us out, better you know now than after a trial.

When you do not need us at all

Worth saying before anything else, because it is true for a lot of people who land on a page like this.

If you are the analyst and you have the workbook, Excel already does this

Solver ships with Excel, it is free, and for a model you run yourself a few times a month it is the right tool. Microsoft also offers a Solver add-in for Excel on the web, which sends your workbook to a cloud service to be solved, because a browser cannot do that compute on its own. There are capable third-party engines for larger problems too.

Every one of those options assumes the person solving the model has Excel and has the file. The moment that stops being true, they stop helping.

The case this page is actually about

One person owns the optimisation. Many people need its answers. That is where a workbook stops working.

The people who need the answer cannot run it

A rep sizing a bundle, an underwriter structuring a layer, a dealer configuring a machine, a customer on your website. None of them will be talked through the Solver dialog, and none of them should see your cost assumptions.

So it becomes a request queue

They email the inputs. Someone opens the workbook, runs Solver, and emails the result back. It works, and it puts a person in the middle of every optimisation you perform.

And the model leaves the building

The alternative is sending the file, which hands over the objective function, the constraint set and the cost data along with the answer. In pricing and engineering that is usually the part worth protecting.

From a working model to something the team can use

Three steps, and none of them involve rebuilding the optimisation. The second one is the only part that is specific to Solver.

  1. Upload the workbook as it is The model arrives intact. Formulas, lookup tables, the constraint calculations, all of it stays as written and keeps calculating server-side. Nothing is ported and nothing is rewritten, so there is no second version of the maths to keep in step with the original.
  2. Point Solver at what the model already contains Name the cell holding your objective and say whether to maximise it, minimise it, or drive it to a value. Name the cells Solver is allowed to change, and list the conditions it has to respect. You are describing the model, not building it.
  3. Put a button on it and decide who may press it Label the button in the language your users speak. Chain a PDF or an email to the result if they need something to send on. Then set who can reach the application: your team, named customers, or anyone with the link.

The person who owns the optimisation still owns it. When a coefficient or a constraint changes, they change it in Excel and re-upload, and everyone is optimising against the new model from that moment.

Will your model qualify? Four things to check

Most Solver models transfer without changes. These are the four conditions that occasionally need a tidy-up first, and what to do about each.

1
Your objective sits in one cell, and it is a formula The thing being optimised has to be a single cell containing a formula, not a hard-coded number and not a range. If it is spread across several cells, add one cell that combines them and optimise that.
2
The cells Solver changes are inputs, not formulas Decision variables have to be plain input cells. Each one becomes a control in the interface, which is also how a user supplies a starting point. If a variable is currently calculated, split it: one input cell for Solver, and the formula downstream of it.
3
Every variable has an upper and a lower bound Bounds keep the search honest and stop it proposing something physically impossible. They are part of the constraint list. If everything in your model is zero or above, switch on the non-negative setting and skip the lower bounds.
4
Decide where whole units matter Staff, pallets, machines and production runs usually cannot be fractional. Variables can be constrained to integers where that is true. If it does not matter, leave them continuous. It is usually faster.

The Solver event documentation has the full configuration reference. If you would rather not work through it, send us the workbook and we will tell you which of the four, if any, needs attention.

Excel Solver as an HTTP call

The part most people do not expect.

Solver can be attached to an API request

Post the inputs, and the response contains the optimised result. Another system can ask your model for the best answer without a person or a browser being involved.

Documentation

Which makes the model a service

Your pricing engine, portal or CRM calls the same optimisation the web form uses. One implementation, maintained in Excel by whoever owns the maths.

See the API

And the workbook is still the source

Change a coefficient or a constraint in the spreadsheet, re-upload, and both the interface and the API are optimising against the new model.

All 501 functions supported

The shapes of problem this suits

Not a list of industries. These are the model shapes that tend to need an optimiser in front of a non-technical user.

Hit a target, find the inputs

Reach a margin, a premium, a payment or a rating, and let the model work out which combination of levers gets there within the rules. If a single input drives it, Goal Seek is the lighter tool.

Allocate something finite

Budget across channels, capacity across products, capital across programmes, staff across shifts. Maximise the outcome subject to what you actually have.

Size or specify to a minimum cost

The cheapest configuration that still satisfies the requirement. Common in engineering, blending and packaging, where the constraint set is the domain expertise.

Optimise across many scenarios

Where the question is distributional rather than single-point, the Simulation event runs the model repeatedly instead, and can be combined with the rest.

Bring the model, not a description of it

Solver needs Private Cloud or a server licence, so there is no self-serve trial for this one. Tell us what the objective is, roughly how many variables and constraints are involved, and where the data has to sit, and we will tell you whether it fits and what it would cost. If it does not fit, we will say so.

Book a technical call Talk to us See pricing